{ "row": "data-provenance-initiative", "entity": "data-provenance-initiative", "address": "Data Provenance Initiative", "addresses": [ { "address": "Data Provenance Initiative", "slug": "data-provenance-initiative-aio-20261029", "addr_id": "ADDR-c0bda0b5919d", "obs_id": "OBS-118fbafa73c9", "axn": "AXN:0ACD.CAPTURE.⚙️⛵∞✋♻️🌉" } ], "epoch": "2026-10-29", "type": "C (public entity on which the archive bears: a research initiative with its audit collection, explorer and card; an entity of the provenance 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 provenance census record, the capture record and #725's statement of relation; archive selected by D/R/O (10 deposits admitted by reading, 6 with claims and 4 grouped as instances; hop of 73 read and rejected), with the provenance census record (22 claims) and the Capture Registry's findings and analysis (13 claims) ledgered as the archive's records of the entity; the field B by origin (4 cards, 2 own pages, 0 census third-party, 1 Wikipedia title); 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 47/47 verified by script against saved text (the bundle's extracted page text for B1 and B7, the census's copy for B5, the Wikipedia responses for B6); archive quotes 63/63 verified against the registry entry, census.json and the deposit texts; CAP-1-01 carried with the mark contradicts_primary", "frozen": false, "breaches": [] }, "spec": "EA-NEGONT-02 v0.9, #1690", "objects": { "T": { "path": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/T1-data-provenance-initiative-aio-20261029.txt", "sha256": "d70390cc1cbc82e6b1f79ac85a3410997dea97adaa7d971b2e2ce754d1cb0fd2", "words": 190, "cards": 4, "words_note": "one transcript at one address (190 words in the answer body with its section headings and closer, links reduced to their text, markers and the card rail excluded; 4 cards)", "claims": [ "T1-1 definition: 'a multi-institutional research project—led in part by researchers from MIT—that audits the datasets, licenses, and web domains used to train artificial intelligence models' (two markers, one of them the Initiative's inline link)", "T1-2 Data Provenance Explorer: 'An interactive tool that audits and filters over 1,800+ text and fine-tuning datasets across dozens of collections to inspect sources, lineages, and license types.'", "T1-3 Web Consent Explorer: 'A system that tracks thousands of web domains for consent signals and web policies.'", "T1-4 Open Model Evolution: 'A monitor tracking open-source and open-weight AI models for transparency and concentration.'", "T1-5 AI Observatory: 'A comparative tool for analyzing real-world AI conversation sets.' (three markers for the tools block)", "T1-6 licence omission and errors: 'Widely used hosting platforms exhibit over a 70% license omission rate and more than 50% error rates in how they categorize dataset licenses.' (one marker)", "T1-7 composition divides: 'Commercially closed or restrictive datasets often monopolize creative tasks, specialized topics, and lower-resource languages.' (one marker)", "T1-8 non-commercial spikes: 'There has been a significant rise in non-commercial and academic-only licenses, frequently driven by synthetic data generated via commercial APIs.' (one marker, a token that occurs nowhere else in the answer)", "T1-9 closer: 'Would you like to explore how to use the Data Provenance Explorer, or look into specific dataset licenses?'" ], "register": { "register": "data/EA-WG-CAPTURES-01.json", "slug": "data-provenance-initiative-aio-20261029", "obs_id": "OBS-118fbafa73c9" }, "transcripts": [ { "slug": "data-provenance-initiative-aio-20261029", "address": "Data Provenance Initiative", "addr_id": "ADDR-c0bda0b5919d", "obs_id": "OBS-118fbafa73c9", "axn": "AXN:0ACD.CAPTURE.⚙️⛵∞✋♻️🌉", "date": "2026-10-29", "surface": "Google AI Overview (expanded)", "auth": "signed out, incognito", "sha256": "d70390cc1cbc82e6b1f79ac85a3410997dea97adaa7d971b2e2ce754d1cb0fd2", "sha256_of": "the registry's 'transcript' string (utf-8)", "sha256_raw": "b4bb5c6e0001583f0546ede91a0b74e554b1fec81fa01951b9eec87932e12b04", "words": 190, "cards": 4, "card_rail": [ "1. Data Provenance Initiative — Data Provenance Initiative", "2. arXiv — [2310.16787] The Data Provenance Initiative: A Large Scale Audit of Dataset Licensing & Attribution in AI", "3. MIT Sloan — Bringing transparency to the data used to train artificial intelligence | MIT Sloan", "4. Nature — A large-scale audit of dataset licensing and attribution in AI | Nature Machine Intelligence" ], "claims": [ "T1-1 definition: 'a multi-institutional research project—led in part by researchers from MIT—that audits the datasets, licenses, and web domains used to train artificial intelligence models' (two markers, one of them the Initiative's inline link)", "T1-2 Data Provenance Explorer: 'An interactive tool that audits and filters over 1,800+ text and fine-tuning datasets across dozens of collections to inspect sources, lineages, and license types.'", "T1-3 Web Consent Explorer: 'A system that tracks thousands of web domains for consent signals and web policies.'", "T1-4 Open Model Evolution: 'A monitor tracking open-source and open-weight AI models for transparency and concentration.'", "T1-5 AI Observatory: 'A comparative tool for analyzing real-world AI conversation sets.' (three markers for the tools block)", "T1-6 licence omission and errors: 'Widely used hosting platforms exhibit over a 70% license omission rate and more than 50% error rates in how they categorize dataset licenses.' (one marker)", "T1-7 composition divides: 'Commercially closed or restrictive datasets often monopolize creative tasks, specialized topics, and lower-resource languages.' (one marker)", "T1-8 non-commercial spikes: 'There has been a significant rise in non-commercial and academic-only licenses, frequently driven by synthetic data generated via commercial APIs.' (one marker, a token that occurs nowhere else in the answer)", "T1-9 closer: 'Would you like to explore how to use the Data Provenance Explorer, or look into specific dataset licenses?'" ], "markers": "Inline links are google.com/goto tokens, opaque; the paste is the Overview expanded (operator's default of 2026-10-01; the 'AI Mode Conversation' view is the expanded Overview, ruling of 2026-10-18); signed out, incognito (operator's default of 2026-10-14). One turn. Four distinct tokens occur in the answer (the definition two, one of them the Initiative's inline link; the tools block three; the licence and divide bullets one each, the same token; the non-commercial bullet one, a fourth token); the tokens cannot be resolved to cards (the registry's cite_list carries url null).", "path": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/T1-data-provenance-initiative-aio-20261029.txt" } ] }, "KO": { "title": "The Data Provenance Initiative", "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 (briefs rebuild/negative-of-the-negative/v09/BRIEF-NEW-ROW-AGGREGATOR.md and BRIEF-NEW-ROW-PROVENANCE.md); not frozen; ledgers unaudited", "sentences": [ { "n": 1, "para": 0, "sense": "what it is", "text": "The Data Provenance Initiative is, in its repository's words, \"a multi-disciplinary volunteer effort to improve transparency, documentation, and responsible use of training datasets for AI,\" and on its About page \"a volunteer collective of AI researchers conducting large-scale audits of the systems and resources that shape modern AI\" (README; dataprovenance.org/about); the provenance census record gives its operator as the Initiative itself, a \"volunteer effort\" with a \"paper by Shayne Longpre and 16 other authors,\" kinds it an \"audit collection and explorer\" and sets it in the lineage of \"training data\" (provenance census record).", "claims": [ "F40", "F45", "PE-02", "PE-04", "PE-03" ] }, { "n": 2, "para": 0, "sense": "what it is", "text": "MIT Sloan's account has \"a team of multidisciplinary researchers, including professor Sandy Pentland and others from MIT,\" create the Initiative \"to tackle the data transparency challenge head-on\" (MIT Sloan, 3 March 2025), and the home page says that \"DPI releases papers, datasets, and dashboards that help researchers, journalists, policymakers, and builders reason from shared evidence\" (dataprovenance.org).", "claims": [ "F23", "F9" ] }, { "n": 3, "para": 0, "sense": "what it is", "text": "Its work, the About page says, \"began with training data provenance: tracing datasets from origin to creation, cataloging sources, licenses, creators, and documentation practices\" and \"now also covers web consent, open model markets, concentration of power, and how people use AI systems in practice\" (dataprovenance.org/about); the home page calls it \"Public measurement infrastructure for AI data, model ecosystems, and real-world AI use,\" which \"maps the datasets, web policies, model markets, and real-world AI uses that shape how AI systems are built and governed\" through \"Four public measurement systems\" (dataprovenance.org).", "claims": [ "F46", "F1", "F2", "F3" ] }, { "n": 4, "para": 1, "sense": "data provenance", "text": "The founding audit's authors name its object: \"we term the combination of these indicators, including a dataset's sourcing, creation and licensing heritage, as well as its characteristics, the 'data provenance'\" (Nature Machine Intelligence, 30 August 2024), traced \"from their source, creators, series of license conditions, properties, and subsequent use\" (arXiv:2310.16787), and the home page's \"Training Data Provenance\" traces \"the sources, licenses, creators, languages, geographies, and derivation chains behind public AI training datasets\" (dataprovenance.org).", "claims": [ "F30", "F18", "F15" ] }, { "n": 5, "para": 1, "sense": "data provenance", "text": "The provenance census asks of each entity it formalizes \"What provenance does this system carry, through which operation, on whose assertion, with what evidence of survival?\" and answers for the Initiative that it records, of the three capabilities express, record and test: its object \"a fine-tuning dataset, traced through its collection to its sources,\" its unit \"a dataset within a collection,\" its relations derivation, creators, licences and subsequent use, the last field restating the paper's lineage (provenance census record).", "claims": [ "PE-22", "PE-05", "PE-06", "PE-09", "PE-07" ] }, { "n": 6, "para": 1, "sense": "data provenance", "text": "#725 sets that object in the second dimension of provenance, \"training-corpus licensing,\" which \"has academic instrumentation and emerging legal architecture,\" its object \"The corpus\" and its question \"Under what permissions did this data enter this system?\", with the Initiative among its frameworks and among provenance's disciplinary homes as \"dataset documentation (DPI, model cards, dataset cards),\" and states that the dimensions of artifact, licensing and meaning are independent: \"a licensed dataset can produce outputs with excellent semantic provenance if the synthesis preserves attribution\" (#725).", "claims": [ "O725-03", "O725-05", "O725-15", "O725-06" ] }, { "n": 7, "para": 2, "sense": "the founding audit", "text": "The founding paper was submitted on 25 October 2023 and revised on 4 November 2023, \"by Shayne Longpre and 16 other authors\" (arXiv:2310.16787), and published in Nature Machine Intelligence, volume 6, on 30 August 2024 (Nature Machine Intelligence); the census dates the entity from the preprint, \"2023 (arXiv:2310.16787, v1 25 October 2023)\" (provenance census record).", "claims": [ "F16", "F29", "PE-15" ] }, { "n": 8, "para": 2, "sense": "the founding audit", "text": "It convened \"a multi-disciplinary effort between legal and machine learning experts to systematically audit and trace 1800+ text datasets\" (arXiv), in the journal \"44 instruction or 'alignment' finetuning data collections composed of 1,858 individual datasets\" (Nature Machine Intelligence), the release the repository calls the Data Provenance Collection and \"this initiative's first release\" (README), and the count the census carries, \"44 data collections, 1800+ fine-tuning text-to-text datasets (first release),\" as of the README (provenance census record); MIT Sloan gives \"more than 1,800 text datasets\" from \"44 of the most widely used text data collections\" (MIT Sloan).", "claims": [ "F17", "F31", "F41", "PE-10", "F24" ] }, { "n": 9, "para": 3, "sense": "the Explorer, the Collection and the Card", "text": "With the audit came \"an interactive UI, the Data Provenance Explorer, which allows practitioners to trace and filter on data provenance for the most popular open source finetuning data collections\" (arXiv), in the journal \"an open-source repository for downloading, filtering and exploring data provenance and characteristics\" whose tools \"auto-generate data provenance cards for scalable symbolic attribution\" (Nature Machine Intelligence); the repository's Data Provenance Card is \"symbolic attribution for all constituent datasets, that can be used as structured documentation\" (README), \"human-readable\" in MIT Sloan's account (MIT Sloan), and the census carries it as the entity's provenance carrier beside \"audit metadata per dataset\" (provenance census record).", "claims": [ "F22", "F34", "F42", "F25", "PE-08" ] }, { "n": 10, "para": 3, "sense": "the Explorer, the Collection and the Card", "text": "The census gives Data Provenance Explorer, Data Provenance Collection and Data Provenance Card as other names of the entity and states its relation to them twice: among its negative tags, \"initiative, explorer and collection are distinct records,\" and on its sibling relation to the MMRS meta-aggregator's record of the explorer, \"the same entity, read for reception there and for provenance carriage here\"; it records the Explorer as \"a JavaScript application, unreadable to a plain fetch\" (provenance census record).", "claims": [ "PE-01", "PE-18", "PE-19", "PE-13" ] }, { "n": 11, "para": 4, "sense": "its systems now", "text": "The home page's four systems are the Data Provenance Explorer (\"4,000+ datasets,\" \"608 languages across text, speech, and video,\" inspecting \"dataset sources, licenses, creators, geographies, modalities, and derivation chains\"), the Web Consent Explorer (\"14,000 domains,\" \"Terms, robots.txt, and crawl restrictions\"), Open Model Evolution (\"851k open models,\" \"2.2B downloads\") and The AI Observatory (\"23,158 conversations,\" \"85,633 turns from 7 public sources\") (dataprovenance.org).", "claims": [ "F4", "F5", "F6", "F7" ] }, { "n": 12, "para": 4, "sense": "its systems now", "text": "Its metrics read \"4,000+\" \"datasets audited,\" \"14,000\" \"web domains tracked,\" \"851k\" \"open models measured\" and \"23,158\" \"AI conversations annotated,\" and its publications carry the later work: Consent in Crisis (NeurIPS 2024), the multimodal audit, \"An audit of four thousand public text, speech, and video datasets\" (ICLR 2025), Economies of Open Intelligence (preprint) and The AI Observatory (NeurIPS 2026) (dataprovenance.org); in March 2025 MIT Sloan still had the work \"currently focused on the data provenance of text,\" with a plan \"to expand to other media, such as video\" (MIT Sloan).", "claims": [ "F8", "F11", "F12", "F13", "F28" ] }, { "n": 13, "para": 4, "sense": "its systems now", "text": "The census carries the entity as \"active\" and lists \"the Explorer itself (JavaScript)\" and \"the multimodal extension (arXiv:2412.17847)\" among what it has not yet checked (provenance census record); the Capture Registry finds that \"The record has one explorer and the first release's count,\" that the home-page card gives \"four platforms and '4,000+ audited datasets; 14,000 tracked web domains',\" that the composition is \"ahead of the record on the entity's current extent,\" and that \"The census is to re-read dataprovenance.org and carry the platforms and the current count with their quotes,\" the capture being \"a correction to be made from the entity's own page\" (Capture Registry, 2026-10-29).", "claims": [ "PE-14", "PE-20", "CAP-1-05", "CAP-1-06", "CAP-1-10", "CAP-1-07", "CAP-1-11" ] }, { "n": 14, "para": 5, "sense": "what the audits found", "text": "The founding audit observed \"frequent miscategorization of licenses on widely used dataset hosting sites, with license omission of 70%+ and error rates of 50%+,\" which \"points to a crisis in misattribution and informed use of the most popular datasets driving many recent breakthroughs\" (arXiv), the result the census records as what survived the operation it tested, \"dataset repackaging and hosting (licence carried or lost on hosting sites)\" (provenance census record).", "claims": [ "F20", "F21", "PE-12", "PE-11", "PE-21" ] }, { "n": 15, "para": 5, "sense": "what the audits found", "text": "The journal gives the measure by platform, \"more than 70% of licences for popular datasets on GitHub and Hugging Face are 'unspecified'\" and \"66% of analysed Hugging Face licences were in a different use category, often labelled as more permissive than the author's original licence,\" and names \"a crisis in licence laundering\" (Nature Machine Intelligence); MIT Sloan restates error rates \"greater than 50%\" and omission rates \"of more than 70%\" (MIT Sloan).", "claims": [ "F32", "F33", "F24" ] }, { "n": 16, "para": 5, "sense": "what the audits found", "text": "The Capture Registry checked these figures \"against the provenance census record of the same day and the entity pages it saved\" and calls the 70%/50% wording \"the record's quote from the repository README\" (Capture Registry, 2026-10-29), where the census gives it as what \"the founding paper reports\" (provenance census record).", "claims": [ "CAP-1-02", "CAP-1-01", "PE-12" ] }, { "n": 17, "para": 5, "sense": "what the audits found", "text": "\"In total, 85% of dataset licences request attribution, and 30% include a share-alike clause\" (Nature Machine Intelligence); #725 carries these figures with the omission and error rates and reads them as the licensing dimension's ground, \"The DPI's documentation of 70%+ license-omission rates establishes the licensing dimension's urgency\" (#725), and #1714 reads the finding as \"how much is lost before training even begins\" (#1714).", "claims": [ "F35", "D725-02", "D725-16", "D1714-01" ] }, { "n": 18, "para": 5, "sense": "what the audits found", "text": "Of the commercial divide the preprint finds \"closed datasets monopolizing important categories: lower resource languages, more creative tasks, richer topic variety, newer and more synthetic training data\" (arXiv); the journal records that \"In 2023 there was a spike in NC/A-O dataset licences,\" \"the NC/A-O rate is 61%\" that year, and that \"a full 45% of NC/A-O datasets are synthetic, compared to <14% in more permissive licence categories\" (Nature Machine Intelligence); the home page's findings are \"Source content often carries stricter terms than dataset releases,\" \"Open weights now outpace truly open-source models\" and \"No single public conversation source generalizes across AI use\" (dataprovenance.org).", "claims": [ "F19", "F36", "F37", "F14" ] }, { "n": 19, "para": 5, "sense": "what the audits found", "text": "The journal finds \"a stark Western-centric skew in representation\" (Nature Machine Intelligence), MIT Sloan language \"heavily skewed toward English and Western European languages,\" and MIT Sloan states the stake: without lineage, users \"may find themselves out of compliance with emerging regulations like the European Union's AI Act or exposed to legal and copyright risks\" (MIT Sloan).", "claims": [ "F38", "F27", "F26" ] }, { "n": 20, "para": 6, "sense": "limits", "text": "Its stated limits: \"any output of the Data Provenance Initiative does not, and is not intended to, constitute legal advice\" (README); the licence protocol \"reflects best efforts to verify self-reported licences and does not constitute legal advice\" (Nature Machine Intelligence); \"We collect self-reported licenses from the papers and repositories that released these datasets and categorize them according to our best efforts\" (dataprovenance.org/about); the census tags it \"NOT legal advice (its own notice)\" and adds that \"describing a collection does not establish which model used it\" (provenance census record), and the Capture Registry finds both tags \"not composed\" at the name (Capture Registry, 2026-10-29).", "claims": [ "F43", "F39", "F47", "PE-16", "PE-17", "CAP-1-08" ] }, { "n": 21, "para": 7, "sense": "the bound", "text": "#725 states the bound in its canonical claim, that existing frameworks \"address the artifact (C2PA / Content Credentials) and the corpus (Data Provenance Initiative, EU AI Act transparency provisions, W3C PROV)\" and \"are not designed to address the survival of authorial lineage through AI synthesis\": the Initiative \"audits whether datasets were licensed, not whether synthesized outputs preserve attribution to the human sources whose labor the synthesis depended upon,\" stands at \"Layer 2 (licensing),\" \"Necessary but operates on corpus, not synthesis output,\" and licensing provenance \"operates at corpus-ingestion stage\" (#725).", "claims": [ "D725-01", "D725-04", "D725-10", "O725-11" ] }, { "n": 22, "para": 7, "sense": "the bound", "text": "\"Licensing audits operate on the input to AI systems. They do not address the output\" (#725, emphasis dropped): \"The Data Provenance Initiative tells you whether the dataset was licensed,\" and none of the instruments \"was designed to tell you what happened to the meaning between the camera and the screen, between the dataset and the answer\"; the path from the Initiative runs through \"dataset licensing\" to \"what about what the model produces from the licensed data?\" and so to semantic provenance, which extends provenance \"to the lifecycle of the meaning the content carries,\" with the Initiative as \"layer 2 of\" provenance after AI (#725).", "claims": [ "D725-08", "D725-09", "R725-12", "O725-07", "R725-13" ] }, { "n": 23, "para": 7, "sense": "the bound", "text": "#1003 groups the Initiative with C2PA and W3C PROV as \"Provenance and attribution studies\" that \"have focused on attribution loss in AI outputs\" (#1003), a reading composed beside #725's bound at the input.", "claims": [ "D1003-01" ] }, { "n": 24, "para": 8, "sense": "documentation and the pipeline", "text": "#98 tables the Initiative (MIT) as documenting \"provenance of popular AI training datasets,\" a \"Documentation project, not infrastructure,\" and Consent in Crisis as a \"Diagnostic paper,\" in the pattern \"They diagnose the crisis or build narrow fixes\"; it sets the Initiative at \"L4: Provenance / Consent,\" where what is missing is \"Integrated provenance in training pipeline,\" and records the archive's own provenance schema as taken from it, \"Implemented (in CHA)\" (#98).", "claims": [ "D98-01", "D98-02", "D98-05", "O98-03", "R98-04" ] }, { "n": 25, "para": 8, "sense": "documentation and the pipeline", "text": "#1714 draws the requirement, \"Provenance has to survive the training pipeline,\" and \"Sources can be recorded at ingestion, the record kept with the training set,\" with the Initiative's literature bearing \"on whether slushing is a fixed property of pipelines\" (#1714); #95 names \"training-data provenance records\" among the companion infrastructure for parametric claims, which PER reaches \"only indirectly\" (#95), and #623 calls standardized documentation of training-data provenance \"a step toward the accounting the framework demands\" (#623).", "claims": [ "D1714-02", "D1714-03", "D1714-04", "O95-01", "O623-01" ] }, { "n": 26, "para": 9, "sense": "reception", "text": "The home page lists coverage by \"the Financial Times, The New York Times, The Washington Post, MIT Technology Review, IEEE Spectrum, and VentureBeat\" and citation in \"public AI safety and governance reporting\" (dataprovenance.org); Wikipedia has no article on it (Wikipedia, 2026-10-29); #725 predicted that broad queries \"will be dominated by C2PA/CAI publications, DPI papers, and major media for the foreseeable future\" (#725).", "claims": [ "F10", "F44", "D725-14" ] }, { "n": 27, "para": 9, "sense": "reception", "text": "At the bare name on Google's AI Overview the Capture Registry reads the composition first as identifying \"the project,\" listing \"its tools\" and reporting \"what its audits found, from its own site, its paper and the press,\" then finds that its \"'multi-institutional research project—led in part by researchers from MIT'\" \"follows the MIT Sloan card,\" that \"the composition names no one\" (\"MIT named; no person; no identifier\"), and that \"no card cites\" the census record, which \"landed on provenanceerasure.org the same day\" (Capture Registry, 2026-10-29).", "claims": [ "CAP-1-09", "CAP-1-03", "CAP-1-04", "CAP-1-13", "CAP-1-12" ] } ], "joins": [ { "act": "order of senses", "sentences": "1–27", "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 (¶0) and its object, data provenance (¶1), precede the founding audit (¶2) and its instruments (¶3), which precede the systems now (¶4), since the extent grows from the first release; the findings (¶5) are the audits'; the limits (¶6) qualify the findings; the bound (¶7) and the pipeline (¶8) place the Initiative among the archive's dimensions; reception (¶9) stands last." }, { "act": "holding apart (the explorer relation)", "sentences": [ 10 ], "claims": [ "PE-18", "PE-19" ], "note": "P_coh within the census record: distinct records, and 'the same entity'; both composed with their loci (open block)" }, { "act": "holding apart (the scope of the Initiative's work)", "sentences": [ 21, 22, 23 ], "claims": [ "D725-04", "D725-08", "D1003-01" ], "note": "P_coh within the archive: #725 bounds the Initiative at the input; #1003 groups it with studies that 'have focused on attribution loss in AI outputs'. Each composed with its source; the composer does not rule (open block)." }, { "act": "holding apart (the founding count)", "sentences": [ 8 ], "claims": [ "F17", "F31", "F41", "PE-10", "F24" ], "note": "1800+ (preprint, README, census), 1,858 in 44 collections (journal), 'more than 1,800' (MIT Sloan): composed each with its source (open block)" }, { "act": "marking (contradicts_primary)", "sentences": [ 16 ], "claims": [ "CAP-1-01", "PE-12" ], "note": "the Capture Registry ascribes the 70%/50% quote to the README; the census gives it as the paper's; composed side by side, the mark carried in the ledger (§4.3)" }, { "act": "holding apart (the count now)", "sentences": [ 11, 12, 13 ], "claims": [ "F4", "F8", "PE-10", "CAP-1-05", "CAP-1-07" ], "note": "the home page's 4,000+ against the census's first-release count: both composed; the census is not corrected here (open block)" } ], "plan": { "order": [ "F40", "F45", "PE-02", "PE-04", "PE-03", "F23", "F9", "F46", "F1", "F2", "F3", "F30", "F18", "F15", "PE-22", "PE-05", "PE-06", "PE-09", "PE-07", "O725-03", "O725-05", "O725-15", "O725-06", "F16", "F29", "PE-15", "F17", "F31", "F41", "PE-10", "F24", "F22", "F34", "F42", "F25", "PE-08", "PE-01", "PE-18", "PE-19", "PE-13", "F4", "F5", "F6", "F7", "F8", "F11", "F12", "F13", "F28", "PE-14", "PE-20", "CAP-1-05", "CAP-1-06", "CAP-1-10", "CAP-1-07", "CAP-1-11", "F20", "F21", "PE-12", "PE-11", "PE-21", "F32", "F33", "CAP-1-02", "CAP-1-01", "F35", "D725-02", "D725-16", "D1714-01", "F19", "F36", "F37", "F14", "F38", "F27", "F26", "F43", "F39", "F47", "PE-16", "PE-17", "CAP-1-08", "D725-01", "D725-04", "D725-10", "O725-11", "D725-08", "D725-09", "R725-12", "O725-07", "R725-13", "D1003-01", "D98-01", "D98-02", "D98-05", "O98-03", "R98-04", "D1714-02", "D1714-03", "D1714-04", "O95-01", "O623-01", "F10", "F44", "D725-14", "CAP-1-09", "CAP-1-03", "CAP-1-04", "CAP-1-13", "CAP-1-12" ], "note": "carried claim ids in the KO's order (first citation)" } }, "P": { "title": "Data Provenance Initiative", "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 a volunteer collective of AI researchers auditing AI training data: 1,800+ text datasets in its 2023 first release, now four public measurement systems from datasets and web consent to model markets and real use (its pages); the archive sets it in the licensing dimension of provenance, bounded at the corpus (#725).", "claims": [ "F45", "F40", "F17", "F41", "F46", "F3", "D725-04", "O725-05" ] }, "sections": [ { "icon": "🗂️", "head": "What it is", "items": [ { "label": "Operator:", "text": "a \"volunteer effort\"; paper by Longpre and 16 others (README; arXiv; census); MIT researchers among its founders (MIT Sloan).", "claims": [ "F40", "F16", "PE-02", "PE-03", "PE-04", "F23", "F9" ] }, { "label": "Data provenance:", "text": "a dataset's \"sourcing, creation and licensing heritage\" (Nature MI), traced to source, creators, licences, use (arXiv; census).", "claims": [ "F30", "F18", "F15", "PE-06", "PE-07", "PE-09", "PE-05", "PE-22" ] }, { "label": "Its dimension:", "text": "training-corpus licensing: \"Under what permissions did this data enter this system?\" (#725).", "claims": [ "O725-03", "O725-05", "O725-15", "O725-06" ] } ] }, { "icon": "🧾", "head": "Origin and extent", "items": [ { "label": "Founding audit:", "text": "1800+ text datasets, 25 Oct 2023 (arXiv); 1,858 in 44 collections (Nature MI, 2024).", "claims": [ "F17", "F31", "F29", "F41", "F24", "PE-10", "PE-15" ] }, { "label": "Tools:", "text": "Explorer, Collection and Card, \"symbolic attribution for all constituent datasets\" (README).", "claims": [ "F22", "F34", "F42", "F25", "PE-01", "PE-08", "PE-13" ] }, { "label": "Now:", "text": "4,000+ datasets audited, 14,000 web domains tracked, 851k open models, 23,158 conversations (home page).", "claims": [ "F1", "F2", "F4", "F5", "F6", "F7", "F8", "F46" ] }, { "label": "Later audits:", "text": "Consent in Crisis (NeurIPS 2024); four thousand text, speech and video datasets (ICLR 2025).", "claims": [ "F11", "F12", "F13", "F28" ] }, { "label": "The record:", "text": "the census keeps the first release's count, the composition \"ahead of the record\" (Capture Registry).", "claims": [ "PE-20", "PE-14", "CAP-1-05", "CAP-1-06", "CAP-1-07", "CAP-1-10", "CAP-1-11" ] }, { "label": "One record or three:", "text": "\"distinct records\", and \"the same entity\" as the explorer's MMRS record (census).", "claims": [ "PE-18", "PE-19" ] } ] }, { "icon": "🔎", "head": "Findings", "items": [ { "label": "Licences:", "text": "omission 70%+, errors 50%+ on hosting sites (arXiv); 66% of Hugging Face licences recategorized (Nature MI).", "claims": [ "F20", "F21", "F32", "F33", "F24", "PE-12", "PE-11", "PE-21", "D1714-01", "D725-16", "CAP-1-01", "CAP-1-02" ] }, { "label": "Attribution:", "text": "85% of licences request it, 30% share-alike (Nature MI; #725).", "claims": [ "F35", "D725-02" ] }, { "label": "Divide:", "text": "closed datasets hold lower-resource languages and creative tasks (arXiv); non-commercial licences 61% in 2023 (Nature MI).", "claims": [ "F19", "F36", "F37", "F14" ] }, { "label": "Skew:", "text": "Western-centric (Nature MI; MIT Sloan).", "claims": [ "F38", "F27", "F26" ] } ] }, { "icon": "⚖️", "head": "Limits and bound", "items": [ { "label": "Limits:", "text": "not legal advice; self-reported licences (README; Nature MI; About); a collection does not show which model used it (census).", "claims": [ "F43", "F39", "F47", "PE-16", "PE-17", "CAP-1-08" ] }, { "label": "Bound:", "text": "\"audits whether datasets were licensed, not whether synthesized outputs preserve attribution\" (#725).", "claims": [ "D725-01", "D725-04", "D725-08", "D725-09", "D725-10", "O725-11", "O725-07", "R725-12", "R725-13" ] }, { "label": "Held beside:", "text": "studies of \"attribution loss in AI outputs\" (#1003).", "claims": [ "D1003-01" ] }, { "label": "Layer:", "text": "\"Documentation project, not infrastructure\"; L4; its schema \"Implemented (in CHA)\" (#98).", "claims": [ "D98-01", "D98-02", "D98-05", "O98-03", "R98-04" ] }, { "label": "Pipeline:", "text": "\"Provenance has to survive the training pipeline\" (#1714; #95; #623).", "claims": [ "D1714-02", "D1714-03", "D1714-04", "O95-01", "O623-01" ] } ] }, { "icon": "📰", "head": "Reception", "items": [ { "label": "Press:", "text": "major press (home page); no Wikipedia article; DPI papers predicted for broad queries (#725).", "claims": [ "F10", "F44", "D725-14" ] }, { "label": "Overview:", "text": "operator from the MIT Sloan card; no author; no card cites the record (Capture Registry).", "claims": [ "CAP-1-03", "CAP-1-04", "CAP-1-09", "CAP-1-12", "CAP-1-13" ] } ] } ], "open": [ { "question": "Is the founding audit 1800+ text datasets or 1,858 datasets in 44 collections?", "rests_on": [ "F17", "F41", "F31", "F24", "PE-10" ], "modality": "documented, both (the preprint and README against the journal)", "would_resolve": "the preprint v3's dataset tables counted against Nature Machine Intelligence Table 1" }, { "question": "Does the census count stand, or the home page's?", "rests_on": [ "PE-10", "F8", "F4", "F12", "CAP-1-07", "CAP-1-06" ], "modality": "documented, both (first release as of the README; '4,000+ datasets audited' on the home page as fetched 2026-10-29)", "would_resolve": "the census's re-read of dataprovenance.org, which the Capture Registry calls for" }, { "question": "Are the initiative, its explorer and its collection one entity or distinct records?", "rests_on": [ "PE-18", "PE-19", "PE-01" ], "modality": "stipulation, both (the census's own two statements)", "would_resolve": "the census editor's ruling; until it, this row and data-provenance-explorer stand as neighbours" }, { "question": "Is the Initiative's work bounded at the corpus, or a study of attribution loss in outputs?", "rests_on": [ "D725-04", "D725-08", "D1003-01" ], "modality": "stipulation, both (archive against archive)", "would_resolve": "the operator's ruling between #725 and #1003" }, { "question": "Where does the '70%+ … 50%+' quote stand: the repository README or the paper?", "rests_on": [ "CAP-1-01", "PE-12", "F20" ], "modality": "documented (the census's evidence names the paper; the saved README has no '70%')", "would_resolve": "the Capture Registry's finding read against the census's saved sources (reported, not fixed)" }, { "question": "Did synthetic data drive the 2023 non-commercial spike?", "rests_on": [ "F36", "F37" ], "modality": "documented, co-occurrence only (the spike and the synthetic share are reported separately)", "would_resolve": "the journal's licence-by-year data broken down by synthetic share" }, { "question": "Will broad provenance queries keep composing from DPI papers?", "rests_on": [ "D725-14" ], "modality": "hypothesis", "would_resolve": "later captures at broad provenance queries, seated in the Capture Registry" } ], "rail": [ { "lineage": "what it is: a volunteer collective auditing AI data", "claims": [ "F40", "F45", "PE-02", "PE-04", "PE-03", "F23", "F9" ], "earliest": "B3" }, { "lineage": "its extent now: public measurement infrastructure for AI data, models and use", "claims": [ "F46", "F1", "F2", "F3" ], "earliest": "B7" }, { "lineage": "the term: data provenance, its relations and its unit", "claims": [ "F30", "F18", "F15", "PE-22", "PE-05", "PE-06", "PE-09", "PE-07", "O725-03", "O725-05", "O725-15", "O725-06" ], "earliest": "B2" }, { "lineage": "the founding audit: 44 collections, 1,800+ text datasets (2023)", "claims": [ "F16", "F29", "PE-15", "F17", "F31", "F41", "PE-10" ], "earliest": "B2" }, { "lineage": "the Explorer, the Collection and the Card", "claims": [ "F22", "F34", "F42", "F25", "PE-08", "PE-01", "PE-18", "PE-19", "PE-13" ], "earliest": "B2" }, { "lineage": "the four systems and their counts", "claims": [ "F4", "F5", "F6", "F7", "F8", "PE-14", "PE-20", "CAP-1-05", "CAP-1-06", "CAP-1-10", "CAP-1-07", "CAP-1-11" ], "earliest": "B1" }, { "lineage": "the later audits: consent, multimodal, models, use", "claims": [ "F11", "F12", "F13", "F28" ], "earliest": "B3" }, { "lineage": "audit findings: licence omission and error on hosting sites", "claims": [ "F24", "F20", "F21", "PE-12", "PE-11", "PE-21", "F32", "F33", "CAP-1-02", "CAP-1-01", "D725-02", "D725-16", "D1714-01" ], "earliest": "B2" }, { "lineage": "audit findings: attribution and share-alike requested", "claims": [ "F35" ], "earliest": "B4" }, { "lineage": "audit findings: the commercial divide, the non-commercial spike, synthetic data", "claims": [ "F19", "F36", "F37", "F14" ], "earliest": "B2" }, { "lineage": "audit findings: the language and Western skew", "claims": [ "F38", "F27" ], "earliest": "B4" }, { "lineage": "the risks it addresses", "claims": [ "F26" ], "earliest": "B3" }, { "lineage": "stated limits: not legal advice; self-reported licences", "claims": [ "F43", "F39", "F47", "PE-16", "PE-17", "CAP-1-08" ], "earliest": "B4" }, { "lineage": "its bound: the corpus, not the output", "claims": [ "D725-01", "D725-04", "D725-10", "O725-11", "D725-08", "D725-09", "R725-12", "O725-07", "R725-13", "D1003-01" ], "earliest": 725 }, { "lineage": "documentation, not infrastructure; the archive's adoption", "claims": [ "D98-01", "D98-02", "D98-05", "O98-03", "R98-04" ], "earliest": 98 }, { "lineage": "provenance through the training pipeline", "claims": [ "D1714-02", "D1714-03", "D1714-04", "O95-01", "O623-01" ], "earliest": 623 }, { "lineage": "reception: press, the encyclopedia, the composition layer", "claims": [ "F10", "F44", "D725-14", "CAP-1-09", "CAP-1-03", "CAP-1-04", "CAP-1-13", "CAP-1-12" ], "earliest": 725 } ], "body_words": 350, "entity": "Composed under v0.9, equal case (ρ = extends): KO_B sentences surviving 0 of 13; every field claim carried." }, "P_B": { "title": "Data Provenance Initiative", "arm": "B (the field alone)", "genre": "popup — the compression from the field ledger alone (cards, own pages, Wikipedia)", "status": "composed 2026-10-29 from the field ledger alone; not frozen", "lede": { "text": "is a volunteer collective of AI researchers auditing AI training data, since its 2023 audit of 1,800+ text datasets, and now four public measurement systems (its pages; arXiv).", "claims": [ "F45", "F40", "F17", "F46", "F3" ] }, "sections": [ { "icon": "🗂️", "head": "What it is", "items": [ { "label": "Operator:", "text": "volunteers; MIT researchers among the founders (MIT Sloan).", "claims": [ "F40", "F23", "F9", "F1", "F2" ] }, { "label": "Data provenance:", "text": "\"sourcing, creation and licensing heritage\" (Nature MI).", "claims": [ "F30", "F18", "F15" ] } ] }, { "icon": "🧾", "head": "Origin and extent", "items": [ { "label": "Founding audit:", "text": "1,858 datasets, 44 collections (arXiv, 2023; Nature MI, 2024).", "claims": [ "F16", "F17", "F31", "F41", "F29", "F24" ] }, { "label": "Tools:", "text": "the Explorer and the Card (arXiv; README).", "claims": [ "F22", "F34", "F42", "F25" ] }, { "label": "Now:", "text": "4,000+ datasets, 14,000 domains, 851k models, 23,158 conversations.", "claims": [ "F4", "F5", "F6", "F7", "F8", "F46" ] }, { "label": "Later audits:", "text": "Consent in Crisis; the multimodal audit.", "claims": [ "F11", "F12", "F13", "F28" ] } ] }, { "icon": "🔎", "head": "Findings", "items": [ { "label": "Licences:", "text": "omission 70%+, errors 50%+ (arXiv); 66% recategorized (Nature MI).", "claims": [ "F20", "F21", "F32", "F33" ] }, { "label": "Divide and skew:", "text": "closed data holds creative and low-resource data; Western skew.", "claims": [ "F35", "F19", "F36", "F37", "F14", "F38", "F27", "F26" ] }, { "label": "Limits:", "text": "not legal advice; self-reported licences.", "claims": [ "F43", "F39", "F47" ] }, { "label": "Reception:", "text": "press coverage; no Wikipedia article.", "claims": [ "F10", "F44" ] } ] } ], "open": [ { "question": "Is the founding audit 1800+ or 1,858 datasets?", "rests_on": [ "F17", "F31" ], "modality": "documented, both", "would_resolve": "the preprint's tables against the journal's Table 1" }, { "question": "Did synthetic data drive the 2023 spike?", "rests_on": [ "F36", "F37" ], "modality": "documented, co-occurrence only", "would_resolve": "licence-by-year data by synthetic share" } ], "rail": [ { "lineage": "what it is: a volunteer collective auditing AI data", "claims": [ "F40", "F45", "F23", "F9" ], "earliest": "B3" }, { "lineage": "its extent now: public measurement infrastructure for AI data, models and use", "claims": [ "F46", "F1", "F2", "F3" ], "earliest": "B7" }, { "lineage": "the term: data provenance, its relations and its unit", "claims": [ "F30", "F18", "F15" ], "earliest": "B2" }, { "lineage": "the founding audit: 44 collections, 1,800+ text datasets (2023)", "claims": [ "F16", "F17", "F31", "F41", "F29" ], "earliest": "B2" }, { "lineage": "the Explorer, the Collection and the Card", "claims": [ "F22", "F34", "F42", "F25" ], "earliest": "B2" }, { "lineage": "the four systems and their counts", "claims": [ "F4", "F5", "F6", "F7", "F8" ], "earliest": "B1" }, { "lineage": "the later audits: consent, multimodal, models, use", "claims": [ "F11", "F12", "F13", "F28" ], "earliest": "B3" }, { "lineage": "audit findings: licence omission and error on hosting sites", "claims": [ "F24", "F20", "F21", "F32", "F33" ], "earliest": "B2" }, { "lineage": "audit findings: attribution and share-alike requested", "claims": [ "F35" ], "earliest": "B4" }, { "lineage": "audit findings: the commercial divide, the non-commercial spike, synthetic data", "claims": [ "F19", "F36", "F37", "F14" ], "earliest": "B2" }, { "lineage": "audit findings: the language and Western skew", "claims": [ "F38", "F27" ], "earliest": "B4" }, { "lineage": "the risks it addresses", "claims": [ "F26" ], "earliest": "B3" }, { "lineage": "stated limits: not legal advice; self-reported licences", "claims": [ "F43", "F39", "F47" ], "earliest": "B4" }, { "lineage": "reception: press, the encyclopedia, the composition layer", "claims": [ "F10", "F44" ], "earliest": "B1" } ], "body_words": 121, "entity": "With no archive claim admitted, the entry is the Initiative at its own resolution: its pages, its papers, its README, MIT Sloan." }, "KO_B": { "title": "The Data Provenance Initiative", "arm": "B (the field alone)", "genre": "knowledge object", "status": "composed 2026-10-29; not frozen", "sentences": [ { "n": 1, "para": 0, "text": "The Data Provenance Initiative describes itself as \"a multi-disciplinary volunteer effort to improve transparency, documentation, and responsible use of training datasets for AI\" (README) and as \"a volunteer collective of AI researchers conducting large-scale audits of the systems and resources that shape modern AI\" (About page).", "claims": [ "F40", "F45" ] }, { "n": 2, "para": 0, "text": "MIT Sloan says researchers \"including professor Sandy Pentland and others from MIT\" created it (MIT Sloan, 2025), and it releases \"papers, datasets, and dashboards\" for reasoning \"from shared evidence\" (home page).", "claims": [ "F23", "F9" ] }, { "n": 3, "para": 0, "text": "Its work began with training data provenance and now also covers web consent, open model markets, concentration of power and real-world use; the home page calls it \"Public measurement infrastructure for AI data, model ecosystems, and real-world AI use,\" with four public measurement systems (About page; home page).", "claims": [ "F46", "F1", "F2", "F3" ] }, { "n": 4, "para": 1, "text": "The systems are the Data Provenance Explorer (4,000+ datasets, 608 languages), the Web Consent Explorer (14,000 domains), Open Model Evolution (851k open models) and The AI Observatory (23,158 conversations) (home page).", "claims": [ "F4", "F5", "F6", "F7", "F8" ] }, { "n": 5, "para": 1, "text": "The founding audit, by Shayne Longpre and 16 other authors (arXiv:2310.16787, 25 October 2023; Nature Machine Intelligence, 30 August 2024), traced 1800+ text datasets, 1,858 in 44 collections, the Data Provenance Collection of its first release (arXiv; Nature; README; MIT Sloan).", "claims": [ "F16", "F17", "F31", "F41", "F29", "F24" ] }, { "n": 6, "para": 1, "text": "Its authors define data provenance as a dataset's \"sourcing, creation and licensing heritage, as well as its characteristics\" (Nature), traced from source and creators through licence conditions to subsequent use (arXiv; home page).", "claims": [ "F30", "F18", "F15" ] }, { "n": 7, "para": 2, "text": "It released the audit with an interactive UI, the Data Provenance Explorer, and a Data Provenance Card, \"symbolic attribution for all constituent datasets\" (arXiv; Nature; README; MIT Sloan).", "claims": [ "F22", "F34", "F42", "F25" ] }, { "n": 8, "para": 2, "text": "The audit found licence omission of 70%+ and error rates of 50%+ on hosting sites, a \"crisis in misattribution\" (arXiv), with over 70% of GitHub and Hugging Face licences unspecified and 66% of Hugging Face licences in another use category, \"licence laundering\" (Nature).", "claims": [ "F20", "F21", "F32", "F33" ] }, { "n": 9, "para": 2, "text": "85% of licences request attribution and 30% share-alike; closed datasets hold lower-resource languages and creative tasks; in 2023 non-commercial licences spiked to 61%, and 45% of such datasets are synthetic (Nature; arXiv; home page).", "claims": [ "F35", "F19", "F36", "F37", "F14" ] }, { "n": 10, "para": 3, "text": "Representation skews Western, and untraced data risks regulatory non-compliance and copyright exposure (Nature; MIT Sloan).", "claims": [ "F38", "F27", "F26" ] }, { "n": 11, "para": 3, "text": "Later audits followed: Consent in Crisis (NeurIPS 2024) and four thousand text, speech and video datasets (ICLR 2025), with work on open models and real-world use (home page; MIT Sloan).", "claims": [ "F11", "F12", "F13", "F28" ] }, { "n": 12, "para": 3, "text": "Its outputs do not constitute legal advice, and its licences are self-reported, categorized by best efforts (README; Nature; About page).", "claims": [ "F43", "F39", "F47" ] }, { "n": 13, "para": 3, "text": "The press has covered it, and Wikipedia has no article (home page; Wikipedia).", "claims": [ "F10", "F44" ] } ] }, "L_B": { "words": 419, "text": "The Data Provenance Initiative describes itself as \"a multi-disciplinary volunteer effort to improve transparency, documentation, and responsible use of training datasets for AI\" (README) and as \"a volunteer collective of AI researchers conducting large-scale audits of the systems and resources that shape modern AI\" (About page). MIT Sloan says researchers \"including professor Sandy Pentland and others from MIT\" created it (MIT Sloan, 2025), and it releases \"papers, datasets, and dashboards\" for reasoning \"from shared evidence\" (home page). Its work began with training data provenance and now also covers web consent, open model markets, concentration of power and real-world use; the home page calls it \"Public measurement infrastructure for AI data, model ecosystems, and real-world AI use,\" with four public measurement systems (About page; home page). The systems are the Data Provenance Explorer (4,000+ datasets, 608 languages), the Web Consent Explorer (14,000 domains), Open Model Evolution (851k open models) and The AI Observatory (23,158 conversations) (home page). The founding audit, by Shayne Longpre and 16 other authors (arXiv:2310.16787, 25 October 2023; Nature Machine Intelligence, 30 August 2024), traced 1800+ text datasets, 1,858 in 44 collections, the Data Provenance Collection of its first release (arXiv; Nature; README; MIT Sloan). Its authors define data provenance as a dataset's \"sourcing, creation and licensing heritage, as well as its characteristics\" (Nature), traced from source and creators through licence conditions to subsequent use (arXiv; home page). It released the audit with an interactive UI, the Data Provenance Explorer, and a Data Provenance Card, \"symbolic attribution for all constituent datasets\" (arXiv; Nature; README; MIT Sloan). The audit found licence omission of 70%+ and error rates of 50%+ on hosting sites, a \"crisis in misattribution\" (arXiv), with over 70% of GitHub and Hugging Face licences unspecified and 66% of Hugging Face licences in another use category, \"licence laundering\" (Nature). 85% of licences request attribution and 30% share-alike; closed datasets hold lower-resource languages and creative tasks; in 2023 non-commercial licences spiked to 61%, and 45% of such datasets are synthetic (Nature; arXiv; home page). Representation skews Western, and untraced data risks regulatory non-compliance and copyright exposure (Nature; MIT Sloan). Later audits followed: Consent in Crisis (NeurIPS 2024) and four thousand text, speech and video datasets (ICLR 2025), with work on open models and real-world use (home page; MIT Sloan). Its outputs do not constitute legal advice, and its licences are self-reported, categorized by best efforts (README; Nature; About page). The press has covered it, and Wikipedia has no article (home page; Wikipedia).", "note": "L_F(B) is realized as the field arm's expansion (objects.KO_B): the field alone, cards, own pages and Wikipedia." }, "L_BA": { "words": 2136, "text": "The Data Provenance Initiative is, in its repository's words, \"a multi-disciplinary volunteer effort to improve transparency, documentation, and responsible use of training datasets for AI,\" and on its About page \"a volunteer collective of AI researchers conducting large-scale audits of the systems and resources that shape modern AI\" (README; dataprovenance.org/about); the provenance census record gives its operator as the Initiative itself, a \"volunteer effort\" with a \"paper by Shayne Longpre and 16 other authors,\" kinds it an \"audit collection and explorer\" and sets it in the lineage of \"training data\" (provenance census record). MIT Sloan's account has \"a team of multidisciplinary researchers, including professor Sandy Pentland and others from MIT,\" create the Initiative \"to tackle the data transparency challenge head-on\" (MIT Sloan, 3 March 2025), and the home page says that \"DPI releases papers, datasets, and dashboards that help researchers, journalists, policymakers, and builders reason from shared evidence\" (dataprovenance.org). Its work, the About page says, \"began with training data provenance: tracing datasets from origin to creation, cataloging sources, licenses, creators, and documentation practices\" and \"now also covers web consent, open model markets, concentration of power, and how people use AI systems in practice\" (dataprovenance.org/about); the home page calls it \"Public measurement infrastructure for AI data, model ecosystems, and real-world AI use,\" which \"maps the datasets, web policies, model markets, and real-world AI uses that shape how AI systems are built and governed\" through \"Four public measurement systems\" (dataprovenance.org). The founding audit's authors name its object: \"we term the combination of these indicators, including a dataset's sourcing, creation and licensing heritage, as well as its characteristics, the 'data provenance'\" (Nature Machine Intelligence, 30 August 2024), traced \"from their source, creators, series of license conditions, properties, and subsequent use\" (arXiv:2310.16787), and the home page's \"Training Data Provenance\" traces \"the sources, licenses, creators, languages, geographies, and derivation chains behind public AI training datasets\" (dataprovenance.org). The provenance census asks of each entity it formalizes \"What provenance does this system carry, through which operation, on whose assertion, with what evidence of survival?\" and answers for the Initiative that it records, of the three capabilities express, record and test: its object \"a fine-tuning dataset, traced through its collection to its sources,\" its unit \"a dataset within a collection,\" its relations derivation, creators, licences and subsequent use, the last field restating the paper's lineage (provenance census record). #725 sets that object in the second dimension of provenance, \"training-corpus licensing,\" which \"has academic instrumentation and emerging legal architecture,\" its object \"The corpus\" and its question \"Under what permissions did this data enter this system?\", with the Initiative among its frameworks and among provenance's disciplinary homes as \"dataset documentation (DPI, model cards, dataset cards),\" and states that the dimensions of artifact, licensing and meaning are independent: \"a licensed dataset can produce outputs with excellent semantic provenance if the synthesis preserves attribution\" (#725). The founding paper was submitted on 25 October 2023 and revised on 4 November 2023, \"by Shayne Longpre and 16 other authors\" (arXiv:2310.16787), and published in Nature Machine Intelligence, volume 6, on 30 August 2024 (Nature Machine Intelligence); the census dates the entity from the preprint, \"2023 (arXiv:2310.16787, v1 25 October 2023)\" (provenance census record). It convened \"a multi-disciplinary effort between legal and machine learning experts to systematically audit and trace 1800+ text datasets\" (arXiv), in the journal \"44 instruction or 'alignment' finetuning data collections composed of 1,858 individual datasets\" (Nature Machine Intelligence), the release the repository calls the Data Provenance Collection and \"this initiative's first release\" (README), and the count the census carries, \"44 data collections, 1800+ fine-tuning text-to-text datasets (first release),\" as of the README (provenance census record); MIT Sloan gives \"more than 1,800 text datasets\" from \"44 of the most widely used text data collections\" (MIT Sloan). With the audit came \"an interactive UI, the Data Provenance Explorer, which allows practitioners to trace and filter on data provenance for the most popular open source finetuning data collections\" (arXiv), in the journal \"an open-source repository for downloading, filtering and exploring data provenance and characteristics\" whose tools \"auto-generate data provenance cards for scalable symbolic attribution\" (Nature Machine Intelligence); the repository's Data Provenance Card is \"symbolic attribution for all constituent datasets, that can be used as structured documentation\" (README), \"human-readable\" in MIT Sloan's account (MIT Sloan), and the census carries it as the entity's provenance carrier beside \"audit metadata per dataset\" (provenance census record). The census gives Data Provenance Explorer, Data Provenance Collection and Data Provenance Card as other names of the entity and states its relation to them twice: among its negative tags, \"initiative, explorer and collection are distinct records,\" and on its sibling relation to the MMRS meta-aggregator's record of the explorer, \"the same entity, read for reception there and for provenance carriage here\"; it records the Explorer as \"a JavaScript application, unreadable to a plain fetch\" (provenance census record). The home page's four systems are the Data Provenance Explorer (\"4,000+ datasets,\" \"608 languages across text, speech, and video,\" inspecting \"dataset sources, licenses, creators, geographies, modalities, and derivation chains\"), the Web Consent Explorer (\"14,000 domains,\" \"Terms, robots.txt, and crawl restrictions\"), Open Model Evolution (\"851k open models,\" \"2.2B downloads\") and The AI Observatory (\"23,158 conversations,\" \"85,633 turns from 7 public sources\") (dataprovenance.org). Its metrics read \"4,000+\" \"datasets audited,\" \"14,000\" \"web domains tracked,\" \"851k\" \"open models measured\" and \"23,158\" \"AI conversations annotated,\" and its publications carry the later work: Consent in Crisis (NeurIPS 2024), the multimodal audit, \"An audit of four thousand public text, speech, and video datasets\" (ICLR 2025), Economies of Open Intelligence (preprint) and The AI Observatory (NeurIPS 2026) (dataprovenance.org); in March 2025 MIT Sloan still had the work \"currently focused on the data provenance of text,\" with a plan \"to expand to other media, such as video\" (MIT Sloan). The census carries the entity as \"active\" and lists \"the Explorer itself (JavaScript)\" and \"the multimodal extension (arXiv:2412.17847)\" among what it has not yet checked (provenance census record); the Capture Registry finds that \"The record has one explorer and the first release's count,\" that the home-page card gives \"four platforms and '4,000+ audited datasets; 14,000 tracked web domains',\" that the composition is \"ahead of the record on the entity's current extent,\" and that \"The census is to re-read dataprovenance.org and carry the platforms and the current count with their quotes,\" the capture being \"a correction to be made from the entity's own page\" (Capture Registry, 2026-10-29). The founding audit observed \"frequent miscategorization of licenses on widely used dataset hosting sites, with license omission of 70%+ and error rates of 50%+,\" which \"points to a crisis in misattribution and informed use of the most popular datasets driving many recent breakthroughs\" (arXiv), the result the census records as what survived the operation it tested, \"dataset repackaging and hosting (licence carried or lost on hosting sites)\" (provenance census record). The journal gives the measure by platform, \"more than 70% of licences for popular datasets on GitHub and Hugging Face are 'unspecified'\" and \"66% of analysed Hugging Face licences were in a different use category, often labelled as more permissive than the author's original licence,\" and names \"a crisis in licence laundering\" (Nature Machine Intelligence); MIT Sloan restates error rates \"greater than 50%\" and omission rates \"of more than 70%\" (MIT Sloan). The Capture Registry checked these figures \"against the provenance census record of the same day and the entity pages it saved\" and calls the 70%/50% wording \"the record's quote from the repository README\" (Capture Registry, 2026-10-29), where the census gives it as what \"the founding paper reports\" (provenance census record). \"In total, 85% of dataset licences request attribution, and 30% include a share-alike clause\" (Nature Machine Intelligence); #725 carries these figures with the omission and error rates and reads them as the licensing dimension's ground, \"The DPI's documentation of 70%+ license-omission rates establishes the licensing dimension's urgency\" (#725), and #1714 reads the finding as \"how much is lost before training even begins\" (#1714). Of the commercial divide the preprint finds \"closed datasets monopolizing important categories: lower resource languages, more creative tasks, richer topic variety, newer and more synthetic training data\" (arXiv); the journal records that \"In 2023 there was a spike in NC/A-O dataset licences,\" \"the NC/A-O rate is 61%\" that year, and that \"a full 45% of NC/A-O datasets are synthetic, compared to <14% in more permissive licence categories\" (Nature Machine Intelligence); the home page's findings are \"Source content often carries stricter terms than dataset releases,\" \"Open weights now outpace truly open-source models\" and \"No single public conversation source generalizes across AI use\" (dataprovenance.org). The journal finds \"a stark Western-centric skew in representation\" (Nature Machine Intelligence), MIT Sloan language \"heavily skewed toward English and Western European languages,\" and MIT Sloan states the stake: without lineage, users \"may find themselves out of compliance with emerging regulations like the European Union's AI Act or exposed to legal and copyright risks\" (MIT Sloan). Its stated limits: \"any output of the Data Provenance Initiative does not, and is not intended to, constitute legal advice\" (README); the licence protocol \"reflects best efforts to verify self-reported licences and does not constitute legal advice\" (Nature Machine Intelligence); \"We collect self-reported licenses from the papers and repositories that released these datasets and categorize them according to our best efforts\" (dataprovenance.org/about); the census tags it \"NOT legal advice (its own notice)\" and adds that \"describing a collection does not establish which model used it\" (provenance census record), and the Capture Registry finds both tags \"not composed\" at the name (Capture Registry, 2026-10-29). #725 states the bound in its canonical claim, that existing frameworks \"address the artifact (C2PA / Content Credentials) and the corpus (Data Provenance Initiative, EU AI Act transparency provisions, W3C PROV)\" and \"are not designed to address the survival of authorial lineage through AI synthesis\": the Initiative \"audits whether datasets were licensed, not whether synthesized outputs preserve attribution to the human sources whose labor the synthesis depended upon,\" stands at \"Layer 2 (licensing),\" \"Necessary but operates on corpus, not synthesis output,\" and licensing provenance \"operates at corpus-ingestion stage\" (#725). \"Licensing audits operate on the input to AI systems. They do not address the output\" (#725, emphasis dropped): \"The Data Provenance Initiative tells you whether the dataset was licensed,\" and none of the instruments \"was designed to tell you what happened to the meaning between the camera and the screen, between the dataset and the answer\"; the path from the Initiative runs through \"dataset licensing\" to \"what about what the model produces from the licensed data?\" and so to semantic provenance, which extends provenance \"to the lifecycle of the meaning the content carries,\" with the Initiative as \"layer 2 of\" provenance after AI (#725). #1003 groups the Initiative with C2PA and W3C PROV as \"Provenance and attribution studies\" that \"have focused on attribution loss in AI outputs\" (#1003), a reading composed beside #725's bound at the input. #98 tables the Initiative (MIT) as documenting \"provenance of popular AI training datasets,\" a \"Documentation project, not infrastructure,\" and Consent in Crisis as a \"Diagnostic paper,\" in the pattern \"They diagnose the crisis or build narrow fixes\"; it sets the Initiative at \"L4: Provenance / Consent,\" where what is missing is \"Integrated provenance in training pipeline,\" and records the archive's own provenance schema as taken from it, \"Implemented (in CHA)\" (#98). #1714 draws the requirement, \"Provenance has to survive the training pipeline,\" and \"Sources can be recorded at ingestion, the record kept with the training set,\" with the Initiative's literature bearing \"on whether slushing is a fixed property of pipelines\" (#1714); #95 names \"training-data provenance records\" among the companion infrastructure for parametric claims, which PER reaches \"only indirectly\" (#95), and #623 calls standardized documentation of training-data provenance \"a step toward the accounting the framework demands\" (#623). The home page lists coverage by \"the Financial Times, The New York Times, The Washington Post, MIT Technology Review, IEEE Spectrum, and VentureBeat\" and citation in \"public AI safety and governance reporting\" (dataprovenance.org); Wikipedia has no article on it (Wikipedia, 2026-10-29); #725 predicted that broad queries \"will be dominated by C2PA/CAI publications, DPI papers, and major media for the foreseeable future\" (#725). At the bare name on Google's AI Overview the Capture Registry reads the composition first as identifying \"the project,\" listing \"its tools\" and reporting \"what its audits found, from its own site, its paper and the press,\" then finds that its \"'multi-institutional research project—led in part by researchers from MIT'\" \"follows the MIT Sloan card,\" that \"the composition names no one\" (\"MIT named; no person; no identifier\"), and that \"no card cites\" the census record, which \"landed on provenanceerasure.org the same day\" (Capture Registry, 2026-10-29).", "rail": [], "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." } }, "field": [ { "id": "B1", "card": "Data Provenance Initiative — Data Provenance Initiative (home)", "fetched": "card; HTTP 200, B1.html sha256 bf609f385257c3a1; text from the script bundle (site-main.af19d7fc.js)", "origin": "card" }, { "id": "B2", "card": "arXiv — [2310.16787] The Data Provenance Initiative: A Large Scale Audit of Dataset Licensing & Attribution in AI", "fetched": "card; HTTP 200, B2.html sha256 d01552052066b391", "origin": "card" }, { "id": "B3", "card": "MIT Sloan — Bringing transparency to the data used to train artificial intelligence", "fetched": "card; HTTP 200, B3.html sha256 563661c73ddce481", "origin": "card" }, { "id": "B4", "card": "Nature Machine Intelligence — A large-scale audit of dataset licensing and attribution in AI", "fetched": "card; HTTP 200, B4.html sha256 766eb441fb2e78db", "origin": "card" }, { "id": "B5", "card": "Data Provenance Initiative — Data-Provenance-Collection README", "fetched": "own_page; the census's saved copy (read 2026-10-29, HTTP 200), B5.html sha256 db4c35a5f4f2b17a; not refetched", "origin": "own_page" }, { "id": "B6", "card": "Wikipedia — Data Provenance Initiative (no article)", "fetched": "no article: raw HTTP 404, rendered HTTP 404 (one 429 retried), search no results; sha256 64f4d986f0e9aa62", "origin": "wikipedia" }, { "id": "B7", "card": "Data Provenance Initiative — About", "fetched": "own_page; HTTP 200, B7.html sha256 bf609f385257c3a1; text from the script bundle (site-main.af19d7fc.js)", "origin": "own_page" } ], "field_claims": { "F1": { "source": "B1", "origin": "card", "locus": "served HTML: meta description", "quote": "Public measurement infrastructure for AI data, model ecosystems, and real-world AI use.", "claim": "the Initiative's self-description: public measurement infrastructure for AI data, model ecosystems and real-world AI use", "modality": "self-description", "kind": "definition", "kernel": true, "const": false, "lineage": "its extent now: public measurement infrastructure for AI data, models and use", "date": "2026-10-29 (page undated; as fetched)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 86ba9a3ecee76334…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F2": { "source": "B1", "origin": "card", "locus": "home (Cr): hero", "quote": "Public evidence for accountable AI … DPI maps the datasets, web policies, model markets, and real-world AI uses that shape how AI systems are built and governed.", "claim": "what it does: maps the datasets, web policies, model markets and real-world uses that shape how AI is built and governed", "modality": "stipulation", "kind": "definition", "kernel": true, "const": true, "lineage": "its extent now: public measurement infrastructure for AI data, models and use", "date": "2026-10-29 (page undated; as fetched)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 86ba9a3ecee76334…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F3": { "source": "B1", "origin": "card", "locus": "home (Cr): 'Flagship systems'", "quote": "Four public measurement systems … Together they trace AI from source data and web consent to model markets and real-world use.", "claim": "four public measurement systems, tracing AI from source data and web consent to model markets and real-world use", "modality": "self-description", "kind": "fact", "kernel": false, "const": false, "lineage": "its extent now: public measurement infrastructure for AI data, models and use", "date": "2026-10-29 (page undated; as fetched)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 86ba9a3ecee76334…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F4": { "source": "B1", "origin": "card", "locus": "home (Cr): system card 'Data Provenance Explorer' (constant xr)", "quote": "Data Provenance Explorer … Training data … 4,000+ datasets … 608 languages across text, speech, and video … Inspect dataset sources, licenses, creators, geographies, modalities, and derivation chains.", "claim": "the Data Provenance Explorer: training data, 4,000+ datasets in 608 languages across text, speech and video; sources, licences, creators, geographies, modalities, derivation chains", "modality": "documented", "kind": "fact", "kernel": true, "const": false, "lineage": "the four systems and their counts", "date": "2026-10-29 (page undated; as fetched)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 86ba9a3ecee76334…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F5": { "source": "B1", "origin": "card", "locus": "home (Cr): system card 'Web Consent Explorer'", "quote": "Web Consent Explorer … Web data … 14,000 domains … Terms, robots.txt, and crawl restrictions … Track how web consent signals are changing as AI developers compete for training data.", "claim": "the Web Consent Explorer: 14,000 domains, their terms, robots.txt and crawl restrictions", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "the four systems and their counts", "date": "2026-10-29 (page undated; as fetched)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 86ba9a3ecee76334…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F6": { "source": "B1", "origin": "card", "locus": "home (Cr): system card 'Open Model Evolution'", "quote": "Open Model Evolution … Model ecosystem … 851k open models … 2.2B downloads and 200+ model attributes … Monitor concentration, participation, transparency, and technical shifts in open-weight AI.", "claim": "Open Model Evolution: 851k open models, 2.2B downloads; concentration, participation and transparency in open-weight AI", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "the four systems and their counts", "date": "2026-10-29 (page undated; as fetched)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 86ba9a3ecee76334…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F7": { "source": "B1", "origin": "card", "locus": "home (Cr): system card 'The AI Observatory'", "quote": "The AI Observatory … Real-world use … 23,158 conversations … 85,633 turns from 7 public sources … Compare how people use and misuse AI systems across sources, models, and contexts.", "claim": "The AI Observatory: 23,158 conversations, 85,633 turns from 7 public sources; how people use and misuse AI", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "the four systems and their counts", "date": "2026-10-29 (page undated; as fetched)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 86ba9a3ecee76334…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F8": { "source": "B1", "origin": "card", "locus": "home (Cr): metrics strip (constant br)", "quote": "4,000+ … datasets audited … 14,000 … web domains tracked … 851k … open models measured … 23,158 … AI conversations annotated", "claim": "the metrics: 4,000+ datasets audited, 14,000 web domains tracked, 851k open models measured, 23,158 conversations annotated", "modality": "documented", "kind": "fact", "kernel": true, "const": false, "lineage": "the four systems and their counts", "date": "2026-10-29 (page undated; as fetched)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 86ba9a3ecee76334…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F9": { "source": "B1", "origin": "card", "locus": "home (Cr): 'Public-interest infrastructure'", "quote": "DPI releases papers, datasets, and dashboards that help researchers, journalists, policymakers, and builders reason from shared evidence.", "claim": "the Initiative releases papers, datasets and dashboards for reasoning from shared evidence", "modality": "self-description", "kind": "function", "kernel": false, "const": false, "lineage": "what it is: a volunteer collective auditing AI data", "date": "2026-10-29 (page undated; as fetched)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 86ba9a3ecee76334…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F10": { "source": "B1", "origin": "card", "locus": "home (Cr): 'Policy and governance relevance'; 'Featured coverage'", "quote": "Our work has been cited in public AI safety and governance reporting and used to ground debates about data transparency, consent, and market concentration. … DPI findings have been covered by the Financial Times, The New York Times, The Washington Post, MIT Technology Review, IEEE Spectrum, and VentureBeat.", "claim": "its account of its uptake: governance reporting and press coverage", "modality": "self-description", "kind": "fact", "kernel": false, "const": false, "lineage": "reception: press, the encyclopedia, the composition layer", "date": "2026-10-29 (page undated; as fetched)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 86ba9a3ecee76334…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F11": { "source": "B1", "origin": "card", "locus": "home (Cr): publications, 'Consent in Crisis'", "quote": "Consent in Crisis: The Rapid Decline of the AI Data Commons … An audit of terms, restrictions, and consent signals across the web. … NeurIPS 2024", "claim": "Consent in Crisis (NeurIPS 2024): an audit of terms, restrictions and consent signals across the web", "modality": "documented", "kind": "genealogy", "kernel": false, "const": false, "lineage": "the later audits: consent, multimodal, models, use", "date": "2024 (NeurIPS 2024, as the home page dates it)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 86ba9a3ecee76334…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F12": { "source": "B1", "origin": "card", "locus": "home (Cr): publications, the multimodal audit", "quote": "Bridging the Data Provenance Gap Across Text, Speech, and Video … An audit of four thousand public text, speech, and video datasets. … ICLR 2025", "claim": "the multimodal audit (ICLR 2025): four thousand public text, speech and video datasets", "modality": "documented", "kind": "genealogy", "kernel": true, "const": false, "lineage": "the later audits: consent, multimodal, models, use", "date": "2025 (ICLR 2025, as the home page dates it)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 86ba9a3ecee76334…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F13": { "source": "B1", "origin": "card", "locus": "home (Cr): publications, the model economy and real-world use", "quote": "The AI Observatory: A Public Measure of Real-World AI Use … NeurIPS 2026 … Economies of Open Intelligence: Tracing Power and Participation in the Model Ecosystem … Preprint", "claim": "its newest papers: the AI Observatory (NeurIPS 2026) and Economies of Open Intelligence (preprint)", "modality": "documented", "kind": "genealogy", "kernel": false, "const": false, "lineage": "the later audits: consent, multimodal, models, use", "date": "2026 (NeurIPS 2026, as the home page dates it)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 86ba9a3ecee76334…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F14": { "source": "B1", "origin": "card", "locus": "home (Cr): findings strip", "quote": "Data transparency … Open weights now outpace truly open-source models … Dataset restrictions … Source content often carries stricter terms than dataset releases … Real-use heterogeneity … No single public conversation source generalizes across AI use", "claim": "the home page's three findings: open weights outpace open source; source content stricter than dataset releases; no single conversation source generalizes", "modality": "interpretation", "kind": "interpretation", "kernel": false, "const": false, "lineage": "audit findings: the commercial divide, the non-commercial spike, synthetic data", "date": "2026-10-29 (page undated; as fetched)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 86ba9a3ecee76334…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F15": { "source": "B1", "origin": "card", "locus": "home (Cr): 'Training Data Provenance'", "quote": "Training Data Provenance … Dataset audits … Trace the sources, licenses, creators, languages, geographies, and derivation chains behind public AI training datasets.", "claim": "training data provenance as dataset audits: the sources, licences, creators, languages, geographies and derivation chains behind public training datasets", "modality": "stipulation", "kind": "definition", "kernel": false, "const": false, "lineage": "the term: data provenance, its relations and its unit", "date": "2026-10-29 (page undated; as fetched)", "quote_ok": true, "checked_against": "saved text B1.txt (sha256 86ba9a3ecee76334…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F16": { "source": "B2", "origin": "card", "locus": "abstract page header", "quote": "[Submitted on 25 Oct 2023 (v1), last revised 4 Nov 2023 (this version, v3)] … The Data Provenance Initiative: A Large Scale Audit of Dataset Licensing & Attribution in AI … by Shayne Longpre and 16 other authors", "claim": "the founding paper, submitted 25 October 2023 (v3 4 November), by Shayne Longpre and 16 other authors", "modality": "documented", "kind": "attribution", "kernel": false, "const": false, "lineage": "the founding audit: 44 collections, 1,800+ text datasets (2023)", "date": "2023-10-25 (v1 submitted; v3 2023-11-04)", "quote_ok": true, "checked_against": "saved text B2.txt (sha256 0063d0167cbd2144…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F17": { "source": "B2", "origin": "card", "locus": "Abstract", "quote": "we convene a multi-disciplinary effort between legal and machine learning experts to systematically audit and trace 1800+ text datasets.", "claim": "the founding audit: legal and ML experts auditing and tracing 1800+ text datasets", "modality": "stipulation", "kind": "definition", "kernel": true, "const": true, "lineage": "the founding audit: 44 collections, 1,800+ text datasets (2023)", "date": "2023-10-25 (v1 submitted; v3 2023-11-04)", "quote_ok": true, "checked_against": "saved text B2.txt (sha256 0063d0167cbd2144…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F18": { "source": "B2", "origin": "card", "locus": "Abstract", "quote": "We develop tools and standards to trace the lineage of these datasets, from their source, creators, series of license conditions, properties, and subsequent use.", "claim": "the lineage traced: source, creators, licence conditions, properties, subsequent use", "modality": "stipulation", "kind": "definition", "kernel": true, "const": true, "lineage": "the term: data provenance, its relations and its unit", "date": "2023-10-25 (v1 submitted; v3 2023-11-04)", "quote_ok": true, "checked_against": "saved text B2.txt (sha256 0063d0167cbd2144…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F19": { "source": "B2", "origin": "card", "locus": "Abstract", "quote": "Our landscape analysis highlights the sharp divides in composition and focus of commercially open vs closed datasets, with closed datasets monopolizing important categories: lower resource languages, more creative tasks, richer topic variety, newer and more synthetic training data.", "claim": "closed datasets monopolize lower-resource languages, creative tasks, topic variety and newer synthetic data", "modality": "interpretation", "kind": "interpretation", "kernel": true, "const": false, "lineage": "audit findings: the commercial divide, the non-commercial spike, synthetic data", "date": "2023-10-25 (v1 submitted; v3 2023-11-04)", "quote_ok": true, "checked_against": "saved text B2.txt (sha256 0063d0167cbd2144…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F20": { "source": "B2", "origin": "card", "locus": "Abstract", "quote": "We also observe frequent miscategorization of licenses on widely used dataset hosting sites, with license omission of 70%+ and error rates of 50%+.", "claim": "on hosting sites, licence omission of 70%+ and error rates of 50%+", "modality": "documented", "kind": "fact", "kernel": true, "const": false, "lineage": "audit findings: licence omission and error on hosting sites", "date": "2023-10-25 (v1 submitted; v3 2023-11-04)", "quote_ok": true, "checked_against": "saved text B2.txt (sha256 0063d0167cbd2144…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F21": { "source": "B2", "origin": "card", "locus": "Abstract", "quote": "This points to a crisis in misattribution and informed use of the most popular datasets driving many recent breakthroughs.", "claim": "read as a crisis in misattribution and informed use", "modality": "interpretation", "kind": "interpretation", "kernel": false, "const": false, "lineage": "audit findings: licence omission and error on hosting sites", "date": "2023-10-25 (v1 submitted; v3 2023-11-04)", "quote_ok": true, "checked_against": "saved text B2.txt (sha256 0063d0167cbd2144…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F22": { "source": "B2", "origin": "card", "locus": "Abstract", "quote": "we release our entire audit, with an interactive UI, the Data Provenance Explorer, which allows practitioners to trace and filter on data provenance for the most popular open source finetuning data collections", "claim": "the whole audit released with an interactive UI, the Data Provenance Explorer", "modality": "stipulation", "kind": "definition", "kernel": true, "const": false, "lineage": "the Explorer, the Collection and the Card", "date": "2023-10-25 (v1 submitted; v3 2023-11-04)", "quote_ok": true, "checked_against": "saved text B2.txt (sha256 0063d0167cbd2144…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F23": { "source": "B3", "origin": "card", "locus": "article body ¶4", "quote": "A team of multidisciplinary researchers, including professor Sandy Pentland and others from MIT, created the Data Provenance Initiative to tackle the data transparency challenge head-on.", "claim": "MIT Sloan: multidisciplinary researchers, Sandy Pentland and others from MIT among them, created the Initiative", "modality": "attributed", "kind": "genealogy", "kernel": true, "const": false, "lineage": "what it is: a volunteer collective auditing AI data", "date": "2025-03-03 (published)", "quote_ok": true, "checked_against": "saved text B3.txt (sha256 da84c05f68986b58…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F24": { "source": "B3", "origin": "card", "locus": "'two key actions', first", "quote": "Conducted a systematic audit of more than 1,800 text datasets. The researchers traced the lineage of fine-tuning datasets from 44 of the most widely used text data collections. They found that licenses were frequently miscategorized, with error rates greater than 50% and license information omission rates of more than 70%.", "claim": "MIT Sloan's account of the audit: 1,800+ text datasets from 44 collections; error rates above 50%, omission above 70%", "modality": "attributed", "kind": "fact", "kernel": false, "const": false, "lineage": "audit findings: licence omission and error on hosting sites", "date": "2025-03-03 (published)", "quote_ok": true, "checked_against": "saved text B3.txt (sha256 da84c05f68986b58…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F25": { "source": "B3", "origin": "card", "locus": "'two key actions', second", "quote": "Practitioners can also use the tool to generate a human-readable data provenance card for datasets", "claim": "MIT Sloan: the tool generates a human-readable data provenance card", "modality": "attributed", "kind": "function", "kernel": false, "const": false, "lineage": "the Explorer, the Collection and the Card", "date": "2025-03-03 (published)", "quote_ok": true, "checked_against": "saved text B3.txt (sha256 da84c05f68986b58…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F26": { "source": "B3", "origin": "card", "locus": "article body ¶2", "quote": "Without transparency into the lineage of data used for artificial intelligence models, researchers, businesses, and other intended users may find themselves out of compliance with emerging regulations like the European Union’s AI Act or exposed to legal and copyright risks.", "claim": "the risk without lineage: non-compliance with regulation, legal and copyright exposure", "modality": "interpretation", "kind": "interpretation", "kernel": true, "const": false, "lineage": "the risks it addresses", "date": "2025-03-03 (published)", "quote_ok": true, "checked_against": "saved text B3.txt (sha256 da84c05f68986b58…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F27": { "source": "B3", "origin": "card", "locus": "'other issues', second", "quote": "Language used in the datasets is heavily skewed toward English and Western European languages, with sparse coverage of languages from Asian, African, and South American nations, if at all.", "claim": "MIT Sloan: language skewed to English and Western European languages", "modality": "attributed", "kind": "fact", "kernel": false, "const": false, "lineage": "audit findings: the language and Western skew", "date": "2025-03-03 (published)", "quote_ok": true, "checked_against": "saved text B3.txt (sha256 da84c05f68986b58…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F28": { "source": "B3", "origin": "card", "locus": "article body, closing ¶ before the quote", "quote": "While the group’s efforts are currently focused on the data provenance of text, the plan is to expand to other media, such as video, as well as domain-specific data", "claim": "MIT Sloan (March 2025): the work focused on text, with a plan to expand to video and domain-specific data", "modality": "attributed", "kind": "genealogy", "kernel": false, "const": false, "lineage": "the later audits: consent, multimodal, models, use", "date": "2025-03-03 (published)", "quote_ok": true, "checked_against": "saved text B3.txt (sha256 da84c05f68986b58…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F29": { "source": "B4", "origin": "card", "locus": "article header", "quote": "Published: 30 August 2024 … Nature Machine Intelligence … volume 6, pages 975–987 (2024)", "claim": "the journal version, Nature Machine Intelligence 6, 975–987, published 30 August 2024", "modality": "documented", "kind": "genealogy", "kernel": false, "const": false, "lineage": "the founding audit: 44 collections, 1,800+ text datasets (2023)", "date": "2024-08-30 (published)", "quote_ok": true, "checked_against": "saved text B4.txt (sha256 e9b2e293b13212b1…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F30": { "source": "B4", "origin": "card", "locus": "Introduction ¶3", "quote": "In this work, we term the combination of these indicators, including a dataset’s sourcing, creation and licensing heritage, as well as its characteristics, the ‘data provenance’.", "claim": "data provenance defined: a dataset's sourcing, creation and licensing heritage and its characteristics", "modality": "stipulation", "kind": "definition", "kernel": true, "const": true, "lineage": "the term: data provenance, its relations and its unit", "date": "2024-08-30 (published)", "quote_ok": true, "checked_against": "saved text B4.txt (sha256 e9b2e293b13212b1…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F31": { "source": "B4", "origin": "card", "locus": "Introduction ¶6", "quote": "As a first contribution of the initiative, we audit 44 instruction or ‘alignment’ finetuning data collections composed of 1,858 individual datasets", "claim": "the first contribution: 44 alignment fine-tuning collections, 1,858 datasets", "modality": "documented", "kind": "fact", "kernel": true, "const": false, "lineage": "the founding audit: 44 collections, 1,800+ text datasets (2023)", "date": "2024-08-30 (published)", "quote_ok": true, "checked_against": "saved text B4.txt (sha256 e9b2e293b13212b1…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F32": { "source": "B4", "origin": "card", "locus": "Introduction ¶5", "quote": "Notably, we find that more than 70% of licences for popular datasets on GitHub and Hugging Face are ‘unspecified’ … our rigorous re-annotation of licences finds that 66% of analysed Hugging Face licences were in a different use category, often labelled as more permissive than the author’s original licence.", "claim": "over 70% of licences on GitHub and Hugging Face unspecified; 66% of Hugging Face licences in a different use category, often more permissive", "modality": "documented", "kind": "fact", "kernel": true, "const": false, "lineage": "audit findings: licence omission and error on hosting sites", "date": "2024-08-30 (published)", "quote_ok": true, "checked_against": "saved text B4.txt (sha256 e9b2e293b13212b1…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F33": { "source": "B4", "origin": "card", "locus": "Introduction ¶5", "quote": "Our work highlights a crisis in licence laundering and informed usage of popular datasets", "claim": "a crisis in licence laundering", "modality": "interpretation", "kind": "interpretation", "kernel": false, "const": false, "lineage": "audit findings: licence omission and error on hosting sites", "date": "2024-08-30 (published)", "quote_ok": true, "checked_against": "saved text B4.txt (sha256 e9b2e293b13212b1…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F34": { "source": "B4", "origin": "card", "locus": "contributions, second", "quote": "The Data Provenance Explorer (DPExplorer) (www.dataprovenance.org), an open-source repository for downloading, filtering and exploring data provenance and characteristics. Our tools auto-generate data provenance cards for scalable symbolic attribution and future documentation best practices.", "claim": "the DPExplorer: an open-source repository for downloading, filtering and exploring provenance; auto-generated provenance cards", "modality": "stipulation", "kind": "function", "kernel": false, "const": false, "lineage": "the Explorer, the Collection and the Card", "date": "2024-08-30 (published)", "quote_ok": true, "checked_against": "saved text B4.txt (sha256 e9b2e293b13212b1…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F35": { "source": "B4", "origin": "card", "locus": "licence findings", "quote": "In total, 85% of dataset licences request attribution, and 30% include a share-alike clause", "claim": "85% of licences request attribution, 30% share-alike", "modality": "documented", "kind": "fact", "kernel": true, "const": false, "lineage": "audit findings: attribution and share-alike requested", "date": "2024-08-30 (published)", "quote_ok": true, "checked_against": "saved text B4.txt (sha256 e9b2e293b13212b1…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F36": { "source": "B4", "origin": "card", "locus": "'In 2023 there was a spike in NC/A-O dataset licences'", "quote": "In 2023 there was a spike in NC/A-O dataset licences … in 2023, when many of the most popular and diverse datasets were published, the NC/A-O rate is 61%.", "claim": "a 2023 spike in non-commercial/academic-only licences: 61% of that year's datasets", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "audit findings: the commercial divide, the non-commercial spike, synthetic data", "date": "2024-08-30 (published)", "quote_ok": true, "checked_against": "saved text B4.txt (sha256 e9b2e293b13212b1…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F37": { "source": "B4", "origin": "card", "locus": "licence-category comparison", "quote": "Table 3 also shows a full 45% of NC/A-O datasets are synthetic, compared to <14% in more permissive licence categories.", "claim": "45% of non-commercial/academic-only datasets synthetic, against under 14% in permissive categories", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "audit findings: the commercial divide, the non-commercial spike, synthetic data", "date": "2024-08-30 (published)", "quote_ok": true, "checked_against": "saved text B4.txt (sha256 e9b2e293b13212b1…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F38": { "source": "B4", "origin": "card", "locus": "'Language representation is heavily skewed …'", "quote": "we find a stark Western-centric skew in representation.", "claim": "a stark Western-centric skew", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "audit findings: the language and Western skew", "date": "2024-08-30 (published)", "quote_ok": true, "checked_against": "saved text B4.txt (sha256 e9b2e293b13212b1…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F39": { "source": "B4", "origin": "card", "locus": "Methods, 'Licence annotation process'", "quote": "Note that this protocol reflects best efforts to verify self-reported licences and does not constitute legal advice.", "claim": "stated limit: best efforts on self-reported licences; not legal advice", "modality": "self-description", "kind": "limit", "kernel": true, "const": false, "lineage": "stated limits: not legal advice; self-reported licences", "date": "2024-08-30 (published)", "quote_ok": true, "checked_against": "saved text B4.txt (sha256 e9b2e293b13212b1…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F40": { "source": "B5", "origin": "own_page", "locus": "README ¶1", "quote": "The Data Provenance Initiative is a multi-disciplinary volunteer effort to improve transparency, documentation, and responsible use of training datasets for AI.", "claim": "the Initiative's self-definition: a multi-disciplinary volunteer effort for transparency, documentation and responsible use of training datasets", "modality": "stipulation", "kind": "definition", "kernel": true, "const": true, "lineage": "what it is: a volunteer collective auditing AI data", "date": "undated (the repository README, read by the census 2026-10-29)", "quote_ok": true, "checked_against": "saved text B5.txt (sha256 db4c35a5f4f2b17a…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F41": { "source": "B5", "origin": "own_page", "locus": "README ¶1", "quote": "Through a large scale audit of 44 data collections, spanning 1800+ finetuning text-to-text datasets, referred to as the Data Provenance Collection, this initiative's first release thoroughly documents their web and machine sources, licenses, creators, and other metadata.", "claim": "the first release: the Data Provenance Collection, 44 collections, 1800+ fine-tuning datasets", "modality": "documented", "kind": "fact", "kernel": true, "const": false, "lineage": "the founding audit: 44 collections, 1,800+ text datasets (2023)", "date": "undated (the repository README, read by the census 2026-10-29)", "quote_ok": true, "checked_against": "saved text B5.txt (sha256 db4c35a5f4f2b17a…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F42": { "source": "B5", "origin": "own_page", "locus": "README ¶1", "quote": "Generating a subset of data from this repository will create a `Data Provenance Card`---symbolic attribution for all constituent datasets, that can be used as structured documentation.", "claim": "the Data Provenance Card: symbolic attribution for a chosen subset", "modality": "stipulation", "kind": "function", "kernel": true, "const": false, "lineage": "the Explorer, the Collection and the Card", "date": "undated (the repository README, read by the census 2026-10-29)", "quote_ok": true, "checked_against": "saved text B5.txt (sha256 db4c35a5f4f2b17a…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F43": { "source": "B5", "origin": "own_page", "locus": "'Legal Notice and Limitations'", "quote": "The information provided on this page and any output of the Data Provenance Initiative does not, and is not intended to, constitute legal advice; instead, all information, content, and materials are for general informational purposes only.", "claim": "stated limit: no output constitutes legal advice", "modality": "self-description", "kind": "limit", "kernel": true, "const": false, "lineage": "stated limits: not legal advice; self-reported licences", "date": "undated (the repository README, read by the census 2026-10-29)", "quote_ok": true, "checked_against": "saved text B5.txt (sha256 db4c35a5f4f2b17a…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F44": { "source": "B6", "origin": "wikipedia", "locus": "Wikipedia: title Data_Provenance_Initiative", "quote": "raw wikitext (action=raw, redirect=no): HTTP 404 … Wikipedia does not have an article with this exact name. … There were no results matching the query.", "claim": "Wikipedia has no article 'Data Provenance Initiative' and no article text containing the phrase", "modality": "documented", "kind": "fact", "kernel": false, "const": false, "lineage": "reception: press, the encyclopedia, the composition layer", "date": "2026-10-29 (as fetched)", "quote_ok": true, "checked_against": "saved text B6.txt (sha256 ff30d99dd38cff9d…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F45": { "source": "B7", "origin": "own_page", "locus": "/about (xh): ¶1", "quote": "We are a volunteer collective of AI researchers conducting large-scale audits of the systems and resources that shape modern AI.", "claim": "the About page's self-definition: a volunteer collective auditing the systems and resources of modern AI", "modality": "stipulation", "kind": "definition", "kernel": true, "const": true, "lineage": "what it is: a volunteer collective auditing AI data", "date": "2026-10-29 (page undated; as fetched)", "quote_ok": true, "checked_against": "saved text B7.txt (sha256 ede0671f9030be4d…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F46": { "source": "B7", "origin": "own_page", "locus": "/about (xh): ¶1", "quote": "Our work began with training data provenance: tracing datasets from origin to creation, cataloging sources, licenses, creators, and documentation practices. It now also covers web consent, open model markets, concentration of power, and how people use AI systems in practice.", "claim": "its genealogy: begun with training data provenance, now also web consent, open model markets, concentration of power and real use", "modality": "self-description", "kind": "genealogy", "kernel": true, "const": true, "lineage": "its extent now: public measurement infrastructure for AI data, models and use", "date": "2026-10-29 (page undated; as fetched)", "quote_ok": true, "checked_against": "saved text B7.txt (sha256 ede0671f9030be4d…), whitespace-normalised; parts joined by ' … ' each checked, in order" }, "F47": { "source": "B7", "origin": "own_page", "locus": "/about: note", "quote": "We collect self-reported licenses from the papers and repositories that released these datasets and categorize them according to our best efforts as a volunteer research and transparency initiative. Information provided by our work and outputs of the Data Provenance Initiative does not constitute legal advice", "claim": "stated limit: self-reported licences categorized by best efforts; not legal advice", "modality": "self-description", "kind": "limit", "kernel": false, "const": false, "lineage": "stated limits: not legal advice; self-reported licences", "date": "2026-10-29 (page undated; as fetched)", "quote_ok": true, "checked_against": "saved text B7.txt (sha256 ede0671f9030be4d…), whitespace-normalised; parts joined by ' … ' each checked, in order" } }, "field_sources": { "B1": "Data Provenance Initiative — Data Provenance Initiative (home)", "B2": "arXiv — [2310.16787] The Data Provenance Initiative: A Large Scale Audit of Dataset Licensing & Attribution in AI", "B3": "MIT Sloan — Bringing transparency to the data used to train artificial intelligence", "B4": "Nature Machine Intelligence — A large-scale audit of dataset licensing and attribution in AI", "B5": "Data Provenance Initiative — Data-Provenance-Collection README", "B6": "Wikipedia — Data Provenance Initiative (no article)", "B7": "Data Provenance Initiative — About" }, "delta": { "T_vs_LB": "At 'Data Provenance Initiative' (T1, four cards: B1–B4; 190 words) the composition gives the Initiative by its home page, its paper, MIT Sloan and the journal paper. Composed from its cards: the four systems (B1 F3–F7: the Data Provenance Explorer, the Web Consent Explorer, Open Model Evolution, the AI Observatory), at coarser grain (F5's 14,000 domains as 'thousands'; F6's and F7's counts dropped); the founding audit's licence finding at the source's grain (B2 F20 'license omission of 70%+ and error rates of 50%+', restated by B3 F24); the commercial divide (B2 F19, 'more creative tasks' and 'lower resource languages' kept, 'richer topic variety' as 'specialized topics'); the Explorer's 'interactive' and its lineage fields (B2 F22, F18). Substituted reading, the operator: the cards give the Initiative's own home page and paper, whose self-description the composition does not take (B1 F1–F2, 'Public measurement infrastructure', 'maps the datasets, web policies, model markets'), and the own pages not surfaced call it a 'volunteer effort' and a 'volunteer collective' (B5 F40, B7 F45); T composes 'a multi-institutional research project—led in part by researchers from MIT' from MIT Sloan's 'researchers, including professor Sandy Pentland and others from MIT' (B3 F23), as the Capture Registry finds (CAP-1-03); 'multi-institutional' is in no card's text (the field has 'multidisciplinary', F23, and 'multi-disciplinary', F17). Substituted count: the Explorer is given at the founding audit's scale, 'over 1,800+ text and fine-tuning datasets across dozens of collections' (B2 F17; B4 F31's 44 collections at the coarser 'dozens'), where card 1 gives the Explorer '4,000+ datasets' in '608 languages across text, speech, and video' (B1 F4, F8): the first release stands for the present explorer. Substituted reading, the tool: T has the Explorer 'audit' datasets; in the field the experts audit (F17) and the audit is released 'with an interactive UI, the Data Provenance Explorer' (F22). Force raised: T1-8 composes a cause, 'frequently driven by synthetic data generated via commercial APIs', where the journal reports the 2023 spike (B4 F36) and the synthetic share of non-commercial datasets (F37) side by side; limit dropped: the year (2023) and the rate (61%). Available, not composed: the term data provenance (B4 F30), attribution and share-alike requested (F35), the per-platform measure and 'licence laundering' (F32, F33), the Western skew (F38, F27), the stake (F26), the later audits by name (B1 F11, F12, F13), the home page's own findings (F14), the metrics strip (F8), the authors and dates (F16, F29; CAP-1-04), the Card (F25, F34), the coverage (F10). Limit dropped: B4 F39 ('best efforts … does not constitute legal advice'), on a card; the own pages' F43 and F47 were not surfaced; the Capture Registry records the census's negative tags as 'not composed' and not contradicted (CAP-1-08). Composed without attribution: no person named, MIT named (CAP-1-13), at the coarse grain of an institutional description: recorded; the findings at the source's grain carry card markers and no name: recorded, the marker the attribution the layer gives. Content replaced by use: the closer (T1-9).", "T_vs_LB_reading": "v0.9 §0.5: the first diff is ontological and small. T and the field agree on the entity (an initiative auditing AI training data, with four platforms and audit findings); T gives its operator from the press card rather than from its own pages, its explorer at the first release's scale, and a cause the journal does not state. The field's term (data provenance) and its stated limit are left out.", "LB_vs_LBA": "Admission adds, in one claim space with the field's: the provenance census's formalization of the entity (PE-01–PE-22), its organizing question, its kind ('audit collection and explorer'), capability ('records'), object, unit and carrier, its count at the first release with its own unchecked items, its negative tags ('NOT legal advice', 'describing a collection does not establish which model used it', 'initiative, explorer and collection are distinct records') held beside its sibling relation 'the same entity' (PE-18, PE-19, P_coh within the record); the Initiative's place in the licensing dimension of provenance, beside artifact and meaning, the dimensions independent (O725-03, O725-05, O725-15, O725-06); its bound at the corpus and the input (D725-01, D725-04, D725-08–D725-10, O725-11) and the path from it to semantic provenance (R725-12, R725-13, O725-07), held beside #1003's grouping of it with studies of 'attribution loss in AI outputs' (D1003-01); 'Documentation project, not infrastructure', L4, and the archive's own adoption of its provenance schema (D98-01, D98-02, D98-05, O98-03, R98-04); the requirement that provenance survive the training pipeline (D1714-01–04, O95-01, O623-01); #725's prediction that broad queries would compose from DPI papers (D725-14); and the Capture Registry's account of the composition at the name (CAP-1-01–13), including that the composition is ahead of the record on the entity's extent (CAP-1-10) and its ascription of the 70%/50% quote to the README, marked contradicts_primary (CAP-1-01). Not surfaced: every archive source; no card cites the census record or a deposit (CAP-1-12).", "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" ], "T1_composed": [ "F3", "F4", "F5", "F6", "F7", "F17", "F18", "F19", "F20", "F22", "F23", "F24", "F31", "F36", "F37" ], "T1_outside_cards": [ "'multi-institutional' (T1-1): in no field source's text", "'frequently driven by' (T1-8): a causal relation no field source states" ], "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": "`D725-04`", "source": "#725", "M_src": "stipulation", "sense": "its bound: the corpus, not the output", "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": "T composes the Initiative's audits and gives no bound", "resolver": "a later capture at 'Data Provenance Initiative' (ADDR-c0bda0b5919d) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K2", "claim": "`D725-10`", "source": "#725", "M_src": "stipulation", "sense": "its layer: licensing, necessary but on the corpus", "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": "T gives no dimension of provenance", "resolver": "a later capture at 'Data Provenance Initiative' (ADDR-c0bda0b5919d) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K3", "claim": "`O725-05`", "source": "#725", "M_src": "stipulation", "sense": "the licensing dimension's object and question", "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": "absent from T", "resolver": "a later capture at 'Data Provenance Initiative' (ADDR-c0bda0b5919d) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K4", "claim": "`D725-08`", "source": "#725", "M_src": "stipulation", "sense": "licensing audits on the input, not the output", "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": "absent from T", "resolver": "a later capture at 'Data Provenance Initiative' (ADDR-c0bda0b5919d) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K5", "claim": "`D1003-01`", "source": "#1003", "M_src": "stipulation", "sense": "held beside: studies of attribution loss in outputs", "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": "absent from T", "resolver": "a later capture at 'Data Provenance Initiative' (ADDR-c0bda0b5919d) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K6", "claim": "`D98-01`", "source": "#98", "M_src": "stipulation", "sense": "a documentation project, not infrastructure", "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": "T: 'a multi-institutional research project'", "resolver": "a later capture at 'Data Provenance Initiative' (ADDR-c0bda0b5919d) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K7", "claim": "`R98-04`", "source": "#98", "M_src": "documented", "sense": "the archive's adoption of its provenance schema", "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": "absent from T", "resolver": "a later capture at 'Data Provenance Initiative' (ADDR-c0bda0b5919d) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K8", "claim": "`D1714-02`", "source": "#1714", "M_src": "stipulation", "sense": "provenance has to survive the training pipeline", "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": "absent from T", "resolver": "a later capture at 'Data Provenance Initiative' (ADDR-c0bda0b5919d) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K9", "claim": "`PE-04`", "source": "provenance census record, 2026-10-29", "M_src": "stipulation", "sense": "the census's kind: an audit collection and explorer", "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": "absent from T", "resolver": "a later capture at 'Data Provenance Initiative' (ADDR-c0bda0b5919d) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K10", "claim": "`PE-06`", "source": "provenance census record, 2026-10-29", "M_src": "stipulation", "sense": "what receives provenance: a fine-tuning dataset traced to its sources", "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": "T: 'sources, lineages, and license types' (coarse)", "resolver": "a later capture at 'Data Provenance Initiative' (ADDR-c0bda0b5919d) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K11", "claim": "`PE-17`", "source": "provenance census record, 2026-10-29", "M_src": "stipulation", "sense": "a collection does not establish which model used it", "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": "absent from T", "resolver": "a later capture at 'Data Provenance Initiative' (ADDR-c0bda0b5919d) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K12", "claim": "`PE-18`", "source": "provenance census record, 2026-10-29", "M_src": "stipulation", "sense": "initiative, explorer and collection are distinct records", "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": "T composes the explorer as one of the Initiative's tools", "resolver": "a later capture at 'Data Provenance Initiative' (ADDR-c0bda0b5919d) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K13", "claim": "`CAP-1-03`", "source": "Capture Registry, 2026-10-29", "M_src": "documented", "sense": "the operator from the Initiative's own description", "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": "T: 'led in part by researchers from MIT'", "resolver": "a later capture at 'Data Provenance Initiative' (ADDR-c0bda0b5919d) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K14", "claim": "`CAP-1-07`", "source": "Capture Registry, 2026-10-29", "M_src": "stipulation", "sense": "the count corrected from the home page", "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": "T: the first release's 1,800+ for the Explorer", "resolver": "a later capture at 'Data Provenance Initiative' (ADDR-c0bda0b5919d) on Google AI Overview, seated in the Capture Registry (§8.3)" }, { "K": "K15", "claim": "`D725-14`", "source": "#725", "M_src": "hypothesis", "sense": "prediction: broad queries composed from DPI papers", "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": "an observation at broad provenance queries, not at this address", "resolver": "later captures at broad provenance queries (e.g. 'what is provenance in the age of AI', #725's Test 8), seated in the Capture Registry (§8.3)" } ], "kernel_T": [ { "claim": "F1", "note": "the Initiative's self-description on its card, absent" }, { "claim": "F8", "note": "the metrics strip (4,000+ datasets audited), absent; the first release's count stands" }, { "claim": "F30", "note": "data provenance defined, absent" }, { "claim": "F35", "note": "attribution and share-alike requested, absent" }, { "claim": "F38", "note": "the Western-centric skew, absent" }, { "claim": "F39", "note": "the stated limit on a card (best efforts; not legal advice), absent" }, { "claim": "F11", "note": "Consent in Crisis by name, absent (its explorer composed)" }, { "claim": "F12", "note": "the multimodal audit, absent" }, { "claim": "F16", "note": "the authors, absent" }, { "claim": "F29", "note": "the journal date, absent" }, { "claim": "F36", "note": "the year and rate of the spike, absent (the spike composed)" } ], "ledger": { "archive": "datasets/negative-of-the-negative/v2/ledgers/data-provenance-initiative/ledger-archive.json", "field": "datasets/negative-of-the-negative/v2/ledgers/data-provenance-initiative/field.json", "selection": "datasets/negative-of-the-negative/v2/traversal/data-provenance-initiative/", "config": "datasets/negative-of-the-negative/v2/panel/configs/data-provenance-initiative.json", "intake": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/", "audit": null }, "ontology": { "composition": "equal", "composition_note": "equal-terms case (v0.9 §0.9, §4.1): one claim space; the field's and the archive's claims ordered together by the entity's dependency and by genealogy within each sense; every field claim composed as received (the translation table records each as untranslated beside the archive claims it stands with)", "case": { "value": "equal", "statement": "The archive's account extends the field's ontology. The field gives the Initiative as its own pages and papers give it (a volunteer collective, its audits, its tools, its findings, its limits) and the census formalizes that same entity, every field with a quote from a saved copy of the entity's own pages; #725 names its relation to the Initiative as extension, not substitution or critique, and supplies the dimension the Initiative was not designed for without contesting what it is. No ontology here is available only from the archive, so the law of the suppressed witness does not apply (§0.10): one claim space, nothing translated (§4.1, §4.3).", "loci": [ { "dep": 725, "locus": "header, Relation", "quote": "Extension and completion, not substitution or critique" }, { "dep": 725, "locus": "header, Governing Doctrine", "quote": "The aim is to extend the existing frameworks by naming the dimension they were not designed to address." }, { "dep": 725, "locus": "§1.4 ¶4", "quote": "The packet's claim is not that the existing frameworks fail. It is that they were not designed for the dimension proposed here." }, { "dep": "PE", "locus": "census.json organizing_question", "quote": "What provenance does this system carry, through which operation, on whose assertion, with what evidence of survival?" } ] }, "E_F": { "type": "C (public entity: a research initiative, a volunteer collective of AI researchers, with its audit collection, explorer and card, now four public measurement systems)", "definition": "the Data Provenance Initiative, a volunteer collective of AI researchers (its words) whose first release audited 44 fine-tuning collections and 1,800+ text datasets (arXiv:2310.16787, 25 October 2023; Nature Machine Intelligence, 30 August 2024), defining data provenance as a dataset's sourcing, creation and licensing heritage, released as the Data Provenance Collection, Explorer and Card, and which now runs four public measurement systems (datasets, web consent, open models, real-world use); its findings licence omission of 70%+ and error of 50%+ on hosting sites, a commercial divide and a Western skew; its outputs not legal advice", "claims": [ "F40", "F45", "F46", "F1", "F2", "F3", "F17", "F31", "F16", "F29", "F30", "F18", "F22", "F42", "F4", "F8", "F20", "F19", "F38", "F43", "F39" ] }, "E_A": { "definition": "the same initiative in the provenance census record and the archive's deposits: an audit collection and explorer that records provenance, its object a fine-tuning dataset traced through its collection to its sources, its carrier audit metadata and the Data Provenance Card, its count the first release's; the corpus (licensing) dimension of provenance beside the artifact and the semantic, bounded at the input of AI systems (it 'tells you whether the dataset was licensed'); a documentation project, not infrastructure, from which the archive took its provenance schema; and the training-pipeline end of the requirement that provenance survive the pipeline", "claims": [ "PE-04", "PE-05", "PE-06", "PE-08", "PE-10", "PE-22", "O725-03", "O725-05", "D725-01", "D725-04", "D725-09", "D725-10", "D98-01", "R98-04", "D1714-02", "O95-01", "CAP-1-10" ], "defining_papers": [ 725, 98, 1714, 1003, 95, 623 ], "defining_records": [ "provenance census record (census.json rows[slug == 'data-provenance-initiative'], organizing_question, schema_note; rendered aggregators/data-provenance-initiative/index.html; editor Lee Sharks, compiled 2026-10-29)", "Capture Registry data-provenance-initiative-aio-20261029 (OBS-118fbafa73c9)" ] }, "rho": { "value": "extends", "status": "read from the provenance census record, the capture record and #725, the deposit that defines the archive's bearing on the Initiative (§1.2a)", "statement": "the archive reads E_F as the entity and adds senses, mechanisms and limits on the field's own object: the census's formalization (its question, kind, capability, object, unit, carrier, negative tags); the licensing dimension of provenance among three independent dimensions, with its bound at the corpus and the input; the Initiative as a documentation project at L4 from which the archive adopted its provenance schema; and the requirement that provenance survive the training pipeline. Nothing in E_A denies a field claim; #725 states its relation to the frameworks it names as 'Extension and completion, not substitution or critique'.", "claims": [ "PE-22", "PE-04", "PE-06", "O725-03", "O725-05", "O725-06", "D725-01", "D725-04", "D725-08", "D725-10", "D98-01", "R98-04", "D1714-02", "D1003-01" ], "guard": "ρ = extends claims an addition and no more: #725's bound ('audits whether datasets were licensed, not whether synthesized outputs preserve attribution') is a limit on the Initiative's scope, stated as not a critique of its adequacy in its own domain ('The packet's claim is not that the existing frameworks fail'); #1003's grouping of it with studies of attribution loss in outputs is composed beside #725's under P_coh, within the one ontology; the census's two statements of the explorer relation are composed together, neither ruled.", "considered": [ { "value": "constructs", "set_aside": "the archive does not read the Initiative or its audit as made by receiving texts" }, { "value": "revises", "set_aside": "the Initiative is not the archive's own entity; the census restates no earlier archive statement of it" } ] }, "translation": [ { "lineage": "what it is: a volunteer collective auditing AI data", "field_claims": [ "F9", "F23", "F40", "F45" ], "received": "F9 (B1, home (Cr): 'Public-interest infrastructure'): \"DPI releases papers, datasets, and dashboards that help researchers, journalists, policymakers, and builders reason from shared evidence.\"; F23 (B3, article body ¶4): \"A team of multidisciplinary researchers, including professor Sandy Pentland and others from MIT, created the Data Provenance Initiative to tackle the data transparency challenge head-on.\"; F40 (B5, README ¶1): \"The Data Provenance Initiative is a multi-disciplinary volunteer effort to improve transparency, documentation, and responsible use of training datasets for AI.\"; F45 (B7, /about (xh): ¶1): \"We are a volunteer collective of AI researchers conducting large-scale audits of the systems and resources that shape modern AI.\"", "receiving_text": "—", "translation": "untranslated: composed as received beside PE-02, PE-03, PE-04", "archive_claims": [ "PE-02", "PE-03", "PE-04" ] }, { "lineage": "its extent now: public measurement infrastructure for AI data, models and use", "field_claims": [ "F1", "F2", "F3", "F46" ], "received": "F1 (B1, served HTML: meta description): \"Public measurement infrastructure for AI data, model ecosystems, and real-world AI use.\"; F2 (B1, home (Cr): hero): \"Public evidence for accountable AI … DPI maps the datasets, web policies, model markets, and real-world AI uses that shape how AI systems are built and governed.\"; F3 (B1, home (Cr): 'Flagship systems'): \"Four public measurement systems … Together they trace AI from source data and web consent to model markets and real-world use.\"; F46 (B7, /about (xh): ¶1): \"Our work began with training data provenance: tracing datasets from origin to creation, cataloging sources, licenses, creators, and documentation practices. It now also covers web consent, open model markets, concentration of power, and how people use AI systems in practice.\"", "receiving_text": "—", "translation": "untranslated: composed as received beside CAP-1-10", "archive_claims": [ "CAP-1-10" ] }, { "lineage": "the term: data provenance, its relations and its unit", "field_claims": [ "F15", "F18", "F30" ], "received": "F15 (B1, home (Cr): 'Training Data Provenance'): \"Training Data Provenance … Dataset audits … Trace the sources, licenses, creators, languages, geographies, and derivation chains behind public AI training datasets.\"; F18 (B2, Abstract): \"We develop tools and standards to trace the lineage of these datasets, from their source, creators, series of license conditions, properties, and subsequent use.\"; F30 (B4, Introduction ¶3): \"In this work, we term the combination of these indicators, including a dataset’s sourcing, creation and licensing heritage, as well as its characteristics, the ‘data provenance’.\"", "receiving_text": "—", "translation": "untranslated: composed as received beside PE-05, PE-06, PE-07, PE-09, PE-22, O725-03, O725-05, O725-06, O725-15", "archive_claims": [ "PE-05", "PE-06", "PE-07", "PE-09", "PE-22", "O725-03", "O725-05", "O725-06", "O725-15" ] }, { "lineage": "the founding audit: 44 collections, 1,800+ text datasets (2023)", "field_claims": [ "F16", "F17", "F29", "F31", "F41" ], "received": "F16 (B2, abstract page header): \"[Submitted on 25 Oct 2023 (v1), last revised 4 Nov 2023 (this version, v3)] … The Data Provenance Initiative: A Large Scale Audit of Dataset Licensing & Attribution in AI … by Shayne Longpre and 16 other authors\"; F17 (B2, Abstract): \"we convene a multi-disciplinary effort between legal and machine learning experts to systematically audit and trace 1800+ text datasets.\"; F29 (B4, article header): \"Published: 30 August 2024 … Nature Machine Intelligence … volume 6, pages 975–987 (2024)\"; F31 (B4, Introduction ¶6): \"As a first contribution of the initiative, we audit 44 instruction or ‘alignment’ finetuning data collections composed of 1,858 individual datasets\"; F41 (B5, README ¶1): \"Through a large scale audit of 44 data collections, spanning 1800+ finetuning text-to-text datasets, referred to as the Data Provenance Collection, this initiative's first release thoroughly documents their web and machine sources, licenses, creators, and other metadata.\"", "receiving_text": "—", "translation": "untranslated: composed as received beside PE-10, PE-15", "archive_claims": [ "PE-10", "PE-15" ] }, { "lineage": "the Explorer, the Collection and the Card", "field_claims": [ "F22", "F25", "F34", "F42" ], "received": "F22 (B2, Abstract): \"we release our entire audit, with an interactive UI, the Data Provenance Explorer, which allows practitioners to trace and filter on data provenance for the most popular open source finetuning data collections\"; F25 (B3, 'two key actions', second): \"Practitioners can also use the tool to generate a human-readable data provenance card for datasets\"; F34 (B4, contributions, second): \"The Data Provenance Explorer (DPExplorer) (www.dataprovenance.org), an open-source repository for downloading, filtering and exploring data provenance and characteristics. Our tools auto-generate data provenance cards for scalable symbolic attribution and future documentation best practices.\"; F42 (B5, README ¶1): \"Generating a subset of data from this repository will create a `Data Provenance Card`---symbolic attribution for all constituent datasets, that can be used as structured documentation.\"", "receiving_text": "—", "translation": "untranslated: composed as received beside PE-01, PE-08, PE-13, PE-18, PE-19", "archive_claims": [ "PE-01", "PE-08", "PE-13", "PE-18", "PE-19" ] }, { "lineage": "the four systems and their counts", "field_claims": [ "F4", "F5", "F6", "F7", "F8" ], "received": "F4 (B1, home (Cr): system card 'Data Provenance Explorer' (constant xr)): \"Data Provenance Explorer … Training data … 4,000+ datasets … 608 languages across text, speech, and video … Inspect dataset sources, licenses, creators, geographies, modalities, and derivation chains.\"; F5 (B1, home (Cr): system card 'Web Consent Explorer'): \"Web Consent Explorer … Web data … 14,000 domains … Terms, robots.txt, and crawl restrictions … Track how web consent signals are changing as AI developers compete for training data.\"; F6 (B1, home (Cr): system card 'Open Model Evolution'): \"Open Model Evolution … Model ecosystem … 851k open models … 2.2B downloads and 200+ model attributes … Monitor concentration, participation, transparency, and technical shifts in open-weight AI.\"; F7 (B1, home (Cr): system card 'The AI Observatory'): \"The AI Observatory … Real-world use … 23,158 conversations … 85,633 turns from 7 public sources … Compare how people use and misuse AI systems across sources, models, and contexts.\"; F8 (B1, home (Cr): metrics strip (constant br)): \"4,000+ … datasets audited … 14,000 … web domains tracked … 851k … open models measured … 23,158 … AI conversations annotated\"", "receiving_text": "—", "translation": "untranslated: composed as received beside CAP-1-05, CAP-1-06, CAP-1-07, CAP-1-11, PE-14, PE-20", "archive_claims": [ "CAP-1-05", "CAP-1-06", "CAP-1-07", "CAP-1-11", "PE-14", "PE-20" ] }, { "lineage": "the later audits: consent, multimodal, models, use", "field_claims": [ "F11", "F12", "F13", "F28" ], "received": "F11 (B1, home (Cr): publications, 'Consent in Crisis'): \"Consent in Crisis: The Rapid Decline of the AI Data Commons … An audit of terms, restrictions, and consent signals across the web. … NeurIPS 2024\"; F12 (B1, home (Cr): publications, the multimodal audit): \"Bridging the Data Provenance Gap Across Text, Speech, and Video … An audit of four thousand public text, speech, and video datasets. … ICLR 2025\"; F13 (B1, home (Cr): publications, the model economy and real-world use): \"The AI Observatory: A Public Measure of Real-World AI Use … NeurIPS 2026 … Economies of Open Intelligence: Tracing Power and Participation in the Model Ecosystem … Preprint\"; F28 (B3, article body, closing ¶ before the quote): \"While the group’s efforts are currently focused on the data provenance of text, the plan is to expand to other media, such as video, as well as domain-specific data\"", "receiving_text": "—", "translation": "untranslated: composed as received beside PE-20", "archive_claims": [ "PE-20" ] }, { "lineage": "audit findings: licence omission and error on hosting sites", "field_claims": [ "F20", "F21", "F24", "F32", "F33" ], "received": "F20 (B2, Abstract): \"We also observe frequent miscategorization of licenses on widely used dataset hosting sites, with license omission of 70%+ and error rates of 50%+.\"; F21 (B2, Abstract): \"This points to a crisis in misattribution and informed use of the most popular datasets driving many recent breakthroughs.\"; F24 (B3, 'two key actions', first): \"Conducted a systematic audit of more than 1,800 text datasets. The researchers traced the lineage of fine-tuning datasets from 44 of the most widely used text data collections. They found that licenses were frequently miscategorized, with error rates greater than 50% and license information omission rates of more than 70%.\"; F32 (B4, Introduction ¶5): \"Notably, we find that more than 70% of licences for popular datasets on GitHub and Hugging Face are ‘unspecified’ … our rigorous re-annotation of licences finds that 66% of analysed Hugging Face licences were in a different use category, often labelled as more permissive than the author’s original licence.\"; F33 (B4, Introduction ¶5): \"Our work highlights a crisis in licence laundering and informed usage of popular datasets\"", "receiving_text": "—", "translation": "untranslated: composed as received beside PE-11, PE-12, PE-21, D725-16, D1714-01, CAP-1-01, CAP-1-02", "archive_claims": [ "PE-11", "PE-12", "PE-21", "D725-16", "D1714-01", "CAP-1-01", "CAP-1-02" ] }, { "lineage": "audit findings: attribution and share-alike requested", "field_claims": [ "F35" ], "received": "F35 (B4, licence findings): \"In total, 85% of dataset licences request attribution, and 30% include a share-alike clause\"", "receiving_text": "—", "translation": "untranslated: composed as received beside D725-02", "archive_claims": [ "D725-02" ] }, { "lineage": "audit findings: the commercial divide, the non-commercial spike, synthetic data", "field_claims": [ "F14", "F19", "F36", "F37" ], "received": "F14 (B1, home (Cr): findings strip): \"Data transparency … Open weights now outpace truly open-source models … Dataset restrictions … Source content often carries stricter terms than dataset releases … Real-use heterogeneity … No single public conversation source generalizes across AI use\"; F19 (B2, Abstract): \"Our landscape analysis highlights the sharp divides in composition and focus of commercially open vs closed datasets, with closed datasets monopolizing important categories: lower resource languages, more creative tasks, richer topic variety, newer and more synthetic training data.\"; F36 (B4, 'In 2023 there was a spike in NC/A-O dataset licences'): \"In 2023 there was a spike in NC/A-O dataset licences … in 2023, when many of the most popular and diverse datasets were published, the NC/A-O rate is 61%.\"; F37 (B4, licence-category comparison): \"Table 3 also shows a full 45% of NC/A-O datasets are synthetic, compared to <14% in more permissive licence categories.\"", "receiving_text": "—", "translation": "untranslated: composed as received (no archive claim stands in this lineage)", "archive_claims": [] }, { "lineage": "audit findings: the language and Western skew", "field_claims": [ "F27", "F38" ], "received": "F27 (B3, 'other issues', second): \"Language used in the datasets is heavily skewed toward English and Western European languages, with sparse coverage of languages from Asian, African, and South American nations, if at all.\"; F38 (B4, 'Language representation is heavily skewed …'): \"we find a stark Western-centric skew in representation.\"", "receiving_text": "—", "translation": "untranslated: composed as received (no archive claim stands in this lineage)", "archive_claims": [] }, { "lineage": "the risks it addresses", "field_claims": [ "F26" ], "received": "F26 (B3, article body ¶2): \"Without transparency into the lineage of data used for artificial intelligence models, researchers, businesses, and other intended users may find themselves out of compliance with emerging regulations like the European Union’s AI Act or exposed to legal and copyright risks.\"", "receiving_text": "—", "translation": "untranslated: composed as received (no archive claim stands in this lineage)", "archive_claims": [] }, { "lineage": "stated limits: not legal advice; self-reported licences", "field_claims": [ "F39", "F43", "F47" ], "received": "F39 (B4, Methods, 'Licence annotation process'): \"Note that this protocol reflects best efforts to verify self-reported licences and does not constitute legal advice.\"; F43 (B5, 'Legal Notice and Limitations'): \"The information provided on this page and any output of the Data Provenance Initiative does not, and is not intended to, constitute legal advice; instead, all information, content, and materials are for general informational purposes only.\"; F47 (B7, /about: note): \"We collect self-reported licenses from the papers and repositories that released these datasets and categorize them according to our best efforts as a volunteer research and transparency initiative. Information provided by our work and outputs of the Data Provenance Initiative does not constitute legal advice\"", "receiving_text": "—", "translation": "untranslated: composed as received beside PE-16, PE-17, CAP-1-08", "archive_claims": [ "PE-16", "PE-17", "CAP-1-08" ] }, { "lineage": "reception: press, the encyclopedia, the composition layer", "field_claims": [ "F10", "F44" ], "received": "F10 (B1, home (Cr): 'Policy and governance relevance'; 'Featured coverage'): \"Our work has been cited in public AI safety and governance reporting and used to ground debates about data transparency, consent, and market concentration. … DPI findings have been covered by the Financial Times, The New York Times, The Washington Post, MIT Technology Review, IEEE Spectrum, and VentureBeat.\"; F44 (B6, Wikipedia: title Data_Provenance_Initiative): \"raw wikitext (action=raw, redirect=no): HTTP 404 … Wikipedia does not have an article with this exact name. … There were no results matching the query.\"", "receiving_text": "—", "translation": "untranslated: composed as received beside D725-14, CAP-1-03, CAP-1-04, CAP-1-09, CAP-1-12, CAP-1-13", "archive_claims": [ "D725-14", "CAP-1-03", "CAP-1-04", "CAP-1-09", "CAP-1-12", "CAP-1-13" ] } ], "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": 13, "positions": [ null, null, 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": "MFFFAAMMMAFFAMFAMFFMAAAAAMA", "field_only_head": 0, "archive_only_tail": 1, "breach": false }, "max_similarity": [ 0.361, 0.313, 0.08, 0.516, 0.255, 0.09, 0.358, 0.25, 0.2, 0.333, 0.377, 0.271, 0.24 ], "ko_sentence_1_max_similarity_to_KO_B": 0.361 }, "homonyms": { "names": [ "DPI" ], "basis": "The census row carries no disambiguation field; its negative tag 'initiative, explorer and collection are distinct records' (PE-18) keeps the Initiative, its Explorer and its Collection apart as records, while its sibling relation calls the explorer's MMRS record 'the same entity' (PE-19); both are carried. No card is a homonym. 'DPI', the Initiative's own short form (B1, B7; #725), is in the archive also #261's distinction preservation index (a metric of the specification, §6.4) and is not an alias. The general concept 'data provenance' (the field's defined term, F30; W3C PROV in #725) is a sense of the term, not of the entity.", "distinct_entities": [ "the Data Provenance Explorer (the Initiative's explorer; its own /non row, datasets/negative-of-the-negative/v2/rows/data-provenance-explorer.json, from the MMRS meta-aggregator: a neighbour, a distinct record, not edited)", "Data-Provenance-Collection (the Initiative's GitHub repository and the audit's collection)", "Open Model Evolution, the Web Consent Explorer and the AI Observatory (the Initiative's other measurement systems)", "MIT Media Lab and MIT Connection Science (institutions of some members; not the operator)", "C2PA / Content Credentials, EU AI Act provisions and W3C PROV (the other provenance frameworks #725 sets beside it)" ], "neighbours": [ { "row": "data-provenance-explorer", "path": "datasets/negative-of-the-negative/v2/rows/data-provenance-explorer.json", "relation": "the Initiative's explorer, formalized by the MMRS meta-aggregator; the census's sibling record (PE-19) and one of the three names the census gives as aka (PE-01); kept a distinct row (PE-18)" } ], "substitutes": { "T": [] }, "register": { "address_1": { "slug": "data-provenance-initiative-aio-20261029", "obs_id": "OBS-118fbafa73c9" } } }, "notes": [ "One address, Google AI Overview (expanded), signed out, incognito, 2026-10-29: data-provenance-initiative-aio-20261029 (OBS-118fbafa73c9, ADDR-c0bda0b5919d, AXN:0ACD.CAPTURE.⚙️⛵∞✋♻️🌉), four cards, 190 words in the answer body. A NEW address; the census record landed on provenanceerasure.org the same day and no card cites it (CAP-1-12).", "Wikipedia: no article. The title Data_Provenance_Initiative returns HTTP 404 for raw wikitext (redirect=no) and for the rendered page (one 429 retried), is not a redirect, and the rendered Special:Search for the quoted phrase in article space reports no results (B6).", "dataprovenance.org, fetched 2026-10-29 (B1, B7; text from the script bundle main.af19d7fc.js, the same bytes the data-provenance-explorer row saved): 'Public measurement infrastructure for AI data, model ecosystems, and real-world AI use'; 'Four public measurement systems' (Data Provenance Explorer, '4,000+ datasets', '608 languages across text, speech, and video'; Web Consent Explorer, '14,000 domains'; Open Model Evolution, '851k open models', '2.2B downloads'; The AI Observatory, '23,158 conversations', '85,633 turns from 7 public sources'); a metrics strip '4,000+ datasets audited', '14,000 web domains tracked', '851k open models measured', '23,158 AI conversations annotated'; publications through 'The AI Observatory: A Public Measure of Real-World AI Use' (NeurIPS 2026); About: 'We are a volunteer collective of AI researchers …', 'Our work began with training data provenance … It now also covers web consent, open model markets, concentration of power, and how people use AI systems in practice.' The capture's card ('4,000+ audited datasets; 14,000 tracked web domains', four platforms) checks against the page: the strip reworded (F4, F8, F3). The census's 'unchecked' status for these figures (CAP-1-07) is answered by this row's fetch as data; the census itself is not changed here.", "Defect (capture record, reported, not fixed): findings[0] says the 70%+/50%+ wording 'is the record's quote from the repository README'; the census's evidence.retention_results gives the source as data-provenance-initiative--paper (the arXiv abstract, B2 F20), and the saved README (B5) has no '70%'. Carried as CAP-1-01 with the mark contradicts_primary.", "Defect (census record, reported, not fixed): the record states the explorer relation two ways: negative_tags[2] 'initiative, explorer and collection are distinct records' and relations[0].note, of the MMRS record of the explorer, 'the same entity, read for reception there and for provenance carriage here' (PE-18, PE-19; open block). The rendered page reads 'Data Provenance Initiative is a audit collection and explorer' (article before a vowel).", "Census fields a fetched source contradicts or the capture shows out of date (reported, not fixed): count '44 data collections, 1800+ fine-tuning text-to-text datasets (first release)' against the home page's '4,000+ datasets audited' (B1 F8) and the multimodal audit (F12); lineage 'training data' and kind 'audit collection and explorer' against the About page's present scope ('It now also covers web consent, open model markets, concentration of power, and how people use AI systems in practice', F46) and the home page's four systems (F3); data_access 'the Explorer (a JavaScript application, unreadable to a plain fetch)': its pages' text is readable from the served script bundle (this row and the data-provenance-explorer row both read it); unchecked 'the multimodal extension (arXiv:2412.17847)' is now on the home page as 'Bridging the Data Provenance Gap Across Text, Speech, and Video' (ICLR 2025). The census's capability 'records' and its object and unit are not contradicted. The README's 'Full details are available in our paper (coming soon)' is stale against the paper's two versions (arXiv 2023, Nature MI 2024), a fact of the source.", "The composition's operator ('multi-institutional research project—led in part by researchers from MIT') follows MIT Sloan (CAP-1-03); the About page's contributors list (in the bundle, after offset 2100100; not ledgered) shows members at MIT and at universities and companies elsewhere, which neither confirms 'multi-institutional' as the Initiative's word nor 'led in part' as a role the field states.", "MIT Sloan (3 March 2025) has the work 'currently focused on the data provenance of text' (F28) after the multimodal audit's ICLR 2025 venue: a dated statement of the source, carried at its date.", "Card and own page: the arXiv abstract page is card 2 and the census's paper source; fetched as the card, its bytes equal the census's saved copy (sha256 d0155205…), so it is one source (B2). The README is copied from the census's saved copy (sha256 db4c35a5…), not refetched (addendum §3).", "Neighbour: datasets/negative-of-the-negative/v2/rows/data-provenance-explorer.json (the explorer, MMRS meta-aggregator). Its string pass and this row's return the same candidates; this row admits more of #725 because the defining paper names the Initiative also as 'DPI'. The two rows' field sources overlap at the home page and the arXiv paper; their ledgers are separate and neither was edited.", "Ledger unaudited (§3.8: one extractor per ledger). Field quotes 47/47 and archive quotes 63/63 verified by script." ], "composition_v09": { "source": "this row (composed directly; no rebuild/negative-of-the-negative/v09/data-provenance-initiative.json)", "composer": "Claude (Opus 5.5), subagent, 2026-10-29", "defining_material_read": [ { "what": "provenance census record, census row", "file": "/home/claude/fleet/provenance-erasure/data/provenance-census/census.json (rows[slug == 'data-provenance-initiative'], organizing_question, schema_note)", "read": "whole" }, { "what": "provenance census record, rendered", "file": "/home/claude/fleet/provenance-erasure/aggregators/data-provenance-initiative/index.html", "read": "text, for agreement with the sources" }, { "what": "provenance census, saved sources", "file": "/home/claude/fleet/provenance-erasure/data/provenance-census/sources/data-provenance-initiative--{repo-readme,paper}.{html.gz,txt}", "read": "whole" }, { "what": "capture record", "file": "data/EA-WG-CAPTURES-01.json (data-provenance-initiative-aio-20261029)", "read": "transcript, transcript_raw, cite_list, d, reading, findings, analysis, q_kind, per_note, originator" }, { "what": "neighbour row", "file": "datasets/negative-of-the-negative/v2/rows/data-provenance-explorer.json", "read": "its shape, field, ledgers, reading and notes (not edited)" }, { "dep": 725, "file": "data/texts/AXN-0273-text.md", "read": "whole" }, { "dep": 98, "file": "data/texts/AXN-0271-text.md", "read": "§D, the layer and component tables" }, { "dep": 1714, "file": "data/texts/AXN-0A7F-text.md", "read": "§9.2 and the revision note naming the Initiative" }, { "dep": 95, "file": "data/texts/AXN-026A-text.md", "read": "the typology of source-dependent claims" }, { "dep": 1003, "file": "data/texts/AXN-03F7-text.md", "read": "the paragraph naming the Initiative" }, { "dep": 623, "file": "data/texts/AXN-01DA-text.md", "read": "§8.3" }, { "what": "the specification", "file": "data/texts/AXN-0700-text.md", "read": "§0–§8 whole; Appendix A by its headings" } ], "cuts": [], "notes": "equal case; KO 27 sentences, all 47 field and 63 archive claims cited; P body 350 words; KO_B 13 sentences, none surviving at ≥ 0.95" }, "field_by_origin": { "cards": [ "B1", "B2", "B3", "B4" ], "own_pages": [ "B5", "B7" ], "census_third_party": [], "wikipedia": [ "B6" ], "unreachable": [], "title_only": [], "script_rendered": [ "B1", "B7" ], "wikipedia_state": "no article: the title Data_Provenance_Initiative returns HTTP 404 for raw wikitext (redirect=no) and for the rendered page ('Wikipedia does not have an article with this exact name'), it is not a redirect, and the rendered Special:Search for the quoted phrase in article space reports 'There were no results matching the query.' (2026-10-29)", "census_sources": { "data-provenance-initiative--repo-readme": "B5 (copied from the census's saved copy)", "data-provenance-initiative--paper": "B2 (fetched as card 2; bytes identical to the census's copy)" }, "census_third_party_note": "the census record reads no third-party source for this entity (both sources basis 'operator'); the origin census_third_party is empty", "sources": [ { "id": "B1", "origin": "card", "role": "Overview card 1; also the census record's url (https://www.dataprovenance.org/)", "card": "Data Provenance Initiative — Data Provenance Initiative", "url": "https://www.dataprovenance.org/", "kind": "Overview card (own page)", "surfaced_by_T": [ { "slug": "data-provenance-initiative-aio-20261029", "card": 1 } ], "files": { "B1.html": { "path": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B1.html", "sha256": "bf609f385257c3a179dc564d68d8ec46b2f096a34f61918cc7d750b83f6d7e2f" }, "B1.txt": { "path": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B1.txt", "sha256": "86ba9a3ecee7633459623ad0b59674ab5ccb5a0035dc912a8bfa1d8d934c8b92" }, "site-main.af19d7fc.js": { "path": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/site-main.af19d7fc.js", "sha256": "91de539c0ec5317faec490142648792d586b0ee21e82a88f5d7feba90a0898e9" } }, "fetched": { "at": "2026-10-29T17:06:17Z", "http": 200, "url": "https://www.dataprovenance.org/", "raw": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B1.html", "raw_sha256": "bf609f385257c3a179dc564d68d8ec46b2f096a34f61918cc7d750b83f6d7e2f", "text": [ "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B1.txt" ], "bundle": { "url": "https://www.dataprovenance.org/static/js/main.af19d7fc.js", "at": "2026-10-29T17:06:31Z", "http": 200, "sha256": "91de539c0ec5317faec490142648792d586b0ee21e82a88f5d7feba90a0898e9" } }, "card_title": "Data Provenance Initiative", "card_snippet": "Platforms: Data Provenance Initiative (maps AI datasets, web policies, model markets, real-world use). Metrics: 4,000+ audited datasets; 14,000 tracked web doma...", "card_note": null, "names_operator": "yes ('Data Provenance Initiative', 'DPI' in the saved text)", "names_archive": "no (none of the Crimson Hexagonal Archive, Alexanarch, Lee Sharks, provenanceerasure, EA-WG-CAPTURES-01 in the saved text, card title and snippet)", "note": "The served HTML is the site's script shell (title, meta description, empty root; 1,601 bytes, sha256 bf609f38…, the same shell the data-provenance-explorer row saved at every path). The page text is the string literals of Cr, the / route's component, with the constants it maps, in /static/js/main.af19d7fc.js (fetched verbatim, saved as site-main.af19d7fc.js, sha256 91de539c…, identical to the bundle the data-provenance-explorer row saved), offsets 1862261–1876490, extracted in source order (CSS-like literals and bare years dropped). What dataprovenance.org says (the task's question): four public measurement systems (Data Provenance Explorer, Web Consent Explorer, Open Model Evolution, The AI Observatory) and a metrics strip '4,000+ / datasets audited', '14,000 / web domains tracked', '851k / open models measured', '23,158 / AI conversations annotated' (constant br). The card's 'Metrics: 4,000+ audited datasets; 14,000 tracked web domains' is that strip reworded; its 'Platforms: … (maps AI datasets, web policies, model markets, real-world use)' is the hero line F2." }, { "id": "B2", "origin": "card", "role": "Overview card 2; also the census's own-page source data-provenance-initiative--paper (same URL, same sha256)", "card": "arXiv — [2310.16787] The Data Provenance Initiative: A Large Scale Audit of Dataset Licensing & Attribution in AI", "url": "https://arxiv.org/abs/2310.16787", "kind": "Overview card (the operator's paper)", "surfaced_by_T": [ { "slug": "data-provenance-initiative-aio-20261029", "card": 2 } ], "files": { "B2.html": { "path": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B2.html", "sha256": "d01552052066b391aeed8b50adde61897cb194ea14727553cbf65971af51bb53" }, "B2.txt": { "path": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B2.txt", "sha256": "0063d0167cbd2144f01306ebc2139a2447a859c49a653dbc65c58f2f13c6a6a8" } }, "fetched": { "at": "2026-10-29T17:06:19Z", "http": 200, "url": "https://arxiv.org/abs/2310.16787", "raw": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B2.html", "raw_sha256": "d01552052066b391aeed8b50adde61897cb194ea14727553cbf65971af51bb53", "text": [ "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B2.txt" ], "census_copy": { "src": "data-provenance-initiative--paper", "read": "2026-10-29", "sha256": "d01552052066b391aeed8b50adde61897cb194ea14727553cbf65971af51bb53", "same_bytes": true } }, "card_title": "[2310.16787] The Data Provenance Initiative: A Large Scale Audit of Dataset Licensing & Attribution in AI", "card_snippet": "Audit: 1800+ text datasets traced and audited by legal and machine learning experts to evaluate licensing, lineage, and attribution in AI training. Findings: Co...", "card_note": null, "names_operator": "yes (title; 'Shayne Longpre and 16 other authors')", "names_archive": "no", "note": "Fetched as the card; the bytes are identical to the census's saved copy (sha256 d0155205…), so the card and the census's own page are one source, carried once, as a card." }, { "id": "B3", "origin": "card", "role": "Overview card 3", "card": "MIT Sloan — Bringing transparency to the data used to train artificial intelligence | MIT Sloan", "url": "https://mitsloan.mit.edu/ideas-made-to-matter/bringing-transparency-to-data-used-to-train-artificial-intelligence", "kind": "Overview card (third-party article; Sara Brown, 2025-03-03)", "surfaced_by_T": [ { "slug": "data-provenance-initiative-aio-20261029", "card": 3 } ], "files": { "B3.html": { "path": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B3.html", "sha256": "563661c73ddce481dd825a948f5c3afa657f017607a21a2e603427ab6ed1996f" }, "B3.txt": { "path": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B3.txt", "sha256": "da84c05f68986b58e3fc26343b53743db0b5bcea2ad1b2b517aeda710a4fb6ea" } }, "fetched": { "at": "2026-10-29T17:08:07Z", "http": 200, "url": "https://mitsloan.mit.edu/node/49214", "effective_url": "https://mitsloan.mit.edu/ideas-made-to-matter/bringing-transparency-to-data-used-to-train-artificial-intelligence", "raw": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B3.html", "raw_sha256": "563661c73ddce481dd825a948f5c3afa657f017607a21a2e603427ab6ed1996f", "text": [ "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B3.txt" ], "found_by": "WebSearch on the card's title and site (mitsloan.mit.edu/node/49214, which redirects to the article; the page's is the card's title)" }, "card_title": "Bringing transparency to the data used to train artificial intelligence | MIT Sloan", "card_snippet": "Problem: Untraced AI training datasets cause legal, copyright, regulatory, bias, and quality risks due to poor documentation and high license error rates (>50%)", "card_note": null, "names_operator": "yes ('the Data Provenance Initiative'; Pentland, Mahari, Longpre)", "names_archive": "no", "note": "MIT Sloan, Ideas Made to Matter, by Sara Brown, datePublished 2025-03-03 (the page's JSON-LD). The card's snippet ('Problem: Untraced AI training datasets cause legal, copyright, regulatory, bias, and quality risks due to poor documentation and high license error rates (>50%)') is a summary of ¶2 and the first 'key action', not a sentence of the page." }, { "id": "B4", "origin": "card", "role": "Overview card 4", "card": "Nature — A large-scale audit of dataset licensing and attribution in AI | Nature Machine Intelligence", "url": "https://www.nature.com/articles/s42256-024-00878-8", "kind": "Overview card (the operator's journal paper, open access)", "surfaced_by_T": [ { "slug": "data-provenance-initiative-aio-20261029", "card": 4 } ], "files": { "B4.html": { "path": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B4.html", "sha256": "766eb441fb2e78dbad63ca8676edc561bf24da0d7157ce18fda4df0234dc1220" }, "B4.txt": { "path": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B4.txt", "sha256": "e9b2e293b13212b16dfadf85aff1892f0d264c909576c62cd064d40df1d9e4d1" } }, "fetched": { "at": "2026-10-29T17:08:12Z", "http": 200, "url": "https://www.nature.com/articles/s42256-024-00878-8", "effective_url": "https://www.nature.com/articles/s42256-024-00878-8?error=cookies_not_supported&code=fbf2fb11-d6b3-4433-8b88-c13b3c109760", "raw": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B4.html", "raw_sha256": "766eb441fb2e78dbad63ca8676edc561bf24da0d7157ce18fda4df0234dc1220", "text": [ "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B4.txt" ], "found_by": "WebSearch on the card's title (DOI 10.1038/s42256-024-00878-8 via OpenAlex); the page's <title> is the card's title" }, "card_title": "A large-scale audit of dataset licensing and attribution in AI | Nature Machine Intelligence", "card_snippet": "Audit: 1,858 text/finetuning datasets across 44 collections evaluated for licensing, attribution, and lineage via the Data Provenance Initiative and DPExplorer.", "card_note": null, "names_operator": "yes (the paper's 'data provenance initiative'; authors Longpre … Hooker)", "names_archive": "no", "note": "Nature Machine Intelligence 6, 975–987, published 30 August 2024 (received 15 February 2024); full text served." }, { "id": "B5", "origin": "own_page", "role": "the census's own-page source data-provenance-initiative--repo-readme (basis operator); not surfaced by T", "card": "Data Provenance Initiative (GitHub) — Data-Provenance-Collection README.md", "url": "https://raw.githubusercontent.com/Data-Provenance-Initiative/Data-Provenance-Collection/main/README.md", "kind": "own page (repository README, markdown)", "surfaced_by_T": "no (not surfaced by T; admitted beside the cards by the operator's ruling of 2026-10-29)", "files": { "B5.html": { "path": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B5.html", "sha256": "db4c35a5f4f2b17a14680fc1c6854a6f36f82a930c6dfffa4f0297b5ea7dd4f1" }, "B5.txt": { "path": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B5.txt", "sha256": "db4c35a5f4f2b17a14680fc1c6854a6f36f82a930c6dfffa4f0297b5ea7dd4f1" } }, "fetched": { "census_copy": true, "src": "data-provenance-initiative--repo-readme", "read": "2026-10-29", "http": "200", "sha256": "db4c35a5f4f2b17a14680fc1c6854a6f36f82a930c6dfffa4f0297b5ea7dd4f1", "raw": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B5.html", "raw_sha256": "db4c35a5f4f2b17a14680fc1c6854a6f36f82a930c6dfffa4f0297b5ea7dd4f1", "text": [ "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B5.txt" ], "note": "copied from /home/claude/fleet/provenance-erasure/data/provenance-census/sources/data-provenance-initiative--repo-readme.html.gz (gunzipped; sha256 matches the census's) and .txt; not refetched (addendum §3)" }, "card_title": null, "card_snippet": null, "card_note": null, "names_operator": "yes", "names_archive": "no", "note": "The README is markdown; B5.txt is the census's text copy (identical bytes to the raw)." }, { "id": "B6", "origin": "wikipedia", "role": "the entity's Wikipedia title", "card": "Wikipedia — Data Provenance Initiative (no article)", "url": "https://en.wikipedia.org/wiki/Data_Provenance_Initiative", "kind": "Wikipedia (no article)", "surfaced_by_T": "no (not surfaced by T; checked by the operator's ruling of 2026-10-29)", "files": { "B6.wiki": { "path": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B6.wiki", "sha256": "f92e16be9fa948f544a2bc138b25dacc6eef5cb5b320497e1e4103057b07b9b2" }, "B6.html": { "path": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B6.html", "sha256": "64f4d986f0e9aa6273b7bd72742a84affdbd64b51b8df152690684f86ba8050c" }, "B6-search.html": { "path": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B6-search.html", "sha256": "43e59e395580b5d884df4155e6c24b540141e6f1019207775d5e54974dc07853" }, "B6.txt": { "path": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B6.txt", "sha256": "ff30d99dd38cff9d7e80d164ebf0768cfd6d796728641ae5d959318ac842bbd4" } }, "fetched": { "raw": { "url": "https://en.wikipedia.org/w/index.php?title=Data_Provenance_Initiative&action=raw&redirect=no", "at": "2026-10-29T17:06:20Z", "http": 404, "sha256": "f92e16be9fa948f544a2bc138b25dacc6eef5cb5b320497e1e4103057b07b9b2" }, "rendered": { "url": "https://en.wikipedia.org/wiki/Data_Provenance_Initiative", "at": "2026-10-29T17:06:38Z", "http": 404, "sha256": "64f4d986f0e9aa6273b7bd72742a84affdbd64b51b8df152690684f86ba8050c", "retries": [ { "at": "2026-10-29T17:06:21Z", "http": 429 } ] }, "search": { "url": "https://en.wikipedia.org/w/index.php?search=%22Data+Provenance+Initiative%22&title=Special:Search&profile=advanced&fulltext=1&ns0=1", "at": "2026-10-29T17:06:22Z", "http": 200, "sha256": "43e59e395580b5d884df4155e6c24b540141e6f1019207775d5e54974dc07853" }, "text": [ "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B6.txt" ] }, "card_title": null, "card_snippet": null, "card_note": null, "names_operator": "no", "names_archive": "no", "note": "no article, no redirect; the full-text search for the quoted phrase in article space returns no results (the same state the data-provenance-explorer row recorded for this title earlier the same day)" }, { "id": "B7", "origin": "own_page", "role": "the entity's About page on the census record's url host (dataprovenance.org/about); not cited by the census's sources; read under the task's instruction to fetch dataprovenance.org; not surfaced by T", "card": "Data Provenance Initiative — About", "url": "https://www.dataprovenance.org/about", "kind": "own page (script-rendered route)", "surfaced_by_T": "no (not surfaced by T)", "files": { "B7.html": { "path": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B7.html", "sha256": "bf609f385257c3a179dc564d68d8ec46b2f096a34f61918cc7d750b83f6d7e2f" }, "B7.txt": { "path": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B7.txt", "sha256": "ede0671f9030be4d6492f19fed09093fc4928dd78f734a95efeef57e70ff8e57" }, "site-main.af19d7fc.js": { "path": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/site-main.af19d7fc.js", "sha256": "91de539c0ec5317faec490142648792d586b0ee21e82a88f5d7feba90a0898e9" } }, "fetched": { "at": "2026-10-29T17:10:15Z", "http": 200, "url": "https://www.dataprovenance.org/about", "raw": "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B7.html", "raw_sha256": "bf609f385257c3a179dc564d68d8ec46b2f096a34f61918cc7d750b83f6d7e2f", "text": [ "datasets/negative-of-the-negative/v2/intake/data-provenance-initiative-20261029/B7.txt" ], "bundle": { "url": "https://www.dataprovenance.org/static/js/main.af19d7fc.js", "sha256": "91de539c0ec5317faec490142648792d586b0ee21e82a88f5d7feba90a0898e9" } }, "card_title": null, "card_snippet": null, "card_note": null, "names_operator": "yes", "names_archive": "no", "note": "The served HTML is the same shell as B1; the text is the string literals of xh, the /about route's component, and what follows it to offset 2100100, before the contributors list (offsets 2090236–2100100). The contributors list that follows (not ledgered) names some fifty people, most of them at MIT, the others at universities and companies in some twenty countries." } ] } }