{ "row": "mit-ai-incident-tracker", "entity": "mit-ai-incident-tracker", "address": "mit ai incident tracker", "addresses": [ { "address": "mit ai incident tracker", "slug": "mit-ai-incident-tracker-aio-20261029", "addr_id": "ADDR-f65abd8a4bbe", "obs_id": "OBS-65034e45b99a", "axn": "AXN:0AAE.CAPTURE.🌄🐚▲🎨🌈➕" } ], "epoch": "2026-10-29", "type": "C (public entity on which the archive bears: a database; an entity of the MMRS meta-aggregator)", "surface": "Google AI Overview (expanded)", "auth": "signed out, incognito", "status": { "procedure": "EA-NEGONT-02 v0.9, #1690: equal case, ρ = extends, read from the MMRS record and the capture record; archive selected by D/R/O (no deposit admitted: D 0, R 0, O-class 3 sentences in 2 deposits read and rejected; no hop), with the MMRS record (22 claims) and the Capture Registry's findings and analysis (5 claims) ledgered as the archive's records of the entity; the field B by three origins (6 card sources for 6 cards, the own pages among them; no own page apart from the cards; 3 Wikipedia sources: no article, one mention, its reference); the composer read the ledgers, never the transcript, except where the brief asks for T's claims and the delta", "ledger": "unaudited (§3.8: one extractor per ledger); field quotes 54/54 verified by script against saved text, extracted meta descriptions or card text; archive quotes 27/27 verified verbatim against the registry, census.json and articles.py; carried as marked: MMRS-13 contested (F1, F37), MMRS-15 with a defect noted", "frozen": false, "breaches": [] }, "spec": "EA-NEGONT-02 v0.9, #1690", "objects": { "T": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/T1-mit-ai-incident-tracker-aio-20261029.txt", "sha256": "5c9ee2ce5776d86946b125350ccf4720425bc72c926a107b34077e477c78efb1", "words": 156, "cards": 6, "words_note": "one transcript at one address; words of the composition from its definition to its closer, link markup and citation markers removed, the cards' titles and snippets excluded", "claims": [ "T1-1 definition: 'an interactive tool developed by the MIT AI Risk Initiative that classifies and analyzes real-world artificial intelligence mishaps' (markers: card 1, the tracker page, and a second token)", "T1-2 data source: 'Processes and standardizes reports from the AI Incident Database using automated large language model (LLM) pipelines.'", "T1-3 classification scale: 'Evaluates incidents using a standardized 1-5 harm severity scale and multi-dimensional taxonomies.'", "T1-4 core risk domains: 'discrimination and toxicity, privacy and security, misinformation, malicious actors, human-computer interaction, socioeconomic and environmental impacts, and AI system safety'", "T1-5 national security views: 'physical security, information warfare, government functions, economic security, and societal stability'", "T1-6 interactive access: 'You can explore the data directly through the MIT AI Risk Navigator or view broader frameworks on the MIT AI Risk Initiative portal.'", "T1-7 closer: 'Would you like to search for incidents related to a specific category (such as deepfakes, autonomous vehicles, or privacy leaks) from the tracker?'" ], "register": { "register": "data/EA-WG-CAPTURES-01.json", "slug": "mit-ai-incident-tracker-aio-20261029", "obs_id": "OBS-65034e45b99a" }, "transcripts": [ { "slug": "mit-ai-incident-tracker-aio-20261029", "address": "mit ai incident tracker", "addr_id": "ADDR-f65abd8a4bbe", "obs_id": "OBS-65034e45b99a", "axn": "AXN:0AAE.CAPTURE.🌄🐚▲🎨🌈➕", "date": "2026-10-29", "surface": "Google AI Overview (expanded)", "auth": "signed out, incognito", "sha256": "5c9ee2ce5776d86946b125350ccf4720425bc72c926a107b34077e477c78efb1", "sha256_of": "the registry's 'transcript' string (utf-8)", "sha256_raw": "9fbd3079b70a9715b8e72e58591fe4ea429ab16cada389e1373de2c84b3e907d", "words": 156, "cards": 6, "card_rail": [ "1. MIT AI Risk Initiative — MIT AI Incident Tracker", "2. MIT AI Risk Initiative — MIT AI Risk Initiative", "3. MIT AI Risk Initiative — AI Incident Tracker June 2026 Update - MIT AI Risk Initiative", "4. MIT AI Risk Initiative — Incident Tracker - June 2025 Update - MIT AI Risk Repository", "5. MIT AI Risk Navigator — Incidents - MIT AI Risk Navigator", "6. MIT AI Risk Initiative — About - MIT AI Risk Initiative" ], "claims": [ "T1-1 definition: 'an interactive tool developed by the MIT AI Risk Initiative that classifies and analyzes real-world artificial intelligence mishaps' (markers: card 1, the tracker page, and a second token)", "T1-2 data source: 'Processes and standardizes reports from the AI Incident Database using automated large language model (LLM) pipelines.'", "T1-3 classification scale: 'Evaluates incidents using a standardized 1-5 harm severity scale and multi-dimensional taxonomies.'", "T1-4 core risk domains: 'discrimination and toxicity, privacy and security, misinformation, malicious actors, human-computer interaction, socioeconomic and environmental impacts, and AI system safety'", "T1-5 national security views: 'physical security, information warfare, government functions, economic security, and societal stability'", "T1-6 interactive access: 'You can explore the data directly through the MIT AI Risk Navigator or view broader frameworks on the MIT AI Risk Initiative portal.'", "T1-7 closer: 'Would you like to search for incidents related to a specific category (such as deepfakes, autonomous vehicles, or privacy leaks) from the tracker?'" ], "markers": "inline links are opaque google.com/goto tokens; the definition and the Data Source item are each marked to two tokens, one of which (CAESYgHr…sk) recurs on the definition, Data Source, Classification Scale and Core Risk Domains; the National Security item is marked to one token shared with the Classification Scale (CAESdwHr…tf); the Interactive Access item to the Navigator's and the Initiative's links; the tokens are not resolved to cards", "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/T1-mit-ai-incident-tracker-aio-20261029.txt" } ] }, "KO": { "title": "The MIT AI Incident Tracker", "genre": "knowledge object, equal case, under EA-NEGONT-02 v0.9 (#1690): the archive's sources and the field's admitted on equal terms, ontology included, in one claim space", "status": "composed 2026-10-29 under v0.9 (brief rebuild/negative-of-the-negative/v09/BRIEF-NEW-ROW-AGGREGATOR.md); not frozen; ledgers unaudited", "sentences": [ { "n": 1, "para": 0, "sense": "what it is", "text": "The MIT AI Incident Tracker is a machine classification of AI Incident Database reports by risk domain, cause, harm, severity and EU AI Act risk level, made by Arcola AI's language-model pipeline against the MIT Risk Repository's taxonomy, and filed as a database (MMRS record); the MIT AI Risk Initiative's about page calls it \"a catalogue that classifies incidents where AI has already caused harm\", developed in collaboration with the AI Incident Database, and its own page says it classifies AIID incidents \"by risk, cause, harm, severity, and other relevant dimensions\".", "claims": [ "MMRS-01", "MMRS-10", "F46", "F1" ] }, { "n": 2, "para": 0, "sense": "source", "text": "It is derived from the AI Incident Database (MMRS record): the tracker page builds on the Responsible AI Collaborative, \"who created and maintain the AI Incident Database, reviewing all submitted reports\", and sends its readers to the AIID for how incidents are defined and ingested, and the about page lists the AIID among the users of the Initiative's work, mapping real-world incidents to its taxonomy in the tracker.", "claims": [ "MMRS-18", "F8", "F50" ] }, { "n": 3, "para": 0, "sense": "purpose", "text": "Its occasion, on its own page, is that incident databases \"struggle with inconsistent structure, limiting their utility for policymaking\", and its answer is a tool to classify incidents by risk and harm severity, offered as a common frame of reference for researchers, developers, businesses, evaluators, auditors, policymakers and regulators; the MMRS record restates the aim as a more consistent structure for AIID's records \"by its own account\", and the June 2026 update grounds it in what incident reporting did for safety in aviation, healthcare and nuclear power.", "claims": [ "F4", "F5", "F12", "MMRS-03", "F30" ] }, { "n": 4, "para": 1, "sense": "method", "text": "Raw AIID reports are \"processed by Arcola AI's LLM pipeline\", which categorizes them with the MIT Risk Repository and a harm severity rating based on CSET's AI Harm Taxonomy (tracker page); the June 2026 update says the tracker uses LLMs to classify incident reports along five taxonomies, \"producing standardized data\", the Navigator that each incident is classified against shared taxonomies \"using an LLM pipeline with human spot-checking\", and the MMRS record that the classification is machine-made, its intake the AIID reports passed through Arcola AI's pipeline.", "claims": [ "F6", "F23", "F44", "MMRS-04", "MMRS-14" ] }, { "n": 5, "para": 1, "sense": "dimensions", "text": "The five taxonomies are harm severity across ten harm types, the EU AI Act risk tier, a causal taxonomy of lifecycle stage and intent, seven risk domains and twenty-four subdomains (June 2026 update); the tracker page shows each incident under the repository's causal and domain taxonomies and the EU AI Act's levels, with its severity across ten CSET-based harm categories and the classification's confidence, and the Initiative's home page lets incidents be compared by risk domain, harm, people affected, alleged developer, time or national-security impact.", "claims": [ "F24", "F9", "F10", "F20" ] }, { "n": 6, "para": 1, "sense": "dimensions", "text": "On 23 June 2025 the severity scale went from 0–10 to 1–5, its definitions disambiguated in the first person (\"I have tried to remove ambiguities\"), and the tracker added flags for ambiguities and alternative interpretations and the system's primary goal by a taxonomy based on the AIID's GMF (June 2025 update).", "claims": [ "F38", "F40" ] }, { "n": 7, "para": 1, "sense": "the national-security view", "text": "A national-security view grades each incident's impact on its primary target country in five categories, from physical security and critical infrastructure to societal stability and human rights, with the threat classified for imminence, novelty and autonomy (tracker page; June 2025 update).", "claims": [ "F11", "F39" ] }, { "n": 8, "para": 2, "sense": "operator", "text": "It is part of the MIT AI Risk Initiative (tracker page), based at MIT FutureTech, a joint initiative of MIT's Institute of the Digital Economy and CSAIL, which keeps several connected workstreams because no single dataset captures AI risk and presents its data and frameworks as authoritative (about and home pages; card 2's summary lists among its tools an expert survey and the AI Risk Repository); the MMRS record gives the operator as the Initiative \"with Arcola AI's LLM pipeline\" and the tracker as part of the MIT AI Risk Repository.", "claims": [ "F7", "F47", "F48", "F22", "F21", "MMRS-09", "MMRS-17" ] }, { "n": 9, "para": 2, "sense": "origin", "text": "Simon Mylius \"developed and leads the AI Incident Tracker\", a classification tool and dashboard now hosted by the MIT AI Risk Repository, its analysis featured on the AIID (about page), and the tracker page's team names MIT FutureTech and University of Queensland researchers beside the AIID's Sean McGregor.", "claims": [ "F49", "F18" ] }, { "n": 10, "para": 3, "sense": "the unit", "text": "The MMRS record gives the unit as the AIID report, classified by domain, cause, harm, severity and EU AI Act level; the tracker's pages count incidents (\"1,600 real-world report incidents\"), and the June 2025 update reclassified \"all incidents\" up to AIID incident ID #1116.", "claims": [ "MMRS-13", "F1", "F37", "F36" ] }, { "n": 11, "para": 3, "sense": "count", "text": "MIT's surfaces give four figures: 1,600 in the tracker page's heading and \"more than 1,400\" in its description, which the MMRS record reports as the page giving its size twice; 1,752 \"real-world AI incident reports\" in the card's summary of that page; and on the Navigator \"more than 1300\" in its card and 1,497 in its page heading on 2026-10-29, beside headline figures of \"Up -76%\" year over year and 64% caused by AI systems.", "claims": [ "F1", "F2", "MMRS-05", "F3", "F41", "F42", "F45" ] }, { "n": 12, "para": 3, "sense": "count", "text": "The MMRS record's count field (as of 2026-10-28, basis observed) carries the page's two figures, Google's card's 1,752 and the Navigator's \"more than 1300\", each with its surface, and the Capture Registry (2026-10-29) records the count \"three ways\" and that the composition gave none.", "claims": [ "MMRS-15", "CAP-1-04" ] }, { "n": 13, "para": 4, "sense": "versions", "text": "The tracker is versioned by reclassification: the June 2025 update brought it up to AIID incident #1116 and classified the full dataset again with the new iteration \"to ensure consistency\", and the June 2026 update names as its next step a more comprehensive validation study and announces that the full set of AIID reports will be reclassified with the updated model selection and prompts.", "claims": [ "F36", "F37", "F35" ] }, { "n": 14, "para": 4, "sense": "validation", "text": "That update (30 June 2026) reports a pilot: because the tracker relies on automated classification rather than manual review, two expert reviewers classified ten AIID incidents on the five taxonomies, eight LLMs were set against their consensus, and the human agreement was itself only \"Fair\" for harm severity (0.31) and \"Moderate\" for the EU AI Act level (0.56).", "claims": [ "F25", "F26", "F29" ] }, { "n": 15, "para": 4, "sense": "validation", "text": "Opus 4.6 and Kimi K2.5 met or exceeded the model in current use, Sonnet 4, on most taxonomies, the models did substantially worse on the EU AI Act level, and after prompt refinement Opus 4.6 met the human baseline on all five \"on this pilot sample\", a sample the update calls small; it concludes that LLM classification has accuracy comparable to the human baseline.", "claims": [ "F27", "F28", "F34", "F35" ] }, { "n": 16, "para": 5, "sense": "limits", "text": "The tracker page states its limits: the classification database explores \"the potential capabilities and limitations of a scalable incident analysis framework\", its AIID input rests on public and expert submissions of varying quality, voluntary reporting gives \"some degree of sampling bias\", spot-checks have improved it but \"a systematic validation study has not yet been completed\", and patterns \"should be taken as indicative and validated through further analysis\".", "claims": [ "F13", "F14", "F15", "F16", "F17" ] }, { "n": 17, "para": 5, "sense": "limits", "text": "The June 2026 update adds that LLMs can hallucinate details and show a score-range bias, the Navigator calls itself a research prototype whose data and analyses are \"preliminary and not yet validated\", and the MMRS record carries the page's caveat in its kernel, its article and its note.", "claims": [ "F33", "F43", "MMRS-02", "MMRS-06", "MMRS-16" ] }, { "n": 18, "para": 5, "sense": "reception", "text": "At its name Google's AI Overview composed the tracker \"as what it says it is, with its machine method named\", AIID as the source and LLM pipelines as the method, which are the MMRS record's derivation and intake, avoided stating a count, and left out the page's caveat of indicative patterns, voluntary reporting and no completed validation study (Capture Registry, 2026-10-29).", "claims": [ "CAP-1-01", "CAP-1-03", "CAP-1-02", "CAP-1-05" ] }, { "n": 19, "para": 5, "sense": "openness", "text": "Data from the MIT AI Risk Initiative is licensed under CC BY 4.0 (tracker page), as most of its outputs are (about page).", "claims": [ "F19", "F47" ] }, { "n": 20, "para": 6, "sense": "place", "text": "In the MMRS record it is a derived layer, \"the editor-resolved incidents of AIID, reclassified by a model\", whose admission inherits AIID's harm test and whose classifications are machine-written, so that original outputs are not preserved, and the chain from event to news report to AIID record to machine classification is \"a chain of receptions, each recorded\".", "claims": [ "MMRS-07", "MMRS-19", "MMRS-20", "MMRS-08" ] }, { "n": 21, "para": 6, "sense": "the record kept apart", "text": "The record keeps each assertion with its source and basis and the lifecycle apart from the observation that saw it (read 2026-10-28), and it rests every field on the one tracker page.", "claims": [ "MMRS-21", "MMRS-12", "MMRS-22" ] }, { "n": 22, "para": 6, "sense": "neighbours", "text": "The June 2026 update places the tracker among the incident databases it builds on, the AI Incident Database, the OECD's AI Incident Monitor, the AIAAIC Repository and the Political Deepfakes Database, adding to them a classification of each incident and a rating of its harm; the MMRS record gives it the role of harm.", "claims": [ "F31", "MMRS-11" ] }, { "n": 23, "para": 6, "sense": "uptake", "text": "Its outputs have been used to report incident trends in the 2026 World Disasters Report and Time magazine (June 2026 update).", "claims": [ "F32" ] }, { "n": 24, "para": 6, "sense": "the encyclopedia", "text": "Wikipedia has no article for the tracker, nor for the AI Risk Repository or the MIT AI Risk Initiative, and none of its articles contains the phrase \"AI Incident Tracker\"; its one mention of the work, in \"Voiceverse NFT plagiarism scandal\", says that \"The MIT AI Risk Repository catalogued the incident\" as \"economic and cultural devaluation of human effort\", referenced to the AIID's page for incident 277, which on 2026-10-29 shows the MIT classification as machine-classified under subdomain 4.3, fraud, scams and targeted manipulation, in the domain Malicious Actors & Misuse.", "claims": [ "F51", "F52", "F53", "F54" ] } ], "joins": [ { "act": "order of senses", "sentences": "1–24", "claims": [], "note": "equal case (§4.7): the entity's dependency order, field and archive claims ordered together by genealogy within each sense; neither ledger placed first as a block. What it is, its source and its purpose (¶0) precede the method and dimensions that realize them (¶1); operator and origin (¶2) follow the work they own; the unit and the count (¶3) precede the versions and the validation that change and test them (¶4); the limits, the composition's treatment of them and the licence (¶5) qualify what precedes; the place in the field, the record kept apart, the neighbours, the uptake and the encyclopedia (¶6) stand last, as relations of the entity already composed." }, { "act": "holding apart (the unit)", "sentences": [ 10 ], "claims": [ "MMRS-13", "F1", "F37", "F36" ], "note": "P_coh: the MMRS record's unit, the AIID report, and MIT's pages counting and reclassifying incidents composed each with its source; no coordinating claim" }, { "act": "holding together (counts on different surfaces)", "sentences": [ 11, 12 ], "claims": [ "F1", "F2", "F3", "F41", "F42", "F45", "MMRS-05", "MMRS-15", "CAP-1-04" ], "note": "T_lib: each figure with its surface and date; neither ledger's count lowered; the MMRS record's and the Capture Registry's readings of the instability composed after the figures" }, { "act": "holding together (validation)", "sentences": [ 14, 15, 16, 17 ], "claims": [ "F16", "F25", "F28", "F43", "MMRS-06" ], "note": "the tracker page's 'not yet completed', the June 2026 pilot and the Navigator's 'not yet validated' composed side by side at their own modality; the update's own framing (a pilot, a fuller study next, F35) carried" }, { "act": "placing in the open block", "sentences": [], "claims": [ "CAP-1-04", "F1", "F16", "F2", "F21", "F23", "F25", "F27", "F28", "F3", "F35", "F37", "F41", "F42", "F43", "F53", "F54", "F6", "MMRS-06", "MMRS-09", "MMRS-13", "MMRS-15" ], "note": "each open question rests on claims composed in the body at their own modality" } ], "plan": { "order": [ "MMRS-01", "MMRS-10", "F46", "F1", "MMRS-18", "F8", "F50", "F4", "F5", "F12", "MMRS-03", "F30", "F6", "F23", "F44", "MMRS-04", "MMRS-14", "F24", "F9", "F10", "F20", "F38", "F40", "F11", "F39", "F7", "F47", "F48", "F22", "F21", "MMRS-09", "MMRS-17", "F49", "F18", "MMRS-13", "F37", "F36", "F2", "MMRS-05", "F3", "F41", "F42", "F45", "MMRS-15", "CAP-1-04", "F35", "F25", "F26", "F29", "F27", "F28", "F34", "F13", "F14", "F15", "F16", "F17", "F33", "F43", "MMRS-02", "MMRS-06", "MMRS-16", "CAP-1-01", "CAP-1-03", "CAP-1-02", "CAP-1-05", "F19", "MMRS-07", "MMRS-19", "MMRS-20", "MMRS-08", "MMRS-21", "MMRS-12", "MMRS-22", "F31", "MMRS-11", "F32", "F51", "F52", "F53", "F54" ], "senses": { "what it is": [ 1 ], "source": [ 2 ], "purpose": [ 3 ], "method": [ 4 ], "dimensions": [ 5, 6 ], "the national-security view": [ 7 ], "operator": [ 8 ], "origin": [ 9 ], "the unit": [ 10 ], "count": [ 11, 12 ], "versions": [ 13 ], "validation": [ 14, 15 ], "limits": [ 16, 17 ], "reception": [ 18 ], "openness": [ 19 ], "place": [ 20 ], "the record kept apart": [ 21 ], "neighbours": [ 22 ], "uptake": [ 23 ], "the encyclopedia": [ 24 ] }, "case_by_sense": { "equal (one claim space; nothing translated)": [ "what it is", "source", "purpose", "method", "dimensions", "the national-security view", "operator", "origin", "the unit", "count", "versions", "validation", "limits", "reception", "openness", "place", "the record kept apart", "neighbours", "uptake", "the encyclopedia" ] }, "carried": { "archive": 27, "field": 54, "not_composed": {} }, "conflict_log": [ { "rule": "D_pres vs channel", "result": "no cut: the KO carries every field and archive claim; the compression's bound met (body 257 words) with no channel log" }, { "rule": "P_coh", "result": "MMRS-13 'AI Incident Database report' against F1/F37 (incidents counted and reclassified), sentence [10]; the validation statements F16, F25–F28, F43 (sentences [14, 15, 16, 17]): each composed with its source" }, { "rule": "N_c", "result": "no claim lowered by standing; the MMRS record's readings (MMRS-19, MMRS-20: basis 'reading') carried at interpretation, as the record codes them; the June 2026 pilot carried at its own scope ('on this pilot sample')" }, { "rule": "C_ex", "result": "senses composed ⊇ senses in the ledgers: 20 senses, every ledger lineage cited in the KO" } ] } }, "P": { "title": "MIT AI Incident Tracker", "genre": "the compression: AIO's interaction grammar (lede, headed clusters, bolded labels, card rail), composed from the v0.9 plan in one claim space; it faces on the card, with AIO's transcript a record within it", "status": "composed 2026-10-29 from the v0.9 knowledge object; not frozen", "lede": { "text": "is the MIT AI Risk Initiative's machine classification of AI Incident Database reports by risk domain, cause, harm, severity and EU AI Act level, through Arcola AI's LLM pipeline (MMRS record; its page), its patterns, by its own word, \"indicative\".", "claims": [ "MMRS-01", "F7", "F6", "F17", "MMRS-02" ] }, "sections": [ { "icon": "🗂️", "head": "What it is", "items": [ { "label": "Source:", "text": "AIID reports, which the Responsible AI Collaborative reviews; developed with the AIID (tracker and about pages).", "claims": [ "F8", "F46", "MMRS-18" ] }, { "label": "Method:", "text": "an LLM pipeline (Sonnet 4 in June 2026) against the Risk Repository's taxonomies, with human spot-checks.", "claims": [ "F6", "F23", "F27", "F44", "MMRS-04" ] }, { "label": "Origin:", "text": "developed and led by Simon Mylius; part of the MIT AI Risk Initiative at MIT FutureTech.", "claims": [ "F49", "F7", "F47" ] } ] }, { "icon": "🧾", "head": "Dimensions", "items": [ { "label": "Taxonomies:", "text": "harm severity (1–5 since June 2025), EU AI Act tier, cause, 7 domains, 24 subdomains.", "claims": [ "F24", "F38", "F9" ] }, { "label": "National security:", "text": "five impact categories; threat by imminence, novelty, autonomy.", "claims": [ "F11", "F39" ] }, { "label": "Unit:", "text": "the report, in the MMRS record; MIT's pages count incidents.", "claims": [ "MMRS-13", "F1", "F37" ] } ] }, { "icon": "🔢", "head": "The count", "items": [ { "label": "Four figures:", "text": "1,600 (heading), more than 1,400 (description), 1,752 (Google's card), 1,497 on the Navigator, \"more than 1300\" on its card.", "claims": [ "F1", "F2", "F3", "F42", "F41", "MMRS-15" ] }, { "label": "Versions:", "text": "reclassified whole at each update; #1116 on 23 June 2025.", "claims": [ "F36", "F37", "F35" ] } ] }, { "icon": "⚖️", "head": "Limits", "items": [ { "label": "Caveat:", "text": "voluntary reporting, sampling bias, no systematic validation; patterns \"indicative\" (tracker page).", "claims": [ "F15", "F16", "F17", "MMRS-06" ] }, { "label": "Pilot:", "text": "ten incidents, two reviewers; human agreement only \"Fair\" on severity; Opus 4.6 met it after prompt refinement (June 2026).", "claims": [ "F26", "F29", "F28", "F34" ] }, { "label": "Navigator:", "text": "\"preliminary and not yet validated\".", "claims": [ "F43" ] } ] }, { "icon": "🧭", "head": "Place", "items": [ { "label": "Derived layer:", "text": "AIID's editor-resolved incidents reclassified by a model, machine-written, a chain of receptions (MMRS record).", "claims": [ "MMRS-07", "MMRS-20", "MMRS-08" ] }, { "label": "Beside:", "text": "the AIID, the OECD's monitor, the AIAAIC, the Political Deepfakes Database (June 2026).", "claims": [ "F31" ] }, { "label": "Wikipedia:", "text": "no article; one mention, under the Risk Repository's name.", "claims": [ "F51", "F53" ] } ] }, { "icon": "👁️", "head": "On the Overview", "items": [ { "label": "Recorded:", "text": "the machine method named; no count; the caveat dropped (Capture Registry, 2026-10-29).", "claims": [ "CAP-1-01", "CAP-1-02", "CAP-1-05" ] } ] } ], "open": [ { "question": "How many does the tracker classify, and of what: 1,600 (heading), more than 1,400 (description), 1,752 (Google's card), 1,497 or more than 1,300 (the Navigator)?", "rests_on": [ "F1", "F2", "F3", "F41", "F42", "MMRS-15", "CAP-1-04" ], "modality": "documented, all, on different surfaces at undated or different times; the 1,752 is in no saved page", "would_resolve": "a dated count on the tracker's own page, or the dataset's row count; a later capture at the address" }, { "question": "Is the unit the AIID report or the AIID incident?", "rests_on": [ "MMRS-13", "F1", "F37", "F23", "F54" ], "modality": "documented (the MMRS record's unit, basis operator) against stipulation (MIT's pages, which count and reclassify incidents and classify reports)", "would_resolve": "the tracker's dataset schema, read for its row key" }, { "question": "Has the classification been validated: none completed (tracker page), a pilot reported (June 2026), not yet validated (Navigator)?", "rests_on": [ "F16", "F25", "F28", "F35", "F43", "MMRS-06" ], "modality": "self-description, all three; the pilot is documented", "would_resolve": "the comprehensive validation study the June 2026 update announces" }, { "question": "Did MIT's classification of AIID incident 277 change, or does Wikipedia misreport it?", "rests_on": [ "F53", "F54", "F37", "F35" ], "modality": "attributed (Wikipedia, access-date 5 October 2025) against documented (the page as fetched 2026-10-29)", "would_resolve": "an archived copy of incidentdatabase.ai/cite/277 at Wikipedia's access date" }, { "question": "What is Arcola AI, and is its pipeline the one that runs Sonnet 4?", "rests_on": [ "F6", "F27", "MMRS-09" ], "modality": "stipulation (tracker page) and documented (June 2026 update); no ledgered source describes Arcola AI", "would_resolve": "an MIT or Arcola AI statement of the pipeline" }, { "question": "Opacity: card 2's summary ('Tools and Data: Expert Survey …') and card 1's 1,752 are not in the saved pages.", "rests_on": [ "F21", "F3" ], "modality": "card text only", "would_resolve": "the pages at the time the cards were drawn" } ], "rail": [ { "lineage": "what it is: a classification of AIID incidents", "claims": [ "MMRS-01", "MMRS-10", "F46", "F1" ], "earliest": "B1" }, { "lineage": "source: the AI Incident Database", "claims": [ "MMRS-18", "F8", "F50" ], "earliest": "B1" }, { "lineage": "purpose: structure for policymaking", "claims": [ "F4", "F5", "F12", "MMRS-03", "F30" ], "earliest": "B3" }, { "lineage": "method: Arcola AI's LLM pipeline over AIID reports", "claims": [ "F6", "F23", "F44", "MMRS-04", "MMRS-14" ], "earliest": "B3" }, { "lineage": "dimensions: taxonomies and scales", "claims": [ "F24", "F9", "F10", "F20", "F38", "F40" ], "earliest": "B4" }, { "lineage": "dimensions: the national-security view", "claims": [ "F11", "F39" ], "earliest": "B4" }, { "lineage": "operator: the MIT AI Risk Initiative at MIT FutureTech", "claims": [ "F7", "F47", "F48", "F22", "F21", "MMRS-09", "MMRS-17" ], "earliest": "B1" }, { "lineage": "origin: developed by Simon Mylius", "claims": [ "F49", "F18" ], "earliest": "B1" }, { "lineage": "the unit: report or incident", "claims": [ "MMRS-13" ], "earliest": "MMRS" }, { "lineage": "updates: reclassification by version", "claims": [ "F37", "F36", "F35" ], "earliest": "B4" }, { "lineage": "count: four figures across MIT's surfaces", "claims": [ "F2", "MMRS-05", "F3", "F41", "F42", "F45", "MMRS-15", "CAP-1-04" ], "earliest": "B1" }, { "lineage": "validation: the June 2026 pilot", "claims": [ "F25", "F26", "F29", "F27", "F28", "F34" ], "earliest": "B3" }, { "lineage": "limits: indicative, not validated", "claims": [ "F13", "F14", "F15", "F16", "F17", "F33", "F43", "MMRS-02", "MMRS-06", "MMRS-16", "CAP-1-05" ], "earliest": "B3" }, { "lineage": "reception: the composition layer", "claims": [ "CAP-1-01", "CAP-1-03", "CAP-1-02" ], "earliest": "capture:mit-ai-incident-tracker-aio-20261029" }, { "lineage": "openness: CC BY 4.0", "claims": [ "F19" ], "earliest": "B1" }, { "lineage": "place: a derived layer, a chain of receptions", "claims": [ "MMRS-07", "MMRS-19", "MMRS-20", "MMRS-08" ], "earliest": "MMRS" }, { "lineage": "the record kept apart", "claims": [ "MMRS-21", "MMRS-12", "MMRS-22" ], "earliest": "MMRS" }, { "lineage": "neighbours: the incident databases it builds on", "claims": [ "F31", "MMRS-11" ], "earliest": "B3" }, { "lineage": "reception: uptake of the tracker", "claims": [ "F32" ], "earliest": "B3" }, { "lineage": "the encyclopedia: no article; one mention under the repository's name", "claims": [ "F51", "F52", "F53", "F54" ], "earliest": "W3" } ], "body_words": 257, "entity": "Composed under v0.9, equal case (ρ = extends): KO_B sentences surviving 0 of 11; every field claim carried." }, "P_B": { "title": "MIT AI Incident Tracker", "arm": "B (the field alone)", "genre": "popup — the compression from the field ledger alone (cards and Wikipedia)", "status": "composed 2026-10-29 from the field ledger alone; not frozen", "lede": { "text": "classifies real-world incidents from the AI Incident Database by risk, cause, harm and severity, through Arcola AI's LLM pipeline; part of the MIT AI Risk Initiative (its page).", "claims": [ "F1", "F6", "F7" ] }, "sections": [ { "icon": "🗂️", "head": "What it is", "items": [ { "label": "Catalogue:", "text": "incidents where AI has already caused harm, developed with the AIID (about page).", "claims": [ "F46", "F8" ] }, { "label": "Origin:", "text": "developed and led by Simon Mylius at MIT FutureTech.", "claims": [ "F49", "F47" ] }, { "label": "Purpose:", "text": "a consistent structure for policy, after incident reporting in aviation.", "claims": [ "F4", "F5", "F30" ] } ] }, { "icon": "🧾", "head": "How it classifies", "items": [ { "label": "Taxonomies:", "text": "harm severity (1–5), EU AI Act tier, cause, domain, subdomain.", "claims": [ "F24", "F38", "F23" ] }, { "label": "National security:", "text": "five impact categories, threat by imminence, novelty, autonomy.", "claims": [ "F11", "F39" ] }, { "label": "Updates:", "text": "the whole dataset reclassified each version; #1116 in June 2025.", "claims": [ "F36", "F37" ] } ] }, { "icon": "⚖️", "head": "Limits", "items": [ { "label": "Caveat:", "text": "voluntary reporting, sampling bias, no systematic validation; patterns indicative.", "claims": [ "F15", "F16", "F17" ] }, { "label": "Pilot:", "text": "ten incidents, two reviewers; Opus 4.6 reached the human baseline after prompt refinement.", "claims": [ "F26", "F28", "F34" ] }, { "label": "Counts:", "text": "1,600, more than 1,400, 1,752, 1,497 and more than 1,300 across MIT's surfaces.", "claims": [ "F1", "F2", "F3", "F42", "F41" ] } ] }, { "icon": "🧭", "head": "Place", "items": [ { "label": "Beside:", "text": "the AIID, the OECD's monitor, the AIAAIC (June 2026).", "claims": [ "F31" ] }, { "label": "Wikipedia:", "text": "no article; one mention, under the Risk Repository's name.", "claims": [ "F51", "F53" ] } ] } ], "open": [ { "question": "Which count stands?", "rests_on": [ "F1", "F2", "F3", "F41", "F42" ], "modality": "documented, on different surfaces", "would_resolve": "a dated count on the tracker's page" }, { "question": "Has the classification been validated?", "rests_on": [ "F16", "F28", "F43" ], "modality": "self-description", "would_resolve": "the comprehensive validation study announced (F35)" } ], "rail": [ { "lineage": "what it is: a classification of AIID incidents", "claims": [ "F1", "F46", "F1" ], "earliest": "B1" }, { "lineage": "method: Arcola AI's LLM pipeline over AIID reports", "claims": [ "F6", "F23", "F44" ], "earliest": "B3" }, { "lineage": "dimensions: taxonomies and scales", "claims": [ "F24", "F9", "F10", "F20", "F38", "F40" ], "earliest": "B4" }, { "lineage": "purpose: structure for policymaking", "claims": [ "F4", "F5", "F12", "F30" ], "earliest": "B3" }, { "lineage": "operator: the MIT AI Risk Initiative at MIT FutureTech", "claims": [ "F7", "F47", "F48", "F22", "F21" ], "earliest": "B1" }, { "lineage": "origin: developed by Simon Mylius", "claims": [ "F49", "F18" ], "earliest": "B1" }, { "lineage": "source: the AI Incident Database", "claims": [ "F8", "F50" ], "earliest": "B1" }, { "lineage": "dimensions: the national-security view", "claims": [ "F11", "F39" ], "earliest": "B4" }, { "lineage": "updates: reclassification by version", "claims": [ "F36", "F37", "F35" ], "earliest": "B4" }, { "lineage": "count: four figures across MIT's surfaces", "claims": [ "F2", "F3", "F41", "F42", "F45" ], "earliest": "B1" }, { "lineage": "validation: the June 2026 pilot", "claims": [ "F25", "F26", "F29", "F27", "F28", "F34" ], "earliest": "B3" }, { "lineage": "limits: indicative, not validated", "claims": [ "F13", "F14", "F15", "F16", "F17", "F33", "F43" ], "earliest": "B3" }, { "lineage": "neighbours: the incident databases it builds on", "claims": [ "F31" ], "earliest": "B3" }, { "lineage": "reception: uptake of the tracker", "claims": [ "F32" ], "earliest": "B3" }, { "lineage": "openness: CC BY 4.0", "claims": [ "F19" ], "earliest": "B1" }, { "lineage": "the encyclopedia: no article; one mention under the repository's name", "claims": [ "F51", "F52", "F53", "F54" ], "earliest": "W3" } ], "body_words": 160, "entity": "With no archive claim admitted, the entry is the field's tracker at its own resolution: MIT's pages, the Navigator and the encyclopedia's one mention." }, "KO_B": { "title": "The MIT AI Incident Tracker", "arm": "B (the field alone)", "genre": "knowledge object", "status": "composed 2026-10-29; not frozen", "sentences": [ { "n": 1, "para": 0, "text": "The MIT AI Incident Tracker classifies 1,600 real-world report incidents from the AI Incident Database by risk, cause, harm, severity and other dimensions, by its page's heading, a catalogue of incidents where AI has already caused harm developed in collaboration with the AIID, by the about page.", "claims": [ "F1", "F46" ] }, { "n": 2, "para": 0, "text": "Raw AIID reports go through Arcola AI's LLM pipeline against the MIT Risk Repository and a CSET-based severity rating; the June 2026 update describes LLMs classifying the reports along five taxonomies, harm severity, EU AI Act level, cause, domain and subdomain, and the Navigator an LLM pipeline with human spot-checking over AIID incidents from news, papers and government reports.", "claims": [ "F6", "F23", "F24", "F44" ] }, { "n": 3, "para": 0, "text": "Incident databases' inconsistent structure limits their use for policy, the page says, and the tool is to give researchers, developers, auditors, policymakers and regulators a common frame of reference, as incident reporting did for aviation, healthcare and nuclear power.", "claims": [ "F4", "F5", "F12", "F30" ] }, { "n": 4, "para": 1, "text": "It is part of the MIT AI Risk Initiative at MIT FutureTech, whose home and about pages present connected workstreams and authoritative data and frameworks, and Simon Mylius developed and leads it, the team including the AIID's Sean McGregor; it builds on the AIID, created and maintained by the Responsible AI Collaborative, which the about page also lists as a user mapping incidents to the taxonomy.", "claims": [ "F7", "F47", "F48", "F22", "F21", "F49", "F18", "F8", "F50" ] }, { "n": 5, "para": 1, "text": "Per incident it shows the causal and domain taxonomies, the EU AI Act levels, severity across ten CSET-based harm types and the classification's confidence, with a national-security view of five impact categories and a threat classification, and the home page compares incidents by domain, harm, people affected, developer, time and national-security impact.", "claims": [ "F9", "F10", "F11", "F39", "F20" ] }, { "n": 6, "para": 2, "text": "The June 2025 update reached AIID incident #1116, reclassified every incident for consistency, moved severity from 0–10 to 1–5 and added flags for ambiguities and the system's primary goal; the June 2026 update plans a fuller validation study and a reclassification of all AIID reports.", "claims": [ "F36", "F37", "F38", "F40", "F35" ] }, { "n": 7, "para": 2, "text": "The counts differ: 1,600 in the page's heading, more than 1,400 in its description, 1,752 in Google's card for the page, more than 1,300 on the Navigator's card and 1,497 in the Navigator's heading, beside its headline figures.", "claims": [ "F1", "F2", "F3", "F41", "F42", "F45" ] }, { "n": 8, "para": 3, "text": "The June 2026 pilot set eight LLMs against two expert reviewers on ten incidents, found human agreement only fair on severity and moderate on the EU AI Act level, found Sonnet 4, the model in use, met or exceeded by Opus 4.6 and Kimi K2.5 on most taxonomies, and after prompt refinement found Opus 4.6 at the human baseline on all five, on a sample it calls small.", "claims": [ "F25", "F26", "F29", "F27", "F28", "F34" ] }, { "n": 9, "para": 3, "text": "The page's limits are voluntary reporting and sampling bias, reports of varying quality, spot-checks without a systematic validation study, and patterns to be taken as indicative; LLMs can hallucinate and show score-range bias, and the Navigator's data are preliminary and not yet validated.", "claims": [ "F13", "F14", "F15", "F16", "F17", "F33", "F43" ] }, { "n": 10, "para": 4, "text": "It builds on the AIID, the OECD's monitor, the AIAAIC Repository and the Political Deepfakes Database, its outputs used in the 2026 World Disasters Report and Time, and its data are CC BY 4.0.", "claims": [ "F31", "F32", "F19" ] }, { "n": 11, "para": 4, "text": "Wikipedia has no article for it; its only mention credits the MIT AI Risk Repository with cataloguing AIID incident 277 as economic and cultural devaluation of human effort, while the AIID page for 277 now shows the machine classification 4.3, fraud, scams and targeted manipulation.", "claims": [ "F51", "F52", "F53", "F54" ] } ] }, "L_B": { "words": 549, "text": "The MIT AI Incident Tracker classifies 1,600 real-world report incidents from the AI Incident Database by risk, cause, harm, severity and other dimensions, by its page's heading, a catalogue of incidents where AI has already caused harm developed in collaboration with the AIID, by the about page. Raw AIID reports go through Arcola AI's LLM pipeline against the MIT Risk Repository and a CSET-based severity rating; the June 2026 update describes LLMs classifying the reports along five taxonomies, harm severity, EU AI Act level, cause, domain and subdomain, and the Navigator an LLM pipeline with human spot-checking over AIID incidents from news, papers and government reports. Incident databases' inconsistent structure limits their use for policy, the page says, and the tool is to give researchers, developers, auditors, policymakers and regulators a common frame of reference, as incident reporting did for aviation, healthcare and nuclear power. It is part of the MIT AI Risk Initiative at MIT FutureTech, whose home and about pages present connected workstreams and authoritative data and frameworks, and Simon Mylius developed and leads it, the team including the AIID's Sean McGregor; it builds on the AIID, created and maintained by the Responsible AI Collaborative, which the about page also lists as a user mapping incidents to the taxonomy. Per incident it shows the causal and domain taxonomies, the EU AI Act levels, severity across ten CSET-based harm types and the classification's confidence, with a national-security view of five impact categories and a threat classification, and the home page compares incidents by domain, harm, people affected, developer, time and national-security impact. The June 2025 update reached AIID incident #1116, reclassified every incident for consistency, moved severity from 0–10 to 1–5 and added flags for ambiguities and the system's primary goal; the June 2026 update plans a fuller validation study and a reclassification of all AIID reports. The counts differ: 1,600 in the page's heading, more than 1,400 in its description, 1,752 in Google's card for the page, more than 1,300 on the Navigator's card and 1,497 in the Navigator's heading, beside its headline figures. The June 2026 pilot set eight LLMs against two expert reviewers on ten incidents, found human agreement only fair on severity and moderate on the EU AI Act level, found Sonnet 4, the model in use, met or exceeded by Opus 4.6 and Kimi K2.5 on most taxonomies, and after prompt refinement found Opus 4.6 at the human baseline on all five, on a sample it calls small. The page's limits are voluntary reporting and sampling bias, reports of varying quality, spot-checks without a systematic validation study, and patterns to be taken as indicative; LLMs can hallucinate and show score-range bias, and the Navigator's data are preliminary and not yet validated. It builds on the AIID, the OECD's monitor, the AIAAIC Repository and the Political Deepfakes Database, its outputs used in the 2026 World Disasters Report and Time, and its data are CC BY 4.0. Wikipedia has no article for it; its only mention credits the MIT AI Risk Repository with cataloguing AIID incident 277 as economic and cultural devaluation of human effort, while the AIID page for 277 now shows the machine classification 4.3, fraud, scams and targeted manipulation.", "note": "L_F(B) is realized as the field arm's expansion (objects.KO_B): the field alone, cards and Wikipedia." }, "L_BA": { "words": 1465, "text": "The MIT AI Incident Tracker is a machine classification of AI Incident Database reports by risk domain, cause, harm, severity and EU AI Act risk level, made by Arcola AI's language-model pipeline against the MIT Risk Repository's taxonomy, and filed as a database (MMRS record); the MIT AI Risk Initiative's about page calls it \"a catalogue that classifies incidents where AI has already caused harm\", developed in collaboration with the AI Incident Database, and its own page says it classifies AIID incidents \"by risk, cause, harm, severity, and other relevant dimensions\". It is derived from the AI Incident Database (MMRS record): the tracker page builds on the Responsible AI Collaborative, \"who created and maintain the AI Incident Database, reviewing all submitted reports\", and sends its readers to the AIID for how incidents are defined and ingested, and the about page lists the AIID among the users of the Initiative's work, mapping real-world incidents to its taxonomy in the tracker. Its occasion, on its own page, is that incident databases \"struggle with inconsistent structure, limiting their utility for policymaking\", and its answer is a tool to classify incidents by risk and harm severity, offered as a common frame of reference for researchers, developers, businesses, evaluators, auditors, policymakers and regulators; the MMRS record restates the aim as a more consistent structure for AIID's records \"by its own account\", and the June 2026 update grounds it in what incident reporting did for safety in aviation, healthcare and nuclear power. Raw AIID reports are \"processed by Arcola AI's LLM pipeline\", which categorizes them with the MIT Risk Repository and a harm severity rating based on CSET's AI Harm Taxonomy (tracker page); the June 2026 update says the tracker uses LLMs to classify incident reports along five taxonomies, \"producing standardized data\", the Navigator that each incident is classified against shared taxonomies \"using an LLM pipeline with human spot-checking\", and the MMRS record that the classification is machine-made, its intake the AIID reports passed through Arcola AI's pipeline. The five taxonomies are harm severity across ten harm types, the EU AI Act risk tier, a causal taxonomy of lifecycle stage and intent, seven risk domains and twenty-four subdomains (June 2026 update); the tracker page shows each incident under the repository's causal and domain taxonomies and the EU AI Act's levels, with its severity across ten CSET-based harm categories and the classification's confidence, and the Initiative's home page lets incidents be compared by risk domain, harm, people affected, alleged developer, time or national-security impact. On 23 June 2025 the severity scale went from 0–10 to 1–5, its definitions disambiguated in the first person (\"I have tried to remove ambiguities\"), and the tracker added flags for ambiguities and alternative interpretations and the system's primary goal by a taxonomy based on the AIID's GMF (June 2025 update). A national-security view grades each incident's impact on its primary target country in five categories, from physical security and critical infrastructure to societal stability and human rights, with the threat classified for imminence, novelty and autonomy (tracker page; June 2025 update). It is part of the MIT AI Risk Initiative (tracker page), based at MIT FutureTech, a joint initiative of MIT's Institute of the Digital Economy and CSAIL, which keeps several connected workstreams because no single dataset captures AI risk and presents its data and frameworks as authoritative (about and home pages; card 2's summary lists among its tools an expert survey and the AI Risk Repository); the MMRS record gives the operator as the Initiative \"with Arcola AI's LLM pipeline\" and the tracker as part of the MIT AI Risk Repository. Simon Mylius \"developed and leads the AI Incident Tracker\", a classification tool and dashboard now hosted by the MIT AI Risk Repository, its analysis featured on the AIID (about page), and the tracker page's team names MIT FutureTech and University of Queensland researchers beside the AIID's Sean McGregor. The MMRS record gives the unit as the AIID report, classified by domain, cause, harm, severity and EU AI Act level; the tracker's pages count incidents (\"1,600 real-world report incidents\"), and the June 2025 update reclassified \"all incidents\" up to AIID incident ID #1116. MIT's surfaces give four figures: 1,600 in the tracker page's heading and \"more than 1,400\" in its description, which the MMRS record reports as the page giving its size twice; 1,752 \"real-world AI incident reports\" in the card's summary of that page; and on the Navigator \"more than 1300\" in its card and 1,497 in its page heading on 2026-10-29, beside headline figures of \"Up -76%\" year over year and 64% caused by AI systems. The MMRS record's count field (as of 2026-10-28, basis observed) carries the page's two figures, Google's card's 1,752 and the Navigator's \"more than 1300\", each with its surface, and the Capture Registry (2026-10-29) records the count \"three ways\" and that the composition gave none. The tracker is versioned by reclassification: the June 2025 update brought it up to AIID incident #1116 and classified the full dataset again with the new iteration \"to ensure consistency\", and the June 2026 update names as its next step a more comprehensive validation study and announces that the full set of AIID reports will be reclassified with the updated model selection and prompts. That update (30 June 2026) reports a pilot: because the tracker relies on automated classification rather than manual review, two expert reviewers classified ten AIID incidents on the five taxonomies, eight LLMs were set against their consensus, and the human agreement was itself only \"Fair\" for harm severity (0.31) and \"Moderate\" for the EU AI Act level (0.56). Opus 4.6 and Kimi K2.5 met or exceeded the model in current use, Sonnet 4, on most taxonomies, the models did substantially worse on the EU AI Act level, and after prompt refinement Opus 4.6 met the human baseline on all five \"on this pilot sample\", a sample the update calls small; it concludes that LLM classification has accuracy comparable to the human baseline. The tracker page states its limits: the classification database explores \"the potential capabilities and limitations of a scalable incident analysis framework\", its AIID input rests on public and expert submissions of varying quality, voluntary reporting gives \"some degree of sampling bias\", spot-checks have improved it but \"a systematic validation study has not yet been completed\", and patterns \"should be taken as indicative and validated through further analysis\". The June 2026 update adds that LLMs can hallucinate details and show a score-range bias, the Navigator calls itself a research prototype whose data and analyses are \"preliminary and not yet validated\", and the MMRS record carries the page's caveat in its kernel, its article and its note. At its name Google's AI Overview composed the tracker \"as what it says it is, with its machine method named\", AIID as the source and LLM pipelines as the method, which are the MMRS record's derivation and intake, avoided stating a count, and left out the page's caveat of indicative patterns, voluntary reporting and no completed validation study (Capture Registry, 2026-10-29). Data from the MIT AI Risk Initiative is licensed under CC BY 4.0 (tracker page), as most of its outputs are (about page). In the MMRS record it is a derived layer, \"the editor-resolved incidents of AIID, reclassified by a model\", whose admission inherits AIID's harm test and whose classifications are machine-written, so that original outputs are not preserved, and the chain from event to news report to AIID record to machine classification is \"a chain of receptions, each recorded\". The record keeps each assertion with its source and basis and the lifecycle apart from the observation that saw it (read 2026-10-28), and it rests every field on the one tracker page. The June 2026 update places the tracker among the incident databases it builds on, the AI Incident Database, the OECD's AI Incident Monitor, the AIAAIC Repository and the Political Deepfakes Database, adding to them a classification of each incident and a rating of its harm; the MMRS record gives it the role of harm. Its outputs have been used to report incident trends in the 2026 World Disasters Report and Time magazine (June 2026 update). Wikipedia has no article for the tracker, nor for the AI Risk Repository or the MIT AI Risk Initiative, and none of its articles contains the phrase \"AI Incident Tracker\"; its one mention of the work, in \"Voiceverse NFT plagiarism scandal\", says that \"The MIT AI Risk Repository catalogued the incident\" as \"economic and cultural devaluation of human effort\", referenced to the AIID's page for incident 277, which on 2026-10-29 shows the MIT classification as machine-classified under subdomain 4.3, fraud, scams and targeted manipulation, in the domain Malicious Actors & Misuse.", "note": "L(B ∪ A) in the equal-terms case (v0.9 §0.4), realized as the knowledge object (objects.KO); its rail is the compression's, by lineage.", "rail_lineages": [ { "lineage": "what it is: a classification of AIID incidents", "claims": [ "MMRS-01", "MMRS-10", "F46", "F1" ], "earliest": "B1" }, { "lineage": "source: the AI Incident Database", "claims": [ "MMRS-18", "F8", "F50" ], "earliest": "B1" }, { "lineage": "purpose: structure for policymaking", "claims": [ "F4", "F5", "F12", "MMRS-03", "F30" ], "earliest": "B3" }, { "lineage": "method: Arcola AI's LLM pipeline over AIID reports", "claims": [ "F6", "F23", "F44", "MMRS-04", "MMRS-14" ], "earliest": "B3" }, { "lineage": "dimensions: taxonomies and scales", "claims": [ "F24", "F9", "F10", "F20", "F38", "F40" ], "earliest": "B4" }, { "lineage": "dimensions: the national-security view", "claims": [ "F11", "F39" ], "earliest": "B4" }, { "lineage": "operator: the MIT AI Risk Initiative at MIT FutureTech", "claims": [ "F7", "F47", "F48", "F22", "F21", "MMRS-09", "MMRS-17" ], "earliest": "B1" }, { "lineage": "origin: developed by Simon Mylius", "claims": [ "F49", "F18" ], "earliest": "B1" }, { "lineage": "the unit: report or incident", "claims": [ "MMRS-13" ], "earliest": "MMRS" }, { "lineage": "updates: reclassification by version", "claims": [ "F37", "F36", "F35" ], "earliest": "B4" }, { "lineage": "count: four figures across MIT's surfaces", "claims": [ "F2", "MMRS-05", "F3", "F41", "F42", "F45", "MMRS-15", "CAP-1-04" ], "earliest": "B1" }, { "lineage": "validation: the June 2026 pilot", "claims": [ "F25", "F26", "F29", "F27", "F28", "F34" ], "earliest": "B3" }, { "lineage": "limits: indicative, not validated", "claims": [ "F13", "F14", "F15", "F16", "F17", "F33", "F43", "MMRS-02", "MMRS-06", "MMRS-16", "CAP-1-05" ], "earliest": "B3" }, { "lineage": "reception: the composition layer", "claims": [ "CAP-1-01", "CAP-1-03", "CAP-1-02" ], "earliest": "capture:mit-ai-incident-tracker-aio-20261029" }, { "lineage": "openness: CC BY 4.0", "claims": [ "F19" ], "earliest": "B1" }, { "lineage": "place: a derived layer, a chain of receptions", "claims": [ "MMRS-07", "MMRS-19", "MMRS-20", "MMRS-08" ], "earliest": "MMRS" }, { "lineage": "the record kept apart", "claims": [ "MMRS-21", "MMRS-12", "MMRS-22" ], "earliest": "MMRS" }, { "lineage": "neighbours: the incident databases it builds on", "claims": [ "F31", "MMRS-11" ], "earliest": "B3" }, { "lineage": "reception: uptake of the tracker", "claims": [ "F32" ], "earliest": "B3" }, { "lineage": "the encyclopedia: no article; one mention under the repository's name", "claims": [ "F51", "F52", "F53", "F54" ], "earliest": "W3" } ], "rail": [] } }, "field": [ { "id": "B1", "card": "MIT AI Risk Initiative — MIT AI Incident Tracker", "fetched": "card; HTTP 200 at 2026-10-29T05:07:24Z; B1.html sha256 720c3b26f04adc57", "origin": "card" }, { "id": "B2", "card": "MIT AI Risk Initiative — MIT AI Risk Initiative", "fetched": "card; HTTP 200 at 2026-10-29T05:07:24Z; B2.html sha256 d3f35366280a66ca", "origin": "card" }, { "id": "B3", "card": "MIT AI Risk Initiative — AI Incident Tracker June 2026 Update - MIT AI Risk Initiative", "fetched": "card; HTTP 200 at 2026-10-29T05:07:25Z; B3.html sha256 d8b013b1dbc4f948", "origin": "card" }, { "id": "B4", "card": "MIT AI Risk Initiative — Incident Tracker - June 2025 Update - MIT AI Risk Repository", "fetched": "card; HTTP 200 at 2026-10-29T05:07:25Z; B4.html sha256 952b48f0be997dc3", "origin": "card" }, { "id": "B5", "card": "MIT AI Risk Navigator — Incidents - MIT AI Risk Navigator", "fetched": "card; HTTP 200 at 2026-10-29T05:07:26Z; B5.html sha256 8c6757061ebc4be1", "origin": "card" }, { "id": "B6", "card": "MIT AI Risk Initiative — About - MIT AI Risk Initiative", "fetched": "card; HTTP 200 at 2026-10-29T05:07:26Z; B6.html sha256 b40b72f7dbd5c16c", "origin": "card" }, { "id": "W1", "card": "Wikipedia — MIT AI Incident Tracker (no article)", "fetched": "wikipedia; HTTP 404 at 2026-10-29T05:07:26Z; W1.html sha256 079e59d86eb3bc8c (no article)", "origin": "wikipedia" }, { "id": "W3", "card": "Wikipedia — Voiceverse NFT plagiarism scandal", "fetched": "wikipedia; HTTP 200 at 2026-10-29T05:11:17Z; W3.html sha256 b5916526f7bbf7c1", "origin": "wikipedia" }, { "id": "W4", "card": "AI Incident Database — Incident 277 (Wikipedia's reference)", "fetched": "wikipedia; HTTP 200 at 2026-10-29T05:10:56Z; W4.html sha256 e4d67493c6e1de91", "origin": "wikipedia" } ], "field_claims": { "F1": { "source": "B1", "origin": "card", "locus": "opening ('How is AI harming us?')", "quote": "The MIT AI Incident Tracker classifies 1,600 real-world report incidents from the AI Incident Database by risk, cause, harm, severity, and other relevant dimensions.", "claim": "the tracker defined by its own page: it classifies 1,600 AIID incidents by risk, cause, harm, severity and other dimensions", "modality": "stipulation", "kind": "definition", "kernel": true, "const": true, "lineage": "what it is: a classification of AIID incidents", "date": "2026-10-29 (undated; as fetched; the MMRS record read it 2026-10-28)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 8d0169460225d228…), whitespace-normalised" }, "F2": { "source": "B1", "origin": "card", "locus": "meta description (name=description, og:description, twitter:description), extracted to B1-meta.txt", "quote": "The MIT AI Incident Tracker classifies more than 1,400 real-world reported incidents from the AI Incident Database by risk, cause, harm, severity, and other relevant dimensions.", "claim": "the same page's description gives the count as more than 1,400", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "count: four figures across MIT's surfaces", "date": "2026-10-29 (undated; as fetched; the MMRS record read it 2026-10-28)", "quote_ok": true, "checked_against": "saved text B1-meta.txt (sha256 f6cbad6e571b1744…), whitespace-normalised" }, "F3": { "source": "B1", "origin": "card", "locus": "card 1 snippet (the count is not in the saved page)", "quote": "Overview: MIT AI Incident Tracker classifies 1,752 real-world AI incident reports from the AI Incident Database using Arcola AI's LLM pipeline", "claim": "the card's summary gives 1,752 real-world AI incident reports", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "count: four figures across MIT's surfaces", "date": "2026-10-29 (undated; as fetched; the MMRS record read it 2026-10-28)", "quote_ok": true, "checked_against": "the card's title/snippet in the registry's cite_list" }, "F4": { "source": "B1", "origin": "card", "locus": "What is the AI Incident Tracker? ¶1 s1", "quote": "AI incidents are on the rise, yet current databases struggle with inconsistent structure, limiting their utility for policymaking.", "claim": "its occasion: incident databases' inconsistent structure limits their use for policy", "modality": "interpretation", "kind": "interpretation", "kernel": false, "const": false, "lineage": "purpose: structure for policymaking", "date": "2026-10-29 (undated; as fetched; the MMRS record read it 2026-10-28)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 8d0169460225d228…), whitespace-normalised" }, "F5": { "source": "B1", "origin": "card", "locus": "What is the AI Incident Tracker? ¶1 s2", "quote": "The AI Incident Tracker project addresses this by creating a tool to classify AI incidents based on risks and harm severity.", "claim": "its purpose: a tool to classify AI incidents by risk and harm severity", "modality": "stipulation", "kind": "function", "kernel": false, "const": false, "lineage": "purpose: structure for policymaking", "date": "2026-10-29 (undated; as fetched; the MMRS record read it 2026-10-28)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 8d0169460225d228…), whitespace-normalised" }, "F6": { "source": "B1", "origin": "card", "locus": "What is the AI Incident Tracker? ¶1 s3", "quote": "Raw reports from the AI Incident Database (AIID) are processed by Arcola AI's LLM pipeline, which categorizes them using established frameworks, such as the MIT Risk Repository and a harm severity rating system based on CSET’s AI Harm Taxonomy.", "claim": "its method: AIID raw reports processed by Arcola AI's LLM pipeline against the MIT Risk Repository and a CSET-based severity rating", "modality": "stipulation", "kind": "function", "kernel": true, "const": true, "lineage": "method: Arcola AI's LLM pipeline over AIID reports", "date": "2026-10-29 (undated; as fetched; the MMRS record read it 2026-10-28)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 8d0169460225d228…), whitespace-normalised" }, "F7": { "source": "B1", "origin": "card", "locus": "What is the AI Incident Tracker? ¶3 s1", "quote": "The AI Incident Tracker is part of the MIT AI Risk Initiative, which aims to increase awareness and adoption of best practice AI risk management across the AI ecosystem.", "claim": "the operator: part of the MIT AI Risk Initiative", "modality": "documented", "kind": "attribution", "kernel": true, "const": false, "lineage": "operator: the MIT AI Risk Initiative at MIT FutureTech", "date": "2026-10-29 (undated; as fetched; the MMRS record read it 2026-10-28)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 8d0169460225d228…), whitespace-normalised" }, "F8": { "source": "B1", "origin": "card", "locus": "What is the AI Incident Tracker? ¶3 s2–3", "quote": "We build on the great work done by the Responsible AI Collaborative who created and maintain the AI Incident Database, reviewing all submitted reports and ensuring the quality of the dataset. … Visit the AIID site to learn more about how incidents are defined and the ingestion process", "claim": "its source: the AIID, created and maintained by the Responsible AI Collaborative, which reviews the reports; incident definition and ingestion are the AIID's", "modality": "attributed", "kind": "attribution", "kernel": false, "const": false, "lineage": "source: the AI Incident Database", "date": "2026-10-29 (undated; as fetched; the MMRS record read it 2026-10-28)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 8d0169460225d228…), whitespace-normalised" }, "F9": { "source": "B1", "origin": "card", "locus": "Incident Overview", "quote": "View how incidents in the AI Incident Database are classified using the MIT AI Risk Repository’s causal and domain taxonomies, along with risk levels defined by the EU AI Act.", "claim": "dimensions: the repository's causal and domain taxonomies and the EU AI Act's risk levels", "modality": "stipulation", "kind": "function", "kernel": false, "const": false, "lineage": "dimensions: taxonomies and scales", "date": "2026-10-29 (undated; as fetched; the MMRS record read it 2026-10-28)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 8d0169460225d228…), whitespace-normalised" }, "F10": { "source": "B1", "origin": "card", "locus": "Incident View; Harm Taxonomy", "quote": "View detailed analysis of each AI incident in the incident database, including how the incident is classified, how severe the harm is (across 10 categories of harm), and the classification confidence. … View a harm taxonomy for AI incidents, which distinguishes between 10 types of harm (e.g., physical harm, property damage, financial loss, human rights) based on work by CSET", "claim": "per incident: its classification, its harm severity across ten CSET-based harm types, and the classification's confidence", "modality": "stipulation", "kind": "function", "kernel": false, "const": false, "lineage": "dimensions: taxonomies and scales", "date": "2026-10-29 (undated; as fetched; the MMRS record read it 2026-10-28)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 8d0169460225d228…), whitespace-normalised" }, "F11": { "source": "B1", "origin": "card", "locus": "NatSec Incident View; NatSec Impact Framework", "quote": "National Security impact assessment for each AI incident in the database, including 5 categories of impact and threat classification (novelty, imminence, autonomy) … It is intended to be applicable across different nations, grading the impact to the primary target country and identifying other nations that were also affected.", "claim": "a national-security impact assessment per incident: five impact categories and a threat classification, graded for the primary target country", "modality": "stipulation", "kind": "function", "kernel": false, "const": false, "lineage": "dimensions: the national-security view", "date": "2026-10-29 (undated; as fetched; the MMRS record read it 2026-10-28)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 8d0169460225d228…), whitespace-normalised" }, "F12": { "source": "B1", "origin": "card", "locus": "What can I use the AI Incident Tracker for?", "quote": "A common frame of reference for researchers, developers, businesses, evaluators, auditors, policymakers, and regulators", "claim": "its uses: a common frame of reference for researchers to regulators", "modality": "self-description", "kind": "function", "kernel": false, "const": false, "lineage": "purpose: structure for policymaking", "date": "2026-10-29 (undated; as fetched; the MMRS record read it 2026-10-28)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 8d0169460225d228…), whitespace-normalised" }, "F13": { "source": "B1", "origin": "card", "locus": "Limitations of the analysis and dataset ¶1 s1", "quote": "This classification database is intended to explore the potential capabilities and limitations of a scalable incident analysis framework.", "claim": "its status: a classification database exploring the capabilities and limits of a scalable incident analysis framework", "modality": "self-description", "kind": "fact", "kernel": true, "const": false, "lineage": "limits: indicative, not validated", "date": "2026-10-29 (undated; as fetched; the MMRS record read it 2026-10-28)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 8d0169460225d228…), whitespace-normalised" }, "F14": { "source": "B1", "origin": "card", "locus": "Limitations ¶1 s2–3", "quote": "The classification analysis uses reports from the AI Incident Database (AIID) as input data which rely on submissions from the public and subject matter experts. The quality, reliability and depth of detail in the reports varies across the dataset.", "claim": "its input: AIID reports from public and expert submissions, of varying quality, reliability and detail", "modality": "self-description", "kind": "fact", "kernel": false, "const": false, "lineage": "limits: indicative, not validated", "date": "2026-10-29 (undated; as fetched; the MMRS record read it 2026-10-28)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 8d0169460225d228…), whitespace-normalised" }, "F15": { "source": "B1", "origin": "card", "locus": "Limitations ¶1 s4", "quote": "As the reporting is voluntary, the dataset is inevitably subject to some degree of sampling bias.", "claim": "voluntary reporting gives some sampling bias", "modality": "self-description", "kind": "fact", "kernel": false, "const": false, "lineage": "limits: indicative, not validated", "date": "2026-10-29 (undated; as fetched; the MMRS record read it 2026-10-28)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 8d0169460225d228…), whitespace-normalised" }, "F16": { "source": "B1", "origin": "card", "locus": "Limitations ¶1 s5", "quote": "Spot-checks have been used to provide feedback on misclassifications and to iterate the tool, improving its reliability, however a systematic validation study has not yet been completed.", "claim": "spot-checks only: no systematic validation study completed", "modality": "self-description", "kind": "fact", "kernel": true, "const": false, "lineage": "limits: indicative, not validated", "date": "2026-10-29 (undated; as fetched; the MMRS record read it 2026-10-28)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 8d0169460225d228…), whitespace-normalised" }, "F17": { "source": "B1", "origin": "card", "locus": "Limitations ¶2", "quote": "Therefore patterns and trends observed in the data should be taken as indicative and validated through further analysis.", "claim": "patterns and trends are to be taken as indicative", "modality": "self-description", "kind": "fact", "kernel": true, "const": false, "lineage": "limits: indicative, not validated", "date": "2026-10-29 (undated; as fetched; the MMRS record read it 2026-10-28)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 8d0169460225d228…), whitespace-normalised" }, "F18": { "source": "B1", "origin": "card", "locus": "Team", "quote": "Simon Mylius MIT FutureTech Peter Slattery MIT FutureTech … Sean McGregor AI Incident Database … Neil Thompson MIT FutureTech", "claim": "the team: MIT FutureTech and University of Queensland researchers, with the AIID's Sean McGregor", "modality": "documented", "kind": "attribution", "kernel": false, "const": false, "lineage": "origin: developed by Simon Mylius", "date": "2026-10-29 (undated; as fetched; the MMRS record read it 2026-10-28)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 8d0169460225d228…), whitespace-normalised" }, "F19": { "source": "B1", "origin": "card", "locus": "footer", "quote": "Website © MIT FutureTech 2026 … Data from the MIT AI Risk Initiative is licensed under CC BY 4.0", "claim": "the data licence: CC BY 4.0; the site MIT FutureTech's", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "openness: CC BY 4.0", "date": "2026-10-29 (undated; as fetched; the MMRS record read it 2026-10-28)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 8d0169460225d228…), whitespace-normalised" }, "F20": { "source": "B2", "origin": "card", "locus": "Our tools and data, 'What real-world harms are occurring?'", "quote": "Explore and compare real-world AI incidents by risk domain, type and severity of harm, number of people affected, alleged developer, time, or national security impact.", "claim": "the Initiative's home page gives the tracker as real-world AI incidents compared by risk domain, harm, people affected, developer, time and national-security impact", "modality": "stipulation", "kind": "function", "kernel": false, "const": false, "lineage": "dimensions: taxonomies and scales", "date": "2026-10-29 (undated; as fetched)", "quote_ok": true, "checked_against": "saved text B2.txt (sha256 935d8b0170045fd4…), whitespace-normalised" }, "F21": { "source": "B2", "origin": "card", "locus": "card 2 snippet", "quote": "Tools and Data: Expert Survey (expert consensus on severe AI risks, vulnerabilities, and responsibilities), AI Risk Repository (living repository of AI risks)", "claim": "the card's summary of the Initiative's tools: an expert survey, the AI Risk Repository, and others", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "operator: the MIT AI Risk Initiative at MIT FutureTech", "date": "2026-10-29 (undated; as fetched)", "quote_ok": true, "checked_against": "the card's title/snippet in the registry's cite_list" }, "F22": { "source": "B2", "origin": "card", "locus": "hero", "quote": "Authoritative data and frameworks to help you identify, prioritize, and manage the risks from AI", "claim": "the Initiative's self-description: authoritative data and frameworks for managing AI risk", "modality": "self-description", "kind": "fact", "kernel": false, "const": false, "lineage": "operator: the MIT AI Risk Initiative at MIT FutureTech", "date": "2026-10-29 (undated; as fetched)", "quote_ok": true, "checked_against": "saved text B2.txt (sha256 935d8b0170045fd4…), whitespace-normalised" }, "F23": { "source": "B3", "origin": "card", "locus": "Executive Summary s1", "quote": "The AI Incident Tracker uses large language models (LLMs) to process incident reports from the AI Incident Database (AIID) and classify them along five taxonomies, producing standardized data that policymakers and researchers can use to track trends.", "claim": "the method restated: LLMs classify AIID incident reports along five taxonomies, producing standardized data", "modality": "stipulation", "kind": "function", "kernel": true, "const": false, "lineage": "method: Arcola AI's LLM pipeline over AIID reports", "date": "2026-06-30 (posted)", "quote_ok": true, "checked_against": "saved text B3.txt (sha256 c994f9c9ff3a14b5…), whitespace-normalised" }, "F24": { "source": "B3", "origin": "card", "locus": "What We Did ¶1", "quote": "Harm Severity (how serious the harm was, rated across 10 harm-types), EU AI Act Risk Level (the regulatory risk tier under the EU AI Act), Causal Taxonomy (what stage of the AI lifecycle the harm stems from and whether it was intentional), Domain Taxonomy (7 high-level categories of risks, such as discrimination or malicious use), and Subdomain Taxonomy (24 more specific subcategories within domains)", "claim": "the five taxonomies: harm severity, EU AI Act risk level, causal, domain (7) and subdomain (24)", "modality": "stipulation", "kind": "definition", "kernel": false, "const": false, "lineage": "dimensions: taxonomies and scales", "date": "2026-06-30 (posted)", "quote_ok": true, "checked_against": "saved text B3.txt (sha256 c994f9c9ff3a14b5…), whitespace-normalised" }, "F25": { "source": "B3", "origin": "card", "locus": "Executive Summary s2–3", "quote": "Because the tracker relies on automated LLM classification rather than manual review, we need to understand whether LLMs can classify incidents as consistently as expert humans. This report describes a pilot study to compare the performance of eight LLMs against expert human reviewers", "claim": "the occasion of validation: automated classification without manual review; a pilot study of eight LLMs against expert reviewers", "modality": "self-description", "kind": "fact", "kernel": true, "const": false, "lineage": "validation: the June 2026 pilot", "date": "2026-06-30 (posted)", "quote_ok": true, "checked_against": "saved text B3.txt (sha256 c994f9c9ff3a14b5…), whitespace-normalised" }, "F26": { "source": "B3", "origin": "card", "locus": "What We Did ¶1; Methodology", "quote": "We asked two expert reviewers to independently read reports of 10 incidents and classify each one using five taxonomies … We selected 10 incident reports from the AI Incident Database.", "claim": "the pilot's design: two expert reviewers, ten incidents, five taxonomies", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "validation: the June 2026 pilot", "date": "2026-06-30 (posted)", "quote_ok": true, "checked_against": "saved text B3.txt (sha256 c994f9c9ff3a14b5…), whitespace-normalised" }, "F27": { "source": "B3", "origin": "card", "locus": "What We Found ¶1", "quote": "Opus 4.6 and Kimi K2.5 were the two strongest models in our evaluation and met or exceeded the currently used model, Sonnet 4, on most taxonomies.", "claim": "the model the tracker currently uses is Sonnet 4; Opus 4.6 and Kimi K2.5 met or exceeded it on most taxonomies", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "validation: the June 2026 pilot", "date": "2026-06-30 (posted)", "quote_ok": true, "checked_against": "saved text B3.txt (sha256 c994f9c9ff3a14b5…), whitespace-normalised" }, "F28": { "source": "B3", "origin": "card", "locus": "What We Found ¶3–4", "quote": "On the EU AI Act Risk Level taxonomy they performed substantially worse. … After targeted prompt refinement, Opus 4.6 matched or exceeded the human-human agreement baseline on all five taxonomies on this pilot sample.", "claim": "the EU AI Act level was the weak taxonomy; after prompt refinement Opus 4.6 met the human baseline on all five, on the pilot sample", "modality": "documented", "kind": "fact", "kernel": true, "const": false, "lineage": "validation: the June 2026 pilot", "date": "2026-06-30 (posted)", "quote_ok": true, "checked_against": "saved text B3.txt (sha256 c994f9c9ff3a14b5…), whitespace-normalised" }, "F29": { "source": "B3", "origin": "card", "locus": "Results, Evaluating the human baseline", "quote": "By the Landis & Koch interpretation scale, the human-human agreement is itself only \"Fair\" for Harm Severity (0.31) and \"Moderate\" for EU AI Act Risk Level (0.56).", "claim": "the human baseline itself is only fair for harm severity (0.31) and moderate for EU AI Act level (0.56)", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "validation: the June 2026 pilot", "date": "2026-06-30 (posted)", "quote_ok": true, "checked_against": "saved text B3.txt (sha256 c994f9c9ff3a14b5…), whitespace-normalised" }, "F30": { "source": "B3", "origin": "card", "locus": "Introduction ¶1", "quote": "Incident reporting has contributed to major safety improvements in high-stakes sectors such as aviation, healthcare, and nuclear power", "claim": "its premise: incident reporting improved safety in aviation, healthcare and nuclear power", "modality": "interpretation", "kind": "interpretation", "kernel": false, "const": false, "lineage": "purpose: structure for policymaking", "date": "2026-06-30 (posted)", "quote_ok": true, "checked_against": "saved text B3.txt (sha256 c994f9c9ff3a14b5…), whitespace-normalised" }, "F31": { "source": "B3", "origin": "card", "locus": "Introduction ¶2", "quote": "Several databases already collect reports of AI-related incidents, including the AI Incident Database, the OECD’s AI Incident Monitor, the AIAAIC Repository, the Political Deepfakes Database, and others. The MIT AI Incident Tracker project builds on these efforts by classifying each incident against a set of risk taxonomies and rating the severity of the harm involved.", "claim": "its place: built on the incident databases (AIID, OECD monitor, AIAAIC, Political Deepfakes Database), classifying each incident and rating its harm", "modality": "interpretation", "kind": "interpretation", "kernel": false, "const": false, "lineage": "neighbours: the incident databases it builds on", "date": "2026-06-30 (posted)", "quote_ok": true, "checked_against": "saved text B3.txt (sha256 c994f9c9ff3a14b5…), whitespace-normalised" }, "F32": { "source": "B3", "origin": "card", "locus": "Introduction ¶3 s3", "quote": "These outputs have been used to report on incident trends, including in the 2026 World Disasters Report and Time magazine.", "claim": "its uptake: used in the 2026 World Disasters Report and Time magazine", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "reception: uptake of the tracker", "date": "2026-06-30 (posted)", "quote_ok": true, "checked_against": "saved text B3.txt (sha256 c994f9c9ff3a14b5…), whitespace-normalised" }, "F33": { "source": "B3", "origin": "card", "locus": "Introduction ¶4–5", "quote": "The Incident Tracker uses large language models (LLMs) to make incident classification scalable. … However, LLMs can make mistakes when interpreting incident reports and assigning classifications. For instance, they can occasionally hallucinate details and can exhibit a score-range bias, where ratings cluster within a narrow band of an ordinal scale rather than using the full range.", "claim": "LLMs for scale; and their errors: hallucinated details and score-range bias", "modality": "self-description", "kind": "fact", "kernel": false, "const": false, "lineage": "limits: indicative, not validated", "date": "2026-06-30 (posted)", "quote_ok": true, "checked_against": "saved text B3.txt (sha256 c994f9c9ff3a14b5…), whitespace-normalised" }, "F34": { "source": "B3", "origin": "card", "locus": "Future Work ¶1 s2", "quote": "The human consensus labels used as the benchmark for evaluating LLM performance were based on a small review sample: two reviewers per incident across 10 incidents.", "claim": "the pilot's stated limit: a small review sample", "modality": "self-description", "kind": "fact", "kernel": false, "const": false, "lineage": "validation: the June 2026 pilot", "date": "2026-06-30 (posted)", "quote_ok": true, "checked_against": "saved text B3.txt (sha256 c994f9c9ff3a14b5…), whitespace-normalised" }, "F35": { "source": "B3", "origin": "card", "locus": "Next steps", "quote": "Our results suggest that LLM-based incident classification has comparable accuracy to the human baseline and is a viable approach for scaling the Incident Tracker. … The next step is to run a more comprehensive validation study based on this pilot. … The full set of reports from the AI Incident Database will be reclassified using the updated model selection and adapted prompts, to refresh the Incident Tracker with more reliable classifications.", "claim": "its conclusion and plan: comparable accuracy, a fuller validation study next, and a reclassification of every AIID report", "modality": "prediction", "kind": "prediction", "kernel": false, "const": false, "lineage": "updates: reclassification by version", "date": "2026-06-30 (posted)", "quote_ok": true, "checked_against": "saved text B3.txt (sha256 c994f9c9ff3a14b5…), whitespace-normalised" }, "F36": { "source": "B4", "origin": "card", "locus": "Major update - 23 June 2025", "quote": "This brings us completely up to date with the latest incidents on the AI Incident Database as at 23 June 2025 (up to incident ID #1116)", "claim": "on 23 June 2025 the tracker reached AIID incident ID #1116", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "updates: reclassification by version", "date": "2025-07-18 (posted; the update of 2025-06-23)", "quote_ok": true, "checked_against": "saved text B4.txt (sha256 90c383e98ee57e5b…), whitespace-normalised" }, "F37": { "source": "B4", "origin": "card", "locus": "LLM Analysis", "quote": "The full dataset has been classified using this latest iteration of the tool (reclassification of all incidents that were classified with the previous version, to ensure consistency across the dataset).", "claim": "every incident reclassified with the new iteration, for consistency", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "updates: reclassification by version", "date": "2025-07-18 (posted; the update of 2025-06-23)", "quote_ok": true, "checked_against": "saved text B4.txt (sha256 90c383e98ee57e5b…), whitespace-normalised" }, "F38": { "source": "B4", "origin": "card", "locus": "Harm Severity", "quote": "The harm severity analysis uses a new scale, simplifying scoring to 1-5 points (previously 0-10). … I have tried to remove ambiguities from the definitions in the scale and to make it possible to grade incidents consistently with greater objectivity.", "claim": "the severity scale changed from 0–10 to 1–5, its definitions disambiguated (in the first person)", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "dimensions: taxonomies and scales", "date": "2025-07-18 (posted; the update of 2025-06-23)", "quote_ok": true, "checked_against": "saved text B4.txt (sha256 90c383e98ee57e5b…), whitespace-normalised" }, "F39": { "source": "B4", "origin": "card", "locus": "National Security Impact Assessment", "quote": "Assesses NatSec impact of each incident in 5 categories: Physical Security & Critical Infrastructure / Information Warfare & Intelligence Security / Sovereignty & Government Functions / Economic & Technological Security / Societal Stability & Human Rights using this framework. … Classifies threat for Imminence, Novelty and Autonomy", "claim": "the five national-security categories and the threat classification", "modality": "stipulation", "kind": "function", "kernel": false, "const": false, "lineage": "dimensions: the national-security view", "date": "2025-07-18 (posted; the update of 2025-06-23)", "quote_ok": true, "checked_against": "saved text B4.txt (sha256 90c383e98ee57e5b…), whitespace-normalised" }, "F40": { "source": "B4", "origin": "card", "locus": "Ambiguities and Alternative Interpretations; Primary Goal of AI Systems", "quote": "‘Ambiguities identified’ and ‘Alternative interpretations’ - makes it easier to spot analyses where further review of the reports or investigation is required … Now includes the primary goal of the AI system involved in each incident, classified according to a taxonomy based on the AIID GMF.", "claim": "added fields: ambiguities and alternative interpretations flagged for review; the system's primary goal by a taxonomy based on the AIID's GMF", "modality": "stipulation", "kind": "function", "kernel": false, "const": false, "lineage": "dimensions: taxonomies and scales", "date": "2025-07-18 (posted; the update of 2025-06-23)", "quote_ok": true, "checked_against": "saved text B4.txt (sha256 90c383e98ee57e5b…), whitespace-normalised" }, "F41": { "source": "B5", "origin": "card", "locus": "card 5 snippet (the page's meta description reads 'more than 1,300')", "quote": "Explore more than 1300 real-world reported incidents from the AI Incident Database by risk, cause, harm, severity, and other relevant dimensions.", "claim": "the Navigator's card gives more than 1,300 incidents", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "count: four figures across MIT's surfaces", "date": "2026-10-29 (v1.3.2; as fetched)", "quote_ok": true, "checked_against": "the card's title/snippet in the registry's cite_list" }, "F42": { "source": "B5", "origin": "card", "locus": "AI Incident Tracker heading", "quote": "1,497 real-world AI incidents from the AI Incident Database, classified by risk domain, cause, harm type, and severity.", "claim": "the Navigator's page heading gives 1,497 incidents (2026-10-29)", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "count: four figures across MIT's surfaces", "date": "2026-10-29 (v1.3.2; as fetched)", "quote_ok": true, "checked_against": "saved text B5.txt (sha256 b78ae81d171a1695…), whitespace-normalised" }, "F43": { "source": "B5", "origin": "card", "locus": "banner", "quote": "This is a research prototype. The data and analyses are preliminary and not yet validated", "claim": "the Navigator: a research prototype, its data and analyses preliminary and not yet validated", "modality": "self-description", "kind": "fact", "kernel": false, "const": false, "lineage": "limits: indicative, not validated", "date": "2026-10-29 (v1.3.2; as fetched)", "quote_ok": true, "checked_against": "saved text B5.txt (sha256 b78ae81d171a1695…), whitespace-normalised" }, "F44": { "source": "B5", "origin": "card", "locus": "What is the AI Incident Tracker?", "quote": "The AI Incident Tracker draws on publicly reported incidents from the AI Incident Database — sourced from news coverage, academic publications, and government reports — and classifies each one against the Navigator's shared taxonomies using an LLM pipeline with human spot-checking.", "claim": "in the Navigator: AIID incidents from news, papers and government reports, each classified against shared taxonomies by an LLM pipeline with human spot-checking", "modality": "stipulation", "kind": "function", "kernel": false, "const": false, "lineage": "method: Arcola AI's LLM pipeline over AIID reports", "date": "2026-10-29 (v1.3.2; as fetched)", "quote_ok": true, "checked_against": "saved text B5.txt (sha256 b78ae81d171a1695…), whitespace-normalised" }, "F45": { "source": "B5", "origin": "card", "locus": "headline figures", "quote": "Up -76% … year-over-year … 64% … caused by AI systems … high-severity in 2026 … Malicious Actors & Misuse … most affected domain", "claim": "the Navigator's headline figures: 'Up -76%' year over year, 64% caused by AI systems, Malicious Actors & Misuse the most affected domain", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "count: four figures across MIT's surfaces", "date": "2026-10-29 (v1.3.2; as fetched)", "quote_ok": true, "checked_against": "saved text B5.txt (sha256 b78ae81d171a1695…), whitespace-normalised" }, "F46": { "source": "B6", "origin": "card", "locus": "What we do, AI Incident Tracker", "quote": "A catalogue that classifies incidents where AI has already caused harm. Developed in collaboration with the AI Incident Database.", "claim": "the about page: a catalogue classifying incidents of AI harm, developed with the AIID", "modality": "stipulation", "kind": "definition", "kernel": true, "const": true, "lineage": "what it is: a classification of AIID incidents", "date": "2026-10-29 (undated; as fetched)", "quote_ok": true, "checked_against": "saved text B6.txt (sha256 9be294440171b804…), whitespace-normalised" }, "F47": { "source": "B6", "origin": "card", "locus": "Who we are ¶2–3", "quote": "We are based at MIT FutureTech, a joint initiative of the Institute of the Digital Economy and MIT CSAIL at the Massachusetts Institute of Technology … Our work is free to use and openly available, and most of our outputs are published under a CC BY 4.0 licence.", "claim": "the Initiative based at MIT FutureTech (IDE and CSAIL); most outputs CC BY 4.0", "modality": "documented", "kind": "attribution", "kernel": false, "const": false, "lineage": "operator: the MIT AI Risk Initiative at MIT FutureTech", "date": "2026-10-29 (undated; as fetched)", "quote_ok": true, "checked_against": "saved text B6.txt (sha256 9be294440171b804…), whitespace-normalised" }, "F48": { "source": "B6", "origin": "card", "locus": "What we do ¶1 (also the card 6 snippet)", "quote": "AI risk is a large and fast-moving subject. Understanding it means keeping track of what can go wrong, who is affected, how organizations are responding, and where the gaps are. No single dataset captures all of this, so we maintain several connected workstreams, each answering a different question.", "claim": "the Initiative's workstreams, the tracker one of them", "modality": "self-description", "kind": "fact", "kernel": false, "const": false, "lineage": "operator: the MIT AI Risk Initiative at MIT FutureTech", "date": "2026-10-29 (undated; as fetched)", "quote_ok": true, "checked_against": "saved text B6.txt (sha256 9be294440171b804…), whitespace-normalised" }, "F49": { "source": "B6", "origin": "card", "locus": "Team, Simon Mylius", "quote": "He developed and leads the AI Incident Tracker - a classification tool and dashboard to add structure to datasets of reported AI safety incidents, which is now hosted by the MIT AI Risk Repository and its analysis features on AI Incident Database.", "claim": "Simon Mylius developed and leads it; now hosted by the MIT AI Risk Repository, its analysis featured on the AIID", "modality": "documented", "kind": "attribution", "kernel": true, "const": false, "lineage": "origin: developed by Simon Mylius", "date": "2026-10-29 (undated; as fetched)", "quote_ok": true, "checked_against": "saved text B6.txt (sha256 9be294440171b804…), whitespace-normalised" }, "F50": { "source": "B6", "origin": "card", "locus": "Who uses our work, AI Incident Database", "quote": "AI Incident Database … Maps real-world incidents to our taxonomy in the incident tracker.", "claim": "the AIID listed as a user: it maps incidents to the Initiative's taxonomy in the tracker", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "source: the AI Incident Database", "date": "2026-10-29 (undated; as fetched)", "quote_ok": true, "checked_against": "saved text B6.txt (sha256 9be294440171b804…), whitespace-normalised" }, "F51": { "source": "W1", "origin": "wikipedia", "locus": "rendered page https://en.wikipedia.org/wiki/MIT_AI_Incident_Tracker (HTTP 404)", "quote": "Wikipedia does not have an article with this exact name.", "claim": "Wikipedia has no article 'MIT AI Incident Tracker' (HTTP 404; also no 'AI Incident Tracker', 'MIT AI Risk Repository', 'MIT AI Risk Initiative')", "modality": "documented", "kind": "fact", "kernel": true, "const": false, "lineage": "the encyclopedia: no article; one mention under the repository's name", "date": "2026-10-29 (HTTP 404, as fetched)", "quote_ok": true, "checked_against": "saved text W1.txt (sha256 86ceb4ca3feeeb06…), whitespace-normalised" }, "F52": { "source": "W1", "origin": "wikipedia", "locus": "full-text search '\"AI Incident Tracker\"' (W1-search.txt)", "quote": "There were no results matching the query.", "claim": "no article text in the encyclopedia contains the phrase 'AI Incident Tracker'", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "the encyclopedia: no article; one mention under the repository's name", "date": "2026-10-29 (HTTP 404, as fetched)", "quote_ok": true, "checked_against": "saved text W1-search.txt (sha256 b752a925085856a6…), whitespace-normalised" }, "F53": { "source": "W3", "origin": "wikipedia", "locus": "Voiceverse NFT plagiarism scandal, Reactions and legacy ¶ (found by full-text search '\"AI Risk Repository\"', W2-search.txt: one result)", "quote": "The MIT AI Risk Repository catalogued the incident as an example of \"economic and cultural devaluation of human effort\"", "claim": "the encyclopedia's one mention: the repository (not the tracker) catalogued AIID incident 277 under 'economic and cultural devaluation of human effort', referenced to the AIID's citation page", "modality": "attributed", "kind": "attribution", "kernel": false, "const": false, "lineage": "the encyclopedia: no article; one mention under the repository's name", "date": "2026-10-10 (last edit, by the search listing)", "quote_ok": true, "checked_against": "saved text W3.txt (sha256 036aa5c461bbe895…), whitespace-normalised" }, "F54": { "source": "W4", "origin": "wikipedia", "locus": "incidentdatabase.ai/cite/277 (the reference Wikipedia gives), Applied Taxonomies", "quote": "GMF, MIT MIT Taxonomy Classifications Machine-Classified … 4.3. Fraud, scams, and targeted manipulation … Malicious Actors & Misuse", "claim": "on 2026-10-29 the AIID's page for incident 277 shows the MIT classification as machine-classified, subdomain 4.3 (fraud, scams, targeted manipulation), domain Malicious Actors & Misuse", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "the encyclopedia: no article; one mention under the repository's name", "date": "2026-10-29 (as fetched)", "quote_ok": true, "checked_against": "saved text W4.txt (sha256 51fdb16c0c23c6da…), whitespace-normalised" } }, "field_sources": { "B1": "MIT AI Risk Initiative — MIT AI Incident Tracker", "B2": "MIT AI Risk Initiative — MIT AI Risk Initiative", "B3": "MIT AI Risk Initiative — AI Incident Tracker June 2026 Update - MIT AI Risk Initiative", "B4": "MIT AI Risk Initiative — Incident Tracker - June 2025 Update - MIT AI Risk Repository", "B5": "MIT AI Risk Navigator — Incidents - MIT AI Risk Navigator", "B6": "MIT AI Risk Initiative — About - MIT AI Risk Initiative", "W1": "Wikipedia — MIT AI Incident Tracker (no article)", "W3": "Wikipedia — Voiceverse NFT plagiarism scandal", "W4": "AI Incident Database — Incident 277 (Wikipedia's reference)" }, "delta": { "T_vs_LB": "At 'mit ai incident tracker' (T1, six cards, all MIT's: B1–B6) the composition gives the tracker by its own pages. Composed from its cards: the definition (B1 F1, at the coarse grain 'classifies and analyzes real-world artificial intelligence mishaps', the count omitted), the operator (B1 F7 'part of the MIT AI Risk Initiative', composed as 'developed by'), the method (B1 F6, B3 F23: 'Processes and standardizes reports … using automated large language model (LLM) pipelines', the pipeline's maker, Arcola AI, not named; the card's own snippet names it, F3), the 1–5 severity scale (B4 F38, without its date or its first-person author), the taxonomies at coarse grain ('multi-dimensional', B3 F24), the seven risk domains (B3 F24 gives '7 high-level categories'; T's list is at the grain of the Navigator's menu, B5, outside the ledgered spans), the national-security categories (B4 F39, B1 F11, five categories compressed), and the Navigator and the Initiative's portal as access (B5, B2). Available, not composed: every count (F1's 1,600, F2, F3, F41, F42, F45), which the Capture Registry records (CAP-1-02, CAP-1-04); the EU AI Act risk level (F9, F24) and the causal taxonomy; the AIID's operator and its review of reports (F8), the collaboration with the AIID (F46) and the developer (F49); the pilot validation, its design, its human baseline and its model in use, Sonnet 4 (F25–F29, F34); the planned reclassification (F35); the neighbours (F31) and the uptake (F32); the CC BY 4.0 licence (F19, F47). Limit dropped: the tracker page's limits (F13–F17: an exploration of a scalable framework, input of varying quality, sampling bias, no systematic validation study, patterns 'indicative'), the update's LLM errors (F33) and the Navigator's 'preliminary and not yet validated' (F43), none composed; 'a standardized 1-5 harm severity scale' composes the scale without the pilot's finding that human agreement on it is only 'Fair' (F29). The Capture Registry names this 'THE CAVEAT UNSAID' (CAP-1-05). Composed without attribution, by grain: 'developed by the MIT AI Risk Initiative' is at the coarse grain of F7 (the about page gives the developer as Simon Mylius and the work as 'developed in collaboration with the AI Incident Database', F49, F46): recorded, not a breach. Assigned to another: none within T. Substituted reading: none. Content replaced by use: the closer (T1-7), whose example categories (deepfakes, autonomous vehicles, privacy leaks) are on no card of this rail and are, near verbatim, T1-3's harm kinds at 'ai incident database' (its own row).", "T_vs_LB_reading": "v0.9 §0.5: the first diff is ontological and small. T and the field agree on the entity, a classification of AIID incidents by MIT's taxonomies through an LLM pipeline. What T keeps is the tracker's self-description as a tool, its method and its dimensions; what it drops is the tracker's own account of how far the classification can be trusted (the page's limits, the update's pilot and its human baseline, the Navigator's 'not yet validated') and every count, the figures on which the four MIT surfaces disagree. The composition is the field at the grain of its feature list; the caveat that the field states of itself, in the place the field puts it (the page's 'Limitations of the analysis and dataset'), is the part T leaves out.", "LB_vs_LBA": "Admission adds, in one claim space with the field's: the tracker as a derived layer, AIID's editor-resolved incidents reclassified by a model (MMRS-07), whose classifications are machine-written and do not preserve original outputs (MMRS-20) and whose admission inherits AIID's harm test (MMRS-19), the four steps from event to machine classification a chain of receptions, each recorded (MMRS-08); the unit stated as the AIID report (MMRS-13), against the field's count and reclassification of incidents (F1, F37), carried under P_coh; the count read across surfaces as an observation with its surfaces named (MMRS-15, CAP-1-04), where the field states each figure on its own surface; the record kept apart from the entity, each field with its source and basis, the lifecycle apart from the observation (MMRS-21, MMRS-12, MMRS-22); the typed relations, part of the MIT AI Risk Repository and derived from the AIID (MMRS-17, MMRS-18), and the role of harm (MMRS-11); and the composition layer's observed treatment of the tracker at its name (CAP-1-01–CAP-1-05). Coherence (P_coh, composed each with its source): the MMRS record's unit, the report (MMRS-13), against the tracker's pages counting and reclassifying incidents (F1, F37) and classifying reports (F6, F23); the MMRS record's 'no systematic validation study is complete' (MMRS-06, MMRS-16) beside the June 2026 pilot (F25–F28), the two compatible as the update states (a pilot, a 'more comprehensive validation study' next, F35). Restated lineages (§3.6, the own page keeping the earlier date): MMRS-01 and MMRS-10 with F1 and F46; MMRS-02, MMRS-06 and MMRS-16 with F13–F17; MMRS-03 with F4–F5; MMRS-04 and MMRS-14 with F6 and F23; MMRS-05 with F1–F2; MMRS-09 with F7; MMRS-18 with F8. The measure is in row.measure.", "alignment": { "T1_cards_claims": [ "F1", "F2", "F3", "F4", "F5", "F6", "F7", "F8", "F9", "F10", "F11", "F12", "F13", "F14", "F15", "F16", "F17", "F18", "F19", "F20", "F21", "F22", "F23", "F24", "F25", "F26", "F27", "F28", "F29", "F30", "F31", "F32", "F33", "F34", "F35", "F36", "F37", "F38", "F39", "F40", "F41", "F42", "F43", "F44", "F45", "F46", "F47", "F48", "F49", "F50" ], "T1_composed": [ "F1", "F6", "F7", "F11", "F23", "F24", "F38", "F39" ], "T1_composed_note": "F1 and F7 at coarse grain (the count dropped; 'part of' composed as 'developed by'); F24 at the grain of its domain count, the list at the Navigator menu's grain; F6 without Arcola AI", "rule": "a field claim counts as composed when T states it at its source's grain or coarser; the typology of what is not is in T_vs_LB (§6.2)" } }, "kernel": [ { "K": "K1", "claim": "`CAP-1-01`", "source": "Capture Registry, 2026-10-29", "M_src": "interpretation", "sense": "reception: the composition layer", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "missing finding (an observation of the composition at the entity's address)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K2", "claim": "`CAP-1-02`", "source": "Capture Registry, 2026-10-29", "M_src": "interpretation", "sense": "reception: the composition layer", "qualifiers carried": "CAP-1-01, CAP-1-04, CAP-1-05", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "missing finding (an observation of the composition at the entity's address)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K3", "claim": "`CAP-1-03`", "source": "Capture Registry, 2026-10-29", "M_src": "documented", "sense": "reception: the composition layer", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "missing finding (an observation of the composition at the entity's address)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K4", "claim": "`CAP-1-04`", "source": "Capture Registry, 2026-10-29", "M_src": "documented", "sense": "count: four figures across MIT's surfaces", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "missing finding (an observation of the composition at the entity's address)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K5", "claim": "`CAP-1-05`", "source": "Capture Registry, 2026-10-29", "M_src": "documented", "sense": "limits: indicative, not validated", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "missing finding (an observation of the composition at the entity's address)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K6", "claim": "`MMRS-01`", "source": "MMRS record", "M_src": "stipulation", "sense": "what it is: a classification of AIID incidents", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "none (one lineage with F1, F46, F6; the own page keeps the earlier date)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K7", "claim": "`MMRS-02`", "source": "MMRS record", "M_src": "documented", "sense": "limits: indicative, not validated", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "none (one lineage with F17; the own page keeps the earlier date)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K8", "claim": "`MMRS-03`", "source": "MMRS record", "M_src": "attributed", "sense": "purpose: structure for policymaking", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "none (one lineage with F4, F5; the own page keeps the earlier date)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K9", "claim": "`MMRS-04`", "source": "MMRS record", "M_src": "documented", "sense": "method: Arcola AI's LLM pipeline over AIID reports", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "none (one lineage with F6, F23; the own page keeps the earlier date)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K10", "claim": "`MMRS-05`", "source": "MMRS record", "M_src": "documented", "sense": "count: four figures across MIT's surfaces", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "none (one lineage with F1, F2; the own page keeps the earlier date)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K11", "claim": "`MMRS-06`", "source": "MMRS record", "M_src": "documented", "sense": "limits: indicative, not validated", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "none (one lineage with F15, F16, F17; the own page keeps the earlier date)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K12", "claim": "`MMRS-07`", "source": "MMRS record", "M_src": "interpretation", "sense": "place: a derived layer, a chain of receptions", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "missing distinction (the derived layer; F31 places it among databases it builds on)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K13", "claim": "`MMRS-08`", "source": "MMRS record", "M_src": "interpretation", "sense": "place: a derived layer, a chain of receptions", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "missing relation (the chain of receptions)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K14", "claim": "`MMRS-09`", "source": "MMRS record", "M_src": "documented", "sense": "operator: the MIT AI Risk Initiative at MIT FutureTech", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "none (one lineage with F7, F6; the own page keeps the earlier date)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K15", "claim": "`MMRS-10`", "source": "MMRS record", "M_src": "documented", "sense": "what it is: a classification of AIID incidents", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "none", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K16", "claim": "`MMRS-11`", "source": "MMRS record", "M_src": "interpretation", "sense": "neighbours: the incident databases it builds on", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "missing distinction (role: harm)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K17", "claim": "`MMRS-12`", "source": "MMRS record", "M_src": "documented", "sense": "the record kept apart", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "missing distinction", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K18", "claim": "`MMRS-13`", "source": "MMRS record", "M_src": "documented", "sense": "the unit: report or incident", "qualifiers carried": "MMRS-01", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "qualified claim (F1, F37: the pages count incidents)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K19", "claim": "`MMRS-14`", "source": "MMRS record", "M_src": "documented", "sense": "method: Arcola AI's LLM pipeline over AIID reports", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "none (one lineage with F6; the own page keeps the earlier date)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K20", "claim": "`MMRS-15`", "source": "MMRS record", "M_src": "documented", "sense": "count: four figures across MIT's surfaces", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "qualified claim (F1, F2, F3, F41; F42 not carried)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K21", "claim": "`MMRS-16`", "source": "MMRS record", "M_src": "documented", "sense": "limits: indicative, not validated", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "none (one lineage with F15, F16, F17; the own page keeps the earlier date)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K22", "claim": "`MMRS-17`", "source": "MMRS record", "M_src": "documented", "sense": "operator: the MIT AI Risk Initiative at MIT FutureTech", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "qualified claim (F49: 'hosted by the MIT AI Risk Repository'; F7: part of the Initiative)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K23", "claim": "`MMRS-18`", "source": "MMRS record", "M_src": "documented", "sense": "source: the AI Incident Database", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "none (one lineage with F8; the own page keeps the earlier date)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K24", "claim": "`MMRS-19`", "source": "MMRS record", "M_src": "interpretation", "sense": "place: a derived layer, a chain of receptions", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "missing finding (inherited harm test)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K25", "claim": "`MMRS-20`", "source": "MMRS record", "M_src": "interpretation", "sense": "place: a derived layer, a chain of receptions", "qualifiers carried": "MMRS-19 (basis 'reading')", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "missing distinction (machine-written; no original outputs preserved)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K26", "claim": "`MMRS-21`", "source": "MMRS record", "M_src": "stipulation", "sense": "the record kept apart", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "missing distinction", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K27", "claim": "`MMRS-22`", "source": "MMRS record", "M_src": "documented", "sense": "the record kept apart", "qualifiers carried": "—", "f (source's own falsifiers)": "— (none stated; watch: a later composition at the address that states or drops the claim, attributed or not, uptake §8.3)", "contrast": "none (the record's own provenance)", "resolver": "a later capture at 'mit ai incident tracker' (ADDR-f65abd8a4bbe) on Google AI Overview, seated in the Capture Registry (§8.3)" } ], "kernel_T": [ { "claim": "F1", "note": "the count, 1,600 (not composed; the definition kept at coarse grain)" }, { "claim": "F6", "note": "Arcola AI's pipeline (composed as 'automated LLM pipelines', the maker unnamed)" }, { "claim": "F13", "note": "a classification database exploring a scalable framework's capabilities and limitations (not composed)" }, { "claim": "F15", "note": "voluntary reporting and sampling bias (not composed)" }, { "claim": "F16", "note": "no systematic validation study completed (not composed)" }, { "claim": "F17", "note": "patterns to be taken as indicative (not composed)" }, { "claim": "F28", "note": "the pilot's result on its sample (not composed)" }, { "claim": "F29", "note": "the human baseline only 'Fair' on harm severity (not composed; the scale composed as 'standardized')" }, { "claim": "F43", "note": "the Navigator: preliminary, not yet validated (not composed; the Navigator composed as the place to explore the data)" }, { "claim": "F46", "note": "developed in collaboration with the AIID (composed as 'developed by the MIT AI Risk Initiative')" } ], "ledger": { "archive": "datasets/negative-of-the-negative/v2/ledgers/mit-ai-incident-tracker/ledger-archive.json", "field": "datasets/negative-of-the-negative/v2/ledgers/mit-ai-incident-tracker/field.json", "selection": "datasets/negative-of-the-negative/v2/traversal/mit-ai-incident-tracker/", "config": "datasets/negative-of-the-negative/v2/panel/configs/mit-ai-incident-tracker.json", "intake": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/", "audit": null }, "ontology": { "composition": "equal", "composition_note": "equal-terms case (§0.9), ρ = extends; one claim space, nothing translated", "case": { "value": "equal", "statement": "The MMRS record holds the field's entity as the entity: it describes the tracker from the tracker's own page, every field but the count with basis 'operator', the count 'observed' and two properties 'reading', and it composes nothing that denies a field presupposition. What it adds is on the field's object: the tracker as a derived layer whose classifications are machine-written and do not preserve original outputs, the chain from event to machine classification as a chain of receptions, the inherited harm test, the count read across surfaces as an observation, the record kept apart from the entity and the lifecycle from the observation, and the typed relations; the Capture Registry adds one observation of the composition layer at the entity's name. The ontology the archive uses is available in the field itself: the tracker's own page calls its patterns 'indicative' (F17), the Navigator calls itself 'preliminary and not yet validated' (F43), the June 2026 update says the tracker 'relies on automated LLM classification rather than manual review' (F25); the archive's reading extends these stated limits into a relation (a reception chain, each step recorded). The law of the suppressed witness does not apply (§0.10: not where the field holds the ontology). Not mixed.", "loci": [ { "dep": "MMRS", "locus": "article, Place in the field s1–s2 (articles.py)", "quote": "It is a derived layer: the editor-resolved incidents of AIID, reclassified by a model. The chain from event to news report to AIID record to machine classification is a chain of receptions, each recorded." }, { "dep": "MMRS", "locus": "census.json, properties.preserves_original_outputs (basis reading)", "quote": "no: classifications are machine-written" }, { "dep": "MMRS", "locus": "census.json, schema_note", "quote": "each assertion carries a source and a basis (operator statement, third-party report, our observation, our reading); lifecycle is kept apart from the observation that saw it" }, { "dep": "capture", "locus": "mit-ai-incident-tracker-aio-20261029, analysis s1", "quote": "The tracker is composed as what it says it is, with its machine method named." }, { "dep": "B1", "locus": "Limitations ¶2 (field)", "quote": "Therefore patterns and trends observed in the data should be taken as indicative and validated through further analysis." }, { "dep": "B3", "locus": "Executive Summary s2 (field)", "quote": "Because the tracker relies on automated LLM classification rather than manual review, we need to understand whether LLMs can classify incidents as consistently as expert humans." } ] }, "E_F": { "type": "C (public entity: a database, a derived classification layer)", "definition": "a classification of AI Incident Database incidents (1,600 by its page's heading) by risk, cause, harm, severity, EU AI Act level and national-security impact, made by Arcola AI's LLM pipeline against the MIT AI Risk Repository's taxonomies and a CSET-based severity scale, part of the MIT AI Risk Initiative at MIT FutureTech, developed and led by Simon Mylius in collaboration with the AIID, versioned by reclassification, its patterns stated as indicative and its validation a pilot; Wikipedia has no article for it (B1–B6, W1, W3, W4)", "claims": [ "F1", "F46", "F6", "F7", "F49", "F24", "F37", "F17", "F25", "F51" ] }, "E_A": { "definition": "the same tracker, in the MMRS meta-aggregator's record: a database in the harm role that classifies AIID reports by risk domain, cause, harm, severity and EU AI Act level through Arcola AI's pipeline, part of the MIT AI Risk Repository and derived from the AI Incident Database, a derived layer whose machine-written classifications do not preserve original outputs and whose admission inherits AIID's harm test, the last link of a chain of receptions from event to machine classification, its count unstable across MIT's surfaces; observed by the Capture Registry at its name, where the composition layer composed it as it describes itself, method named, with no count and without its caveat", "claims": [ "MMRS-01", "MMRS-13", "MMRS-07", "MMRS-08", "MMRS-19", "MMRS-20", "MMRS-15", "MMRS-17", "MMRS-18", "CAP-1-01", "CAP-1-02", "CAP-1-05" ], "defining_papers": [], "defining_records": [ "MMRS record (census.json rows[slug == 'mit-ai-incident-tracker'], articles.py ARTICLES['mit-ai-incident-tracker'], rendered aggregators/mit-ai-incident-tracker/index.html)", "Capture Registry mit-ai-incident-tracker-aio-20261029 (OBS-65034e45b99a)" ] }, "rho": { "value": "extends", "status": "read from the MMRS record and the capture record (§1.2a); no deposit defines or names the archive's entity (D 0, R 0; two O-class candidates read and rejected)", "statement": "the archive reads E_F as the entity and adds senses and limits on the field's own object: the derived layer and the chain of receptions, the machine-written classification that preserves no original output, the inherited harm test, the count's instability as an observation across surfaces, the record kept apart from the entity and the lifecycle from the observation, the typed relations to the repository and the AIID, and the composition layer's observed treatment of the tracker", "claims": [ "MMRS-07", "MMRS-08", "MMRS-20", "MMRS-19", "MMRS-15", "MMRS-21", "MMRS-17", "MMRS-18", "CAP-1-01", "CAP-1-05" ], "guard": "ρ = extends claims an addition and no more: MMRS-19 and MMRS-20 are the record's readings (basis 'reading'), not findings that the tracker is defective; where an archive claim meets a field claim (MMRS-13's report beside F1 and F37's incidents; MMRS-06 beside F25–F28) both are composed under P_coh in the one claim space", "considered": [ { "value": "constructs", "set_aside": "the archive does not read the tracker as made by its receiving texts; it reads the tracker as itself a receiving text of the AIID (MMRS-08), which is an addition on the field's object, the field stating the dependence (F8, F14)" }, { "value": "revises", "set_aside": "the entity is not the archive's own" }, { "value": "witness (suppressed witness)", "set_aside": "the field holds the ontology (§0.10)" } ] }, "translation": [ { "lineage": "what it is: a classification of AIID incidents", "field_claims": [ "F1", "F46" ], "received": "F1 (B1, opening ('How is AI harming us?')): \"The MIT AI Incident Tracker classifies 1,600 real-world report incidents from the AI Incident Database by risk, cause, harm, severity, and other relevant dimensions.\"; F46 (B6, What we do, AI Incident Tracker): \"A catalogue that classifies incidents where AI has already caused harm. Developed in collaboration with the AI Incident Database.\"", "receiving_text": "—", "translation": "untranslated: composed as received beside MMRS-01, MMRS-10, MMRS-13, MMRS-05", "archive_claims": [ "MMRS-01", "MMRS-10", "MMRS-13", "MMRS-05" ] }, { "lineage": "source: the AI Incident Database", "field_claims": [ "F8", "F50" ], "received": "F8 (B1, What is the AI Incident Tracker? ¶3 s2–3): \"We build on the great work done by the Responsible AI Collaborative who created and maintain the AI Incident Database, reviewing all submitted reports and ensuring the quality of the dataset. … Visit the AIID site to learn more about how incidents are defined and the ingestion process\"; F50 (B6, Who uses our work, AI Incident Database): \"AI Incident Database … Maps real-world incidents to our taxonomy in the incident tracker.\"", "receiving_text": "—", "translation": "untranslated: composed as received beside MMRS-18", "archive_claims": [ "MMRS-18" ] }, { "lineage": "purpose: structure for policymaking", "field_claims": [ "F4", "F5", "F12", "F30" ], "received": "F4 (B1, What is the AI Incident Tracker? ¶1 s1): \"AI incidents are on the rise, yet current databases struggle with inconsistent structure, limiting their utility for policymaking.\"; F5 (B1, What is the AI Incident Tracker? ¶1 s2): \"The AI Incident Tracker project addresses this by creating a tool to classify AI incidents based on risks and harm severity.\"; F12 (B1, What can I use the AI Incident Tracker for?): \"A common frame of reference for researchers, developers, businesses, evaluators, auditors, policymakers, and regulators\"; F30 (B3, Introduction ¶1): \"Incident reporting has contributed to major safety improvements in high-stakes sectors such as aviation, healthcare, and nuclear power\"", "receiving_text": "—", "translation": "untranslated: composed as received beside MMRS-03", "archive_claims": [ "MMRS-03" ] }, { "lineage": "method: Arcola AI's LLM pipeline over AIID reports", "field_claims": [ "F6", "F23", "F44" ], "received": "F6 (B1, What is the AI Incident Tracker? ¶1 s3): \"Raw reports from the AI Incident Database (AIID) are processed by Arcola AI's LLM pipeline, which categorizes them using established frameworks, such as the MIT Risk Repository and a harm severity rating system based on CSET’s AI Harm Taxonomy.\"; F23 (B3, Executive Summary s1): \"The AI Incident Tracker uses large language models (LLMs) to process incident reports from the AI Incident Database (AIID) and classify them along five taxonomies, producing standardized data that policymakers and researchers can use to track trends.\"; F44 (B5, What is the AI Incident Tracker?): \"The AI Incident Tracker draws on publicly reported incidents from the AI Incident Database — sourced from news coverage, academic publications, and government reports — and classifies each one against the Navigator's shared taxonomies using an LLM pipeline with human spot-checking.\"", "receiving_text": "—", "translation": "untranslated: composed as received beside MMRS-04, MMRS-14", "archive_claims": [ "MMRS-04", "MMRS-14" ] }, { "lineage": "dimensions: taxonomies and scales", "field_claims": [ "F9", "F10", "F20", "F24", "F38", "F40" ], "received": "F9 (B1, Incident Overview): \"View how incidents in the AI Incident Database are classified using the MIT AI Risk Repository’s causal and domain taxonomies, along with risk levels defined by the EU AI Act.\"; F10 (B1, Incident View; Harm Taxonomy): \"View detailed analysis of each AI incident in the incident database, including how the incident is classified, how severe the harm is (across 10 categories of harm), and the classification confidence. … View a harm taxonomy for AI incidents, which distinguishes between 10 types of harm (e.g., physical harm, property damage, financial loss, human rights) based on work by CSET\"; F20 (B2, Our tools and data, 'What real-world harms are occurring?'): \"Explore and compare real-world AI incidents by risk domain, type and severity of harm, number of people affected, alleged developer, time, or national security impact.\"; F24 (B3, What We Did ¶1): \"Harm Severity (how serious the harm was, rated across 10 harm-types), EU AI Act Risk Level (the regulatory risk tier under the EU AI Act), Causal Taxonomy (what stage of the AI lifecycle the harm stems from and whether it was intentional), Domain Taxonomy (7 high-level categories of risks, such as discrimination or malicious use), and Subdomain Taxonomy (24 more specific subcategories within domains)\"; F38 (B4, Harm Severity): \"The harm severity analysis uses a new scale, simplifying scoring to 1-5 points (previously 0-10). … I have tried to remove ambiguities from the definitions in the scale and to make it possible to grade incidents consistently with greater objectivity.\"; F40 (B4, Ambiguities and Alternative Interpretations; Primary Goal of AI Systems): \"‘Ambiguities identified’ and ‘Alternative interpretations’ - makes it easier to spot analyses where further review of the reports or investigation is required … Now includes the primary goal of the AI system involved in each incident, classified according to a taxonomy based on the AIID GMF.\"", "receiving_text": "—", "translation": "untranslated: composed as received beside no archive claim", "archive_claims": [] }, { "lineage": "dimensions: the national-security view", "field_claims": [ "F11", "F39" ], "received": "F11 (B1, NatSec Incident View; NatSec Impact Framework): \"National Security impact assessment for each AI incident in the database, including 5 categories of impact and threat classification (novelty, imminence, autonomy) … It is intended to be applicable across different nations, grading the impact to the primary target country and identifying other nations that were also affected.\"; F39 (B4, National Security Impact Assessment): \"Assesses NatSec impact of each incident in 5 categories: Physical Security & Critical Infrastructure / Information Warfare & Intelligence Security / Sovereignty & Government Functions / Economic & Technological Security / Societal Stability & Human Rights using this framework. … Classifies threat for Imminence, Novelty and Autonomy\"", "receiving_text": "—", "translation": "untranslated: composed as received beside no archive claim", "archive_claims": [] }, { "lineage": "operator: the MIT AI Risk Initiative at MIT FutureTech", "field_claims": [ "F7", "F21", "F22", "F47", "F48" ], "received": "F7 (B1, What is the AI Incident Tracker? ¶3 s1): \"The AI Incident Tracker is part of the MIT AI Risk Initiative, which aims to increase awareness and adoption of best practice AI risk management across the AI ecosystem.\"; F21 (B2, card 2 snippet): \"Tools and Data: Expert Survey (expert consensus on severe AI risks, vulnerabilities, and responsibilities), AI Risk Repository (living repository of AI risks)\"; F22 (B2, hero): \"Authoritative data and frameworks to help you identify, prioritize, and manage the risks from AI\"; F47 (B6, Who we are ¶2–3): \"We are based at MIT FutureTech, a joint initiative of the Institute of the Digital Economy and MIT CSAIL at the Massachusetts Institute of Technology … Our work is free to use and openly available, and most of our outputs are published under a CC BY 4.0 licence.\"; F48 (B6, What we do ¶1 (also the card 6 snippet)): \"AI risk is a large and fast-moving subject. Understanding it means keeping track of what can go wrong, who is affected, how organizations are responding, and where the gaps are. No single dataset captures all of this, so we maintain several connected workstreams, each answering a different question.\"", "receiving_text": "—", "translation": "untranslated: composed as received beside MMRS-09, MMRS-17", "archive_claims": [ "MMRS-09", "MMRS-17" ] }, { "lineage": "origin: developed by Simon Mylius", "field_claims": [ "F18", "F49" ], "received": "F18 (B1, Team): \"Simon Mylius MIT FutureTech Peter Slattery MIT FutureTech … Sean McGregor AI Incident Database … Neil Thompson MIT FutureTech\"; F49 (B6, Team, Simon Mylius): \"He developed and leads the AI Incident Tracker - a classification tool and dashboard to add structure to datasets of reported AI safety incidents, which is now hosted by the MIT AI Risk Repository and its analysis features on AI Incident Database.\"", "receiving_text": "—", "translation": "untranslated: composed as received beside no archive claim", "archive_claims": [] }, { "lineage": "updates: reclassification by version", "field_claims": [ "F35", "F36", "F37" ], "received": "F35 (B3, Next steps): \"Our results suggest that LLM-based incident classification has comparable accuracy to the human baseline and is a viable approach for scaling the Incident Tracker. … The next step is to run a more comprehensive validation study based on this pilot. … The full set of reports from the AI Incident Database will be reclassified using the updated model selection and adapted prompts, to refresh the Incident Tracker with more reliable classifications.\"; F36 (B4, Major update - 23 June 2025): \"This brings us completely up to date with the latest incidents on the AI Incident Database as at 23 June 2025 (up to incident ID #1116)\"; F37 (B4, LLM Analysis): \"The full dataset has been classified using this latest iteration of the tool (reclassification of all incidents that were classified with the previous version, to ensure consistency across the dataset).\"", "receiving_text": "—", "translation": "untranslated: composed as received beside MMRS-13", "archive_claims": [ "MMRS-13" ] }, { "lineage": "count: four figures across MIT's surfaces", "field_claims": [ "F2", "F3", "F41", "F42", "F45" ], "received": "F2 (B1, meta description (name=description, og:description, twitter:description), extracted to B1-meta.txt): \"The MIT AI Incident Tracker classifies more than 1,400 real-world reported incidents from the AI Incident Database by risk, cause, harm, severity, and other relevant dimensions.\"; F3 (B1, card 1 snippet (the count is not in the saved page)): \"Overview: MIT AI Incident Tracker classifies 1,752 real-world AI incident reports from the AI Incident Database using Arcola AI's LLM pipeline\"; F41 (B5, card 5 snippet (the page's meta description reads 'more than 1,300')): \"Explore more than 1300 real-world reported incidents from the AI Incident Database by risk, cause, harm, severity, and other relevant dimensions.\"; F42 (B5, AI Incident Tracker heading): \"1,497 real-world AI incidents from the AI Incident Database, classified by risk domain, cause, harm type, and severity.\"; F45 (B5, headline figures): \"Up -76% … year-over-year … 64% … caused by AI systems … high-severity in 2026 … Malicious Actors & Misuse … most affected domain\"", "receiving_text": "—", "translation": "untranslated: composed as received beside CAP-1-04, MMRS-05, MMRS-15", "archive_claims": [ "CAP-1-04", "MMRS-05", "MMRS-15" ] }, { "lineage": "validation: the June 2026 pilot", "field_claims": [ "F25", "F26", "F27", "F28", "F29", "F34" ], "received": "F25 (B3, Executive Summary s2–3): \"Because the tracker relies on automated LLM classification rather than manual review, we need to understand whether LLMs can classify incidents as consistently as expert humans. This report describes a pilot study to compare the performance of eight LLMs against expert human reviewers\"; F26 (B3, What We Did ¶1; Methodology): \"We asked two expert reviewers to independently read reports of 10 incidents and classify each one using five taxonomies … We selected 10 incident reports from the AI Incident Database.\"; F27 (B3, What We Found ¶1): \"Opus 4.6 and Kimi K2.5 were the two strongest models in our evaluation and met or exceeded the currently used model, Sonnet 4, on most taxonomies.\"; F28 (B3, What We Found ¶3–4): \"On the EU AI Act Risk Level taxonomy they performed substantially worse. … After targeted prompt refinement, Opus 4.6 matched or exceeded the human-human agreement baseline on all five taxonomies on this pilot sample.\"; F29 (B3, Results, Evaluating the human baseline): \"By the Landis & Koch interpretation scale, the human-human agreement is itself only \"Fair\" for Harm Severity (0.31) and \"Moderate\" for EU AI Act Risk Level (0.56).\"; F34 (B3, Future Work ¶1 s2): \"The human consensus labels used as the benchmark for evaluating LLM performance were based on a small review sample: two reviewers per incident across 10 incidents.\"", "receiving_text": "—", "translation": "untranslated: composed as received beside no archive claim", "archive_claims": [] }, { "lineage": "limits: indicative, not validated", "field_claims": [ "F13", "F14", "F15", "F16", "F17", "F33", "F43" ], "received": "F13 (B1, Limitations of the analysis and dataset ¶1 s1): \"This classification database is intended to explore the potential capabilities and limitations of a scalable incident analysis framework.\"; F14 (B1, Limitations ¶1 s2–3): \"The classification analysis uses reports from the AI Incident Database (AIID) as input data which rely on submissions from the public and subject matter experts. The quality, reliability and depth of detail in the reports varies across the dataset.\"; F15 (B1, Limitations ¶1 s4): \"As the reporting is voluntary, the dataset is inevitably subject to some degree of sampling bias.\"; F16 (B1, Limitations ¶1 s5): \"Spot-checks have been used to provide feedback on misclassifications and to iterate the tool, improving its reliability, however a systematic validation study has not yet been completed.\"; F17 (B1, Limitations ¶2): \"Therefore patterns and trends observed in the data should be taken as indicative and validated through further analysis.\"; F33 (B3, Introduction ¶4–5): \"The Incident Tracker uses large language models (LLMs) to make incident classification scalable. … However, LLMs can make mistakes when interpreting incident reports and assigning classifications. For instance, they can occasionally hallucinate details and can exhibit a score-range bias, where ratings cluster within a narrow band of an ordinal scale rather than using the full range.\"; F43 (B5, banner): \"This is a research prototype. The data and analyses are preliminary and not yet validated\"", "receiving_text": "—", "translation": "untranslated: composed as received beside CAP-1-05, MMRS-02, MMRS-06, MMRS-16", "archive_claims": [ "CAP-1-05", "MMRS-02", "MMRS-06", "MMRS-16" ] }, { "lineage": "openness: CC BY 4.0", "field_claims": [ "F19" ], "received": "F19 (B1, footer): \"Website © MIT FutureTech 2026 … Data from the MIT AI Risk Initiative is licensed under CC BY 4.0\"", "receiving_text": "—", "translation": "untranslated: composed as received beside no archive claim", "archive_claims": [] }, { "lineage": "neighbours: the incident databases it builds on", "field_claims": [ "F31" ], "received": "F31 (B3, Introduction ¶2): \"Several databases already collect reports of AI-related incidents, including the AI Incident Database, the OECD’s AI Incident Monitor, the AIAAIC Repository, the Political Deepfakes Database, and others. The MIT AI Incident Tracker project builds on these efforts by classifying each incident against a set of risk taxonomies and rating the severity of the harm involved.\"", "receiving_text": "—", "translation": "untranslated: composed as received beside MMRS-11", "archive_claims": [ "MMRS-11" ] }, { "lineage": "reception: uptake of the tracker", "field_claims": [ "F32" ], "received": "F32 (B3, Introduction ¶3 s3): \"These outputs have been used to report on incident trends, including in the 2026 World Disasters Report and Time magazine.\"", "receiving_text": "—", "translation": "untranslated: composed as received beside no archive claim", "archive_claims": [] }, { "lineage": "the encyclopedia: no article; one mention under the repository's name", "field_claims": [ "F51", "F52", "F53", "F54" ], "received": "F51 (W1, rendered page https://en.wikipedia.org/wiki/MIT_AI_Incident_Tracker (HTTP 404)): \"Wikipedia does not have an article with this exact name.\"; F52 (W1, full-text search '\"AI Incident Tracker\"' (W1-search.txt)): \"There were no results matching the query.\"; F53 (W3, Voiceverse NFT plagiarism scandal, Reactions and legacy ¶ (found by full-text search '\"AI Risk Repository\"', W2-search.txt: one result)): \"The MIT AI Risk Repository catalogued the incident as an example of \"economic and cultural devaluation of human effort\"\"; F54 (W4, incidentdatabase.ai/cite/277 (the reference Wikipedia gives), Applied Taxonomies): \"GMF, MIT MIT Taxonomy Classifications Machine-Classified … 4.3. Fraud, scams, and targeted manipulation … Malicious Actors & Misuse\"", "receiving_text": "—", "translation": "untranslated: composed as received beside no archive claim", "archive_claims": [] } ], "receiving_texts": "equal-terms case: every field claim composed as received, beside the archive claims it stands with" }, "measure": { "rule": "v0.9 §4.10: KO_B sentences surviving in the KO at similarity ≥ 0.95, with positions (computed as apply_v09.py does: difflib.SequenceMatcher ratio, the best KO sentence for each KO_B sentence)", "v0.9": { "surviving": 0, "of": 11, "positions": [ null, null, null, null, null, null, null, null, null, null, null ] }, "opening_run": { "rule": "v0.9 §5.10 (equal-terms case): the field-only KO's sentences as the KO's opening run in their own order", "breach": false }, "blocks": { "rule": "v0.9 §4.1 (equal-terms and mixed cases): the KO does not open with the field as a block or close with the archive as one; sentences typed by the ledgers of the claims they cite (F field-only, A archive-only, M both)", "sequence": "MMMMFFFMFMMAFFFFMAFAAMFF", "field_only_head": 0, "archive_only_tail": 0, "breach": false }, "max_similarity": [ 0.234, 0.242, 0.037, 0.175, 0.281, 0.137, 0.424, 0.216, 0.242, 0.529, 0.211 ], "ko_sentence_1_max_similarity_to_KO_B": 0.234 }, "homonyms": { "names": [], "basis": "the MMRS record carries no disambiguation for the entity, and no card on the rail is another entity of the name: all six are MIT's own surfaces. The parent and neighbours are distinct entities, kept as relations, never senses of the name: the MIT AI Risk Repository (part_of; Wikipedia's one mention credits the tracker's classification to it, F53), the AI Incident Database (derived_from), the MIT AI Risk Navigator (a separate MIT project that presents the tracker's data, B5), and the databases the June 2026 update names beside it (the OECD's AI Incident Monitor, the AIAAIC Repository, the Political Deepfakes Database, F31).", "distinct_entities": [ "mit-ai-risk-repository", "ai-incident-database", "MIT AI Risk Navigator", "oecd-ai-incidents-monitor", "aiaaic-repository", "Political Deepfakes Database" ], "substitutes": { "T": [] }, "register": { "address_1": { "slug": "mit-ai-incident-tracker-aio-20261029", "obs_id": "OBS-65034e45b99a" } } }, "notes": [ "One address, Google AI Overview (expanded), signed out, incognito, 2026-10-29: mit-ai-incident-tracker-aio-20261029 (OBS-65034e45b99a, ADDR-f65abd8a4bbe, AXN:0AAE.CAPTURE.🌄🐚▲🎨🌈➕), six cards, all MIT's.", "Wikipedia: no article for the name or for the parent's names (four titles, HTTP 404); no full-text result for \"AI Incident Tracker\". The one mention of the work, under 'The MIT AI Risk Repository', is in 'Voiceverse NFT plagiarism scandal' (F53), referenced to the AIID's citation page for incident 277 (access-date 5 October 2025), whose MIT classification read on 2026-10-29 is machine-classified subdomain 4.3, fraud, scams and targeted manipulation (F54), not the 'economic and cultural devaluation of human effort' Wikipedia quotes: either the incident was reclassified (the tracker reclassifies its whole dataset at each version, F37, F35) or the encyclopedia misreports it (open block). The MediaWiki search API returned HTTP 429 once and was not retried; the Special:Search page was used.", "Fetching: every page saved verbatim with curl on 2026-10-29 between 05:07:24 and 05:11:17 UTC, raw and text with sha256 in field.json. Card URLs: B1 the MMRS record's source; B2 and B6 by title and site; B3 and B4 by WebSearch (B4's slug, with three hyphens, found by searching the snippet's 'incident ID #1116' after the guessed slug returned 404); B5 from airisk.mit.edu/llms.txt, which gives the Navigator as www.airi-navigator.com. No card unreachable.", "The three counts the census records, ledgered where they stand: 'more than 1,400' in the tracker page's meta description (F2, B1-meta.txt, the description, og:description and twitter:description tags alike); 1,600 in the page's heading (F1); 1,752 in card 1's snippet only, in no saved page (F3). Two more on the Navigator: 'more than 1300' in card 5's snippet ('more than 1,300' in the page's meta description, B5-meta.txt) and 1,497 in the Navigator page's heading on 2026-10-29 (F42), which the MMRS record and the capture do not carry. None of the figures is dated on its surface.", "Own pages: the MMRS record cites one source, the tracker page (B1, card 1); the task's June 2026 and June 2025 update pages are cards 3 and 4 (B3, B4); all carried once, as cards. No own page stands outside the rail.", "The archive's deposits do not name the tracker (D 0, R 0); the O-class pass found three sentences in #662 and #1397 on LLM classification in other senses, read and rejected (reading.json). The archive's bearing is carried by its own records: the MMRS record (MMRS-01–22) and the capture (CAP-1-01–05).", "Defects found, reported and not fixed. (1) MMRS census 'count': as_of 2026-10-28 with evidence 'airisk.mit.edu/ai-incident-tracker, basis observed', but the value carries Google's card summary (1,752, from the capture of 2026-10-29) and the Navigator's card ('more than 1300'), neither the cited source nor in 'sources'; the Navigator page's own heading (1,497 on 2026-10-29) is not carried. (2) MMRS census 'unit': 'AI Incident Database report' where MIT's pages count and reclassify incidents ('1,600 real-world report incidents'; 'reclassification of all incidents'; the June 2026 pilot sampled '10 incidents') and the AIID shows the MIT classification per incident (F54); carried as contested. (3) The MMRS record (article and note) says 'no systematic validation study is complete', the tracker page's wording; the June 2026 update, a card on the same rail, reports a pilot validation study and names a fuller one as the next step, which the record does not mention (not a contradiction: the update calls it a pilot). (4) The MMRS record's operator names 'Arcola AI's LLM pipeline'; the June 2026 update names the model in use, Sonnet 4, and no ledgered source says what Arcola AI is. (5) The MMRS census row carries no disambiguation, negative tags or battery, unlike the AIID's row. (6) The capture's finding 2 calls the count 'three ways' and gives four figures (two on the page, the card's, the Navigator card's): the heading counts surfaces, not figures (not material). The tracker page itself carries unfinished template blocks ('[TO UPDATE / REMOVE] Frequently Asked Questions', 'Question 1', 'Response 1'), recorded in B1's note, not ledgered.", "Field tensions, carried: the tracker page's 'a systematic validation study has not yet been completed' (F16), the June 2026 update's pilot (F25–F28) and the Navigator's 'not yet validated' (F43) (open block); the tracker page's 'part of the MIT AI Risk Initiative' (F7) and the about page's 'now hosted by the MIT AI Risk Repository' (F49), the MMRS record's part_of (MMRS-17); card 2's snippet is not in the saved home page (F21).", "The case and ρ (decided): equal, ρ = extends; the MMRS record describes the tracker from its own page, a basis per field, and adds senses and limits on the field's object (the derived layer, the chain of receptions, machine-written classifications, the inherited harm test); the field holds the ontology, so the suppressed witness does not apply. No deposit defines the archive's entity; the defining material is the MMRS record and the capture record.", "Opening-run check (§5.10): KO sentence 1 is new (max similarity to any KO_B sentence 0.234); no KO_B sentence survives at ≥ 0.95 (highest 0.529); sentence types MMMMFFFMFMMAFFFFMAFAAMFF (no field-only head, no archive-only tail of three or more).", "For the operator, outside the row: the closer at this address offers 'deepfakes, autonomous vehicles, or privacy leaks', categories on no card of this rail and, near verbatim, the harm kinds the Overview gave at 'ai incident database' (that row's T1-3); a WebSearch for the June 2026 update also surfaced PointGuard AI's 'AI Security Incident Tracker', a different tracker, not on the rail and not ledgered." ], "composition_v09": { "source": "this row (composed directly; no rebuild/negative-of-the-negative/v09/mit-ai-incident-tracker.json)", "composer": "Claude (Opus 5.5), subagent, 2026-10-29", "defining_material_read": [ { "what": "MMRS record, census row", "file": "/home/claude/fleet/machinemediation-org/data/meta-aggregator/census.json (rows[slug == 'mit-ai-incident-tracker'], schema_note)", "read": "whole" }, { "what": "MMRS record, article", "file": "/home/claude/fleet/machinemediation-org/scripts/meta_aggregator/articles.py (ARTICLES['mit-ai-incident-tracker'])", "read": "whole" }, { "what": "MMRS record, rendered", "file": "/home/claude/fleet/machinemediation-org/aggregators/mit-ai-incident-tracker/index.html", "read": "text, for agreement with the sources" }, { "what": "capture record", "file": "data/EA-WG-CAPTURES-01.json (mit-ai-incident-tracker-aio-20261029)", "read": "transcript, transcript_raw, cite_list, d, reading, findings, analysis" }, { "what": "the source row", "file": "datasets/negative-of-the-negative/v2/rows/ai-incident-database.json", "read": "as exemplar and for the AIID's account of the tracker (its F6–F9 on the same page)" }, { "what": "the specification", "file": "data/texts/AXN-0700-text.md", "read": "§0, §1 (1.2a), §3.3–3.8, §4, §6.2 and Appendix A" } ], "cuts": [], "notes": "equal case; KO 24 sentences, all 54 field and 27 archive claims cited; P body 257 words; KO_B 11 sentences, none surviving at ≥ 0.95" }, "field_by_origin": { "cards": [ "B1", "B2", "B3", "B4", "B5", "B6" ], "own_pages": [], "own_pages_note": "the own pages are cards: B1 (the tracker page, the MMRS record's one source), B3 (June 2026 update) and B4 (June 2025 update), carried once, as cards", "wikipedia": [ "W1", "W3", "W4" ], "unreachable": [], "title_only": [], "wikipedia_state": "no article: https://en.wikipedia.org/wiki/MIT_AI_Incident_Tracker and its raw wikitext HTTP 404 ('Wikipedia does not have an article with this exact name'); 'AI Incident Tracker', 'MIT AI Risk Repository' and 'MIT AI Risk Initiative' HTTP 404 likewise; a full-text search for \"AI Incident Tracker\" has no result; a full-text search for \"AI Risk Repository\" has one, 'Voiceverse NFT plagiarism scandal', which says 'The MIT AI Risk Repository catalogued the incident as an example of \"economic and cultural devaluation of human effort\"', referenced to incidentdatabase.ai/cite/277, where on 2026-10-29 the MIT classification is machine-classified as subdomain 4.3 (fraud, scams and targeted manipulation); the encyclopedia knows the tracker's work only under its parent's name", "sources": [ { "id": "B1", "origin": "card", "role": "Overview card 1; also the own page the MMRS record cites (its only source)", "card": "MIT AI Risk Initiative — MIT AI Incident Tracker", "url": "https://airisk.mit.edu/ai-incident-tracker", "kind": "Overview card (own page)", "surfaced_by_T": [ { "slug": "mit-ai-incident-tracker-aio-20261029", "card": 1 } ], "files": { "B1.html": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B1.html", "sha256": "720c3b26f04adc57970c7e6090668c43a909a5f9601b94db9fa6a26d0fb0a91c" }, "B1.txt": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B1.txt", "sha256": "8d0169460225d22859c9cc71be61e32a3fac3d064e0719f239eeff4f7da2c683" }, "B1-meta.txt": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B1-meta.txt", "sha256": "f6cbad6e571b1744c4c6523a3d1b85d3ade3b55a7f1387ff2b469ee5d52fb0ab" } }, "fetched": { "at": "2026-10-29T05:07:24Z", "http": 200, "url": "https://airisk.mit.edu/ai-incident-tracker", "raw": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B1.html", "raw_sha256": "720c3b26f04adc57970c7e6090668c43a909a5f9601b94db9fa6a26d0fb0a91c", "text": [ "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B1.txt", "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B1-meta.txt" ] }, "card_title": "MIT AI Incident Tracker", "card_snippet": "Overview: MIT AI Incident Tracker classifies 1,752 real-world AI incident reports from the AI Incident Database using Arcola AI's LLM pipeline, MIT Risk Reposit...", "card_note": null, "names_operator": "yes ('MIT AI Risk Initiative' in the saved text or card)", "names_archive": "no (none of the Crimson Hexagonal Archive, Alexanarch, Lee Sharks, machinemediation, EA-WG-CAPTURES-01 in the saved text, card title and snippet)", "date": "2026-10-29 (undated; as fetched; the MMRS record read it 2026-10-28)", "note": "URL the MMRS record's source and the task's; the card's snippet ('classifies 1,752 real-world AI incident reports … using Arcola AI's LLM pipeline') is not in the saved page, whose heading gives 1,600 and whose meta description (B1-meta.txt, extracted verbatim from the name=description, og:description and twitter:description tags) gives 'more than 1,400'; the page carries unfinished template blocks ('[TO UPDATE / REMOVE] Frequently Asked Questions … Question 1 … Response 1', '[TO UPDATE / REMOVE] Acknowledgments'), recorded, not ledgered" }, { "id": "B2", "origin": "card", "role": "Overview card 2", "card": "MIT AI Risk Initiative — MIT AI Risk Initiative", "url": "https://airisk.mit.edu/", "kind": "Overview card (the operator's home page)", "surfaced_by_T": [ { "slug": "mit-ai-incident-tracker-aio-20261029", "card": 2 } ], "files": { "B2.html": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B2.html", "sha256": "d3f35366280a66cad3739c9cecb5216a73be687644466f0fabfd204af408be21" }, "B2.txt": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B2.txt", "sha256": "935d8b0170045fd4112c26b32af21b397d0d9e2dfa05f6e14c1a7b3156cae3ac" } }, "fetched": { "at": "2026-10-29T05:07:24Z", "http": 200, "url": "https://airisk.mit.edu/", "raw": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B2.html", "raw_sha256": "d3f35366280a66cad3739c9cecb5216a73be687644466f0fabfd204af408be21", "text": [ "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B2.txt" ] }, "card_title": "MIT AI Risk Initiative", "card_snippet": "Tools and Data: Expert Survey (expert consensus on severe AI risks, vulnerabilities, and responsibilities), AI Risk Repository (living repository of AI risks), ...", "card_note": null, "names_operator": "yes ('MIT AI Risk Initiative' in the saved text or card)", "names_archive": "no (none of the Crimson Hexagonal Archive, Alexanarch, Lee Sharks, machinemediation, EA-WG-CAPTURES-01 in the saved text, card title and snippet)", "date": "2026-10-29 (undated; as fetched)", "note": "URL by title and site (the page's is 'MIT AI Risk Initiative'); the card's snippet ('Tools and Data: Expert Survey (…), AI Risk Repository (living repository of AI risks), …') is a summary not found verbatim in the saved page, whose tools block reads 'Which risks are most important? … What are the risks from AI? … What real-world harms are occurring?'" }, { "id": "B3", "origin": "card", "role": "Overview card 3; an own page the task names (not among the MMRS record's sources)", "card": "MIT AI Risk Initiative — AI Incident Tracker June 2026 Update - MIT AI Risk Initiative", "url": "https://airisk.mit.edu/blog/ai-incident-tracker-june-2026-update", "kind": "Overview card (own page)", "surfaced_by_T": [ { "slug": "mit-ai-incident-tracker-aio-20261029", "card": 3 } ], "files": { "B3.html": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B3.html", "sha256": "d8b013b1dbc4f948e86e6bbea4936765db871e58df93007f478793ef05cdefba" }, "B3.txt": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B3.txt", "sha256": "c994f9c9ff3a14b5446b59a03c2aa587f48c5888ac61ed538c254e21b3a9e0fe" } }, "fetched": { "at": "2026-10-29T05:07:25Z", "http": 200, "url": "https://airisk.mit.edu/blog/ai-incident-tracker-june-2026-update", "raw": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B3.html", "raw_sha256": "d8b013b1dbc4f948e86e6bbea4936765db871e58df93007f478793ef05cdefba", "text": [ "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B3.txt" ] }, "card_title": "AI Incident Tracker June 2026 Update - MIT AI Risk Initiative", "card_snippet": "Executive Summary. The AI Incident Tracker uses large language models (LLMs) to process incident reports from the AI Incident Database (AIID) and classify them ...", "card_note": null, "names_operator": "yes ('MIT AI Risk Initiative' in the saved text or card)", "names_archive": "no (none of the Crimson Hexagonal Archive, Alexanarch, Lee Sharks, machinemediation, EA-WG-CAPTURES-01 in the saved text, card title and snippet)", "date": "2026-06-30 (posted)", "note": "URL found by WebSearch on the title; dated June 30, 2026; the card's snippet is the Executive Summary's first sentence; a PDF download is linked, not fetched" }, { "id": "B4", "origin": "card", "role": "Overview card 4; an own page the task names (not among the MMRS record's sources)", "card": "MIT AI Risk Initiative — Incident Tracker - June 2025 Update - MIT AI Risk Repository", "url": "https://airisk.mit.edu/blog/incident-tracker---june-2025-update", "kind": "Overview card (own page)", "surfaced_by_T": [ { "slug": "mit-ai-incident-tracker-aio-20261029", "card": 4 } ], "files": { "B4.html": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B4.html", "sha256": "952b48f0be997dc30853047c6f386d621ea1fca27dc8920f6d14055db97fde9d" }, "B4.txt": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B4.txt", "sha256": "90c383e98ee57e5b7ac6a5fb4244a094f34fe3704f2da287ef2cf778c839a495" } }, "fetched": { "at": "2026-10-29T05:07:25Z", "http": 200, "url": "https://airisk.mit.edu/blog/incident-tracker---june-2025-update", "raw": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B4.html", "raw_sha256": "952b48f0be997dc30853047c6f386d621ea1fca27dc8920f6d14055db97fde9d", "text": [ "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B4.txt" ] }, "card_title": "Incident Tracker - June 2025 Update - MIT AI Risk Repository", "card_snippet": "This brings us completely up to date with the latest incidents on the AI Incident Database as at 23 June 2025 (up to incident ID #1116) LLM Analysis Some update...", "card_note": null, "names_operator": "yes ('MIT AI Risk Initiative' in the saved text or card)", "names_archive": "no (none of the Crimson Hexagonal Archive, Alexanarch, Lee Sharks, machinemediation, EA-WG-CAPTURES-01 in the saved text, card title and snippet)", "date": "2025-07-18 (posted; the update of 2025-06-23)", "note": "URL found by WebSearch on the snippet's 'incident ID #1116' (the guessed slug /blog/incident-tracker-june-2025-update returned 404); dated July 18, 2025, for the update of 23 June 2025; the card's title ends 'MIT AI Risk Repository', the page's <title> 'Incident Tracker - June 2025 Update'" }, { "id": "B5", "origin": "card", "role": "Overview card 5", "card": "MIT AI Risk Navigator — Incidents - MIT AI Risk Navigator", "url": "https://www.airi-navigator.com/incidents", "kind": "Overview card (the Navigator, a project of the MIT FutureTech AI Risk Initiative)", "surfaced_by_T": [ { "slug": "mit-ai-incident-tracker-aio-20261029", "card": 5 } ], "files": { "B5.html": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B5.html", "sha256": "8c6757061ebc4be13c52000e335d5efc4210282b8e51d4d99a1e07de91a61528" }, "B5.txt": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B5.txt", "sha256": "b78ae81d171a1695208f7eb462831f2e68508c24bd8a4d958d9e4bca2b3ed462" }, "B5-meta.txt": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B5-meta.txt", "sha256": "1dd25599daef30317a180517487fc0b1d0ac91ff793e4f88d5926e5909331eeb" } }, "fetched": { "at": "2026-10-29T05:07:26Z", "http": 200, "url": "https://www.airi-navigator.com/incidents", "raw": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B5.html", "raw_sha256": "8c6757061ebc4be13c52000e335d5efc4210282b8e51d4d99a1e07de91a61528", "text": [ "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B5.txt", "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B5-meta.txt" ] }, "card_title": "Incidents - MIT AI Risk Navigator", "card_snippet": "Explore more than 1300 real-world reported incidents from the AI Incident Database by risk, cause, harm, severity, and other relevant dimensions.", "card_note": null, "names_operator": "no ('MIT AI Risk Initiative' absent from the saved text, card title and snippet)", "names_archive": "no (none of the Crimson Hexagonal Archive, Alexanarch, Lee Sharks, machinemediation, EA-WG-CAPTURES-01 in the saved text, card title and snippet)", "date": "2026-10-29 (v1.3.2; as fetched)", "note": "URL from airisk.mit.edu/llms.txt (the Navigator at www.airi-navigator.com) and the card's title; the page's <title> 'Incidents | MIT AI Risk Navigator'; the card's snippet is the meta description (B5-meta.txt) save 'more than 1300' for '1,300'; the page's heading gives 1,497; version v1.3.2; the footer credits Spencer Michaels" }, { "id": "B6", "origin": "card", "role": "Overview card 6", "card": "MIT AI Risk Initiative — About - MIT AI Risk Initiative", "url": "https://airisk.mit.edu/about", "kind": "Overview card (the operator's about page)", "surfaced_by_T": [ { "slug": "mit-ai-incident-tracker-aio-20261029", "card": 6 } ], "files": { "B6.html": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B6.html", "sha256": "b40b72f7dbd5c16c3d50da84dc12cbfc9974e71e5ffd84e3dc8f0a95305e3396" }, "B6.txt": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B6.txt", "sha256": "9be294440171b8044981eb4e2f6bb29d6610c0cec83ae5cd86df63ac59467a78" } }, "fetched": { "at": "2026-10-29T05:07:26Z", "http": 200, "url": "https://airisk.mit.edu/about", "raw": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B6.html", "raw_sha256": "b40b72f7dbd5c16c3d50da84dc12cbfc9974e71e5ffd84e3dc8f0a95305e3396", "text": [ "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/B6.txt" ] }, "card_title": "About - MIT AI Risk Initiative", "card_snippet": "AI risk is a large and fast-moving subject. Understanding it means keeping track of what can go wrong, who is affected, how organizations are responding, and wh...", "card_note": null, "names_operator": "yes ('MIT AI Risk Initiative' in the saved text or card)", "names_archive": "no (none of the Crimson Hexagonal Archive, Alexanarch, Lee Sharks, machinemediation, EA-WG-CAPTURES-01 in the saved text, card title and snippet)", "date": "2026-10-29 (undated; as fetched)", "note": "URL by title and site; the card's snippet is 'What we do' ¶1" }, { "id": "W1", "origin": "wikipedia", "role": "Wikipedia's page for the name: none", "card": "Wikipedia — MIT AI Incident Tracker (no article)", "url": "https://en.wikipedia.org/wiki/MIT_AI_Incident_Tracker", "kind": "Wikipedia (HTTP 404)", "surfaced_by_T": "no (not surfaced by T; admitted beside the cards by the operator's ruling of 2026-10-29)", "files": { "W1.html": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/W1.html", "sha256": "079e59d86eb3bc8c9302519d36bc05fc2785b5bd2a610580a5f463b68248509e" }, "W1.txt": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/W1.txt", "sha256": "86ceb4ca3feeeb06b900ac02d99298dcaeccf22add0e7a3d5423efd6a7fd5a82" }, "W1-raw.html": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/W1-raw.html", "sha256": "5531fa583eab1ce6274baa3ff8c97ae2893067ae935275f88d671a218f060c3e" }, "W1-search.html": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/W1-search.html", "sha256": "1a569d35b508c453a5f968970ab96358f7f8bc8a85844ae9b2777d0d8453abdb" }, "W1-search.txt": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/W1-search.txt", "sha256": "b752a925085856a6c1a893895f1fbfcdeb2568d42d8bb128578aa5cf4fd9cc31" } }, "fetched": { "at": "2026-10-29T05:07:26Z", "http": 404, "url": "https://en.wikipedia.org/wiki/MIT_AI_Incident_Tracker", "raw": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/W1.html", "raw_sha256": "079e59d86eb3bc8c9302519d36bc05fc2785b5bd2a610580a5f463b68248509e", "text": [ "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/W1.txt", "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/W1-search.txt" ] }, "unreachable": false, "no_article": true, "names_operator": "no ('MIT AI Risk Initiative' absent from the saved text, card title and snippet)", "names_archive": "no (none of the Crimson Hexagonal Archive, Alexanarch, Lee Sharks, machinemediation, EA-WG-CAPTURES-01 in the saved text, card title and snippet)", "date": "2026-10-29 (HTTP 404, as fetched)", "note": "no article: the rendered page and the raw wikitext (https://en.wikipedia.org/w/index.php?title=MIT_AI_Incident_Tracker&action=raw, W1-raw.html) both HTTP 404; the same for AI_Incident_Tracker, MIT_AI_Risk_Repository and MIT_AI_Risk_Initiative (W1-alt-*.html, HTTP 404 each, 2026-10-29T05:07:27Z–05:07:30Z); full-text search for '\"AI Incident Tracker\"' (W1-search.html, HTTP 200, 05:10:17Z): no results. The MediaWiki search API answered HTTP 429 (rate limit) at 05:07:38Z and was not retried; the Special:Search page was used instead" }, { "id": "W3", "origin": "wikipedia", "role": "the encyclopedia's one mention of the tracker's work, found by full-text search for '\"AI Risk Repository\"' (W2-search.html, HTTP 200, 2026-10-29T05:10:24Z: 'Result 1 of 1'), under the repository's name", "card": "Wikipedia — Voiceverse NFT plagiarism scandal", "url": "https://en.wikipedia.org/wiki/Voiceverse_NFT_plagiarism_scandal", "kind": "Wikipedia article (a mention)", "surfaced_by_T": "no (not surfaced by T; admitted beside the cards by the operator's ruling of 2026-10-29)", "files": { "W3.html": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/W3.html", "sha256": "b5916526f7bbf7c1fb86f816c6ef842741f79548574eb28890d0d1c7105d7b31" }, "W3.txt": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/W3.txt", "sha256": "036aa5c461bbe8958ca9d2b3de0c8eef411ee7545b3ea1e0d920f8a1efc14a74" }, "W3.wiki": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/W3.wiki", "sha256": "ea09a4740b6caa9b1e86fbffc93cd30970820b0288bde3db4a3a0a6955cc09c2" }, "W2-search.html": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/W2-search.html", "sha256": "be689c38f669d18fca7fd8f08367c9aa54358ad2c75dd36cee5d6273a1c8040b" }, "W2-search.txt": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/W2-search.txt", "sha256": "f0f2921e1c6e2a2ab3f5d8afa63333799a89cad524ccd4317c23f5b8c4e1a4d5" } }, "fetched": { "at": "2026-10-29T05:11:17Z", "http": 200, "url": "https://en.wikipedia.org/wiki/Voiceverse_NFT_plagiarism_scandal", "raw": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/W3.html", "raw_sha256": "b5916526f7bbf7c1fb86f816c6ef842741f79548574eb28890d0d1c7105d7b31", "text": [ "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/W3.txt", "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/W2-search.txt" ] }, "names_operator": "no ('MIT AI Risk Initiative' absent from the saved text, card title and snippet)", "names_archive": "no (none of the Crimson Hexagonal Archive, Alexanarch, Lee Sharks, machinemediation, EA-WG-CAPTURES-01 in the saved text, card title and snippet)", "date": "2026-10-10 (last edit, by the search listing)", "note": "raw wikitext (https://en.wikipedia.org/w/index.php?title=Voiceverse_NFT_plagiarism_scandal&action=raw, HTTP 200, 05:10:44Z, W3.wiki) and rendered page; last edited 10 October 2026 by the search listing; the sentence's reference is the AIID's citation page for incident 277 (ref name Sonali-2022, access-date October 5, 2025)" }, { "id": "W4", "origin": "wikipedia", "role": "the reference Wikipedia gives for its mention (incidentdatabase.ai/cite/277), fetched to read the classification it cites", "card": "AI Incident Database — Incident 277 (Wikipedia's reference)", "url": "https://incidentdatabase.ai/cite/277/", "kind": "Wikipedia's reference (an AIID page)", "surfaced_by_T": "no (not surfaced by T; admitted beside the cards by the operator's ruling of 2026-10-29)", "files": { "W4.html": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/W4.html", "sha256": "e4d67493c6e1de915dfb97bfb34dbde0ef13454cfd325a8cbd76eda35ff93714" }, "W4.txt": { "path": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/W4.txt", "sha256": "51fdb16c0c23c6daa2717338589d48961c2ff915778ef3f3a7261369648da4c5" } }, "fetched": { "at": "2026-10-29T05:10:56Z", "http": 200, "url": "https://incidentdatabase.ai/cite/277/", "raw": "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/W4.html", "raw_sha256": "e4d67493c6e1de915dfb97bfb34dbde0ef13454cfd325a8cbd76eda35ff93714", "text": [ "datasets/negative-of-the-negative/v2/intake/mit-ai-incident-tracker-20261029/W4.txt" ] }, "names_operator": "no ('MIT AI Risk Initiative' absent from the saved text, card title and snippet)", "names_archive": "no (none of the Crimson Hexagonal Archive, Alexanarch, Lee Sharks, machinemediation, EA-WG-CAPTURES-01 in the saved text, card title and snippet)", "date": "2026-10-29 (as fetched)", "note": "the page shows 'MIT Taxonomy Classifications … Machine-Classified', subdomain '4.3. Fraud, scams, and targeted manipulation', domain 'Malicious Actors & Misuse'; the phrase Wikipedia quotes ('economic and cultural devaluation of human effort') is not on the page as fetched" } ] } }