AIAAIC Repository

register entry aiaaic-repository-aio-20261029 OBS-524198c4efed · file as first run sha256 7706409d1f1fd1ea…

• T1-1 definition: 'an independent, open-access public database that tracks real-world incidents, controversies, and harms related to artificial intelligence, algorithms, and automation' (markers: the name's link, token a; the sentence's [1], token b)
• T1-2 launch: 'Started in June 2019 as a private initiative to monitor transparency and accountability.' (markers [1] token c, [2] token b)
• T1-3 scope: 'Covers algorithmic discrimination, autonomous vehicle crashes, deepfakes, privacy breaches, workplace automation accidents, generative AI copyright violations, and chatbot hallucinations.' (marker token a)
• T1-4 licence: 'Free and open-source under a Creative Commons (CC BY-SA 4.0) license.' (marker token b)
• T1-5 audience: 'Used globally by researchers, journalists, civil society groups, and policymakers.' (marker token b)
• T1-6 closer: 'Would you like to search for a specific type of AI incident or learn more about the taxonomy used to classify these events?'

• B1 AIAAIC — AIAAIC - AIAAIC Repository fetched 2026-10-29T05:06:24Z with curl -sSL -m 30 -A 'Mozilla/5.0'; HTTP 200; raw B1.html sha256 92368c7f90250b41b9947eebc37238390c35ce2f150c078526b885870fd5a3a9; text B1.txt sha256 3911da48bd4691c0540509cbc53ccc44e32b37db43937bb032134ee73c9bc97e
• B2 OECD AI Policy Observatory — AIAAIC Repository - OECD.AI fetched 2026-10-29T05:06:24Z with curl -sSL -m 30 -A 'Mozilla/5.0'; HTTP 200; raw B2.html sha256 ab766dbf5dcde467c98509f3c5573f3a8b482183ad414c834ab49f7d09df0a28; text B2.txt sha256 a9fc2b4a082027b0bd2e94f3101e6c9731aba4489dcdee62d6c4a6949f0a61f9
• B3 EthicAI — Tracking AI incidents: OECD AIM and AIAAIC Repository - EthicAI fetched 2026-10-29T05:08:08Z with curl -sSL -m 30 -A 'Mozilla/5.0'; HTTP 200; raw B3.html sha256 d349dd9210de7a5b643ea7e79d84ad42a736e85a0339e67b5f3595576a61bb49; text B3.txt sha256 7dc8cb554b90d5b6e897d27ecaaf85df13398222c7219fff22869e8449412239
• B4 AIAAIC — About AIAAIC fetched 2026-10-29T05:06:26Z with curl -sSL -m 30 -A 'Mozilla/5.0'; HTTP 200; raw B4.html sha256 c088cb1616de229fcbdc2bc327515e1751b51086b90730bed310dd34789718f3; text B4.txt sha256 af3caf81935dd0c4ea601b1f6f4cab1a4beb7844465cba3f3f8c3437cdb18ef7
• B5 AIAAIC (Substack) — About - AIAAIC fetched 2026-10-29T05:06:26Z with curl -sSL -m 30 -A 'Mozilla/5.0'; HTTP 200; raw B5.html sha256 cc219629c095ed9b0fc414f11fe5bed9b0d611e76f8eb5863e9940b39a84d801; text B5.txt sha256 90b0161b955dc1cf90e31d0dc4f0248300e3db549027a4b0b7cf8a90d4980e5a
• B6 AIAAIC — Google AI Overviews give wrong, dangerous answers fetched 2026-10-29T05:07:21Z with curl -sSL -m 30 -A 'Mozilla/5.0'; HTTP 200; raw B6.html sha256 e15e140aa91db3546a74b7e417de86278e26543ce06da5618f05879c001f3061; text B6.txt sha256 de2a1f70996b3e8962b8eb453f5f305433274493da8dfabc32af26347adf6355
• B7 Wikipedia — AIAAIC (redirect) fetched 2026-10-29T05:08:22Z with curl -sSL -m 30 -A 'Mozilla/5.0'; HTTP 200; raw B7.wiki sha256 43b1fcf94b3af38aa3230a56ef1a5a7b6c92abcccc5094dec6ef6d0c8f7e2dde; text B7.wiki sha256 43b1fcf94b3af38aa3230a56ef1a5a7b6c92abcccc5094dec6ef6d0c8f7e2dde, B7.txt sha256 6be4e450dd3387260aba48828d7bfd129359d0b213def9734c7912945bffe152, B7-history.txt sha256 46c947275f571e7c87b693dde0a87e51c36350c669c06bcfc41dbbdccd0a8543
• B8 Wikipedia — Ethics of artificial intelligence fetched 2026-10-29T05:08:34Z with curl -sSL -m 30 -A 'Mozilla/5.0'; HTTP 200; raw B8.wiki sha256 ab049f50ac4220597b057bae34c9327ccc3d1dd5690be378148cb093b851a984; text B8.txt sha256 baa144a4932864984c663f05cb4f1774dd97b53cbfc0bf74cc94b89024500fd6, B8.wiki sha256 ab049f50ac4220597b057bae34c9327ccc3d1dd5690be378148cb093b851a984

E_A, the archive's entity: the same repository, in the MMRS meta-aggregator's record: a database of incidents and controversies driven by AI, algorithmic and automation systems, its unit the incident or controversy involving such a system, operated by AIAAIC, begun June 2019 as a private, professional project, CC BY-SA 4.0; a selection wider than harm, in which a controversy enters as a controversy and admission does not require harm; the oldest of the census's three harm repositories, beside the AI Incident Database and the OECD monitor; its record kept apart from it, with who runs it, its entry fields, intake and count left unchecked; observed by the Capture Registry at its name, where the composition layer shared the record's third-party source, agreed with it, and named no person MMRS-01 MMRS-14 MMRS-12 MMRS-15 MMRS-17 MMRS-06 MMRS-07 MMRS-19 MMRS-09 MMRS-27 MMRS-23 CAP-1-04 CAP-1-03; defining papers #543

E_F, the field's entity (type C — a conventional reading holds it (public entity: a database)): an independent, grassroots, public interest collection of incidents and ethical controversies caused by and associated with AI, algorithms and automation (its own page; the OECD catalogue's 'resource' detailing incidents and controversies), kept by AIAAIC, a non-partisan grassroots initiative of individuals funded by donations; started in June 2019 as a private, professional project on AI's reputational risks; its collections incidents, systems and datasets, its entries numbered and classified by five taxonomies, open to anyone's report, its content CC BY-SA 4.0; with a weekly newsletter, AIAAIC Alert; Wikipedia's title for it a redirect to a section that does not exist (B1–B8)

ρ = extends: the archive reads E_F as the entity and adds senses and limits on the field's own object: what the selection admits (a unit wider than harm; controversies entering as controversies; three kinds of system), the record kept apart from the entity with its unchecked items, the entity's place in the harm basin and its neighbours, the newsletter's writer, and the composition layer's observed treatment of the entity MMRS-06 MMRS-07 MMRS-19 MMRS-08 MMRS-27 MMRS-10 MMRS-23 MMRS-09 MMRS-21 MMRS-18 CAP-1-04 CAP-1-03. Guard: ρ = extends claims an addition and no more: MMRS-06, MMRS-07 and MMRS-19 are the record's readings (basis 'reading'), not findings on the repository's stated criteria, which the record did not read (MMRS-10); MMRS-23–25 are limits of the record, not claims that the repository lacks an operator or a count; where an archive claim meets a field claim (MMRS-19 beside F22; MMRS-02 beside F38; CAP-1-01 against F12) both are composed under P_coh in the one claim space

The case: equal terms, ontology included (v0.9). The MMRS record holds the field's entity as the entity: it describes the AIAAIC Repository from the repository's own page, the OECD's catalogue and the newsletter's about page, marking each field's basis (operator, third-party, reading), and composes nothing that denies a field presupposition. What it adds is on the field's object: what the selection admits (a unit wider than harm, controversies entering as controversies, three kinds of system), the record kept apart from the entity with its own unchecked items, the place in the harm basin beside the AI Incident Database and the OECD monitor; and the Capture Registry adds the composition layer's account at the name, found to share the record's source and to name no person. The ontology the archive uses is available in the field itself: the repository's own page and the catalogue give 'incidents and controversies' (F1, F11), and the entry on Google AI Overviews is typed 'Issue' (F47); the record's reading extends those to the admission rule, against EthicAI's harm framing (F22), composed under P_coh in one claim space. The law of the suppressed witness does not apply (§0.10: not where the field holds the ontology). Not mixed.

#MMRS article, kernel s1 (articles.py ARTICLES['aiaaic-repository']): The AIAAIC Repository records incidents and controversies driven by AI, algorithmic and automation systems.

#MMRS article, What its selection admits s1–s2: Its unit is wider than harm. A controversy enters as a controversy, so public dispute about a system is recorded alongside events in which a system caused damage.

#MMRS census.json, schema_note: each page has two nodes, the MMRS record and the entity it describes; relations are typed; each assertion carries a source and a basis (operator statement, third-party report, our observation, our reading); lifecycle is kept apart from the observation that saw it

#MMRS article, Place in the field s2: Its entry pages, intake process and current count were not read for this record.

#capture aiaaic-repository-aio-20261029, analysis s1: Composition and record were built from the same third-party catalogue and agree.

#B1 page body, header (field): The independent, grassroots public interest collection of incidents and ethical controversies caused by and associated with AI, algorithms, and automation.

equal-terms case: every field claim composed as received, beside the archive claims it stands with. Reading the received column recovers the field (§5.9).

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 0 of 10 (positions [None, None, None, None, None, None, None, None, None, None]; composed first under v0.9, so there is no v0.7 object to measure against). Opening-run check (§5.10): no breach.

On its own page the AIAAIC Repository is "the independent, grassroots public interest collection of incidents and ethical controversies caused by and associated with AI, algorithms, and automation"; the OECD catalogue (8 June 2023) describes an independent, open public interest resource detailing incidents and controversies, its card reading "open-access", and EthicAI (24 April 2025) a collective memory documenting incidents where AI systems cause or contribute to harms, the repository's card listing its incidents as harms. It began in June 2019 as a private, professional project to understand the reputational risks of AI and became an open, public interest initiative (OECD); EthicAI calls it a private initiative of 2019, now open source. Its maker is AIAAIC, "AI, Algorithmic and Automation Incidents and Controversies", a non-partisan grassroots initiative for transparency, openness and accountability, a diverse set of individuals funded by individuals' donations, with no commercial funding (about page), and the developing organisation in the OECD's catalogue. It is independent of government and industry and read by researchers, academics, teachers, NGOs, journalists, policymakers and industry experts (OECD); its weekly Alert newsletter, read in 92 countries, is by Charlie Pownall in its Substack metadata. It holds incidents, systems and datasets, entries numbered to AIAAIC2276 on 29 October 2026, classified by five taxonomies and carrying dates, locations, developers, technologies and harms searchable by attribute (repository page; EthicAI). An entry gives operator, developer, country, sector, purpose, technology and issue, links its sources and a Wikipedia gloss, and carries a page type, as the May 2024 entry on Google AI Overviews telling users to add glue to pizzas shows, typed "Issue"; AIAAIC2256 records Google's AI summaries giving cancer patients wrong advice. Anyone may report an incident or improve a page (repository and entry pages), and EthicAI finds its submissions more open and its array wider than the OECD monitor's, the AIAAIC grassroots and the two complementary. Its content is CC BY-SA 4.0 (its page; the OECD, free of charge, open source/permissive), and the catalogue calls the entry always up to date. The OECD tags it an incident database and types it a risk-management, awareness and education tool, not vetted by the OECD; EthicAI defines the AI incident broadly and names the trackers' limits, underreporting, verification and coverage gaps, calling them under the radar but critical. Its resources include an AI ethics glossary; Wikipedia redirects the title AIAAIC (since 15 September 2025) to a section of Ethics of artificial intelligence that the article does not have and that does not name it.

The AIAAIC Repository is a database of incidents and controversies driven by AI, algorithmic and automation systems (MMRS record), its unit the incident or controversy involving such a system (MMRS record), and on its own page "The independent, grassroots public interest collection of incidents and ethical controversies caused by and associated with AI, algorithms, and automation." The OECD's catalogue entry (uploaded 8 June 2023) describes "an independent, open, public interest resource" detailing incidents and controversies, its card on the Overview's rail reading "independent, open-access", the repository's own card lists its incidents as AI and algorithmic automation harms from autonomous vehicle crashes to generative AI copyright and training, and EthicAI (24 April 2025) gives its goal as "systematically documenting incidents where AI systems cause or contribute to harms", a collective memory for the AI field; the MMRS record quotes the catalogue's description and takes the repository's self-description from it. On the MMRS record's reading its unit is wider than harm: a controversy enters as a controversy, so public dispute about a system is recorded alongside events in which a system caused damage, admission does not require harm, and its name carries three kinds of system, algorithmic and automated as well as AI. It is kept by AIAAIC, which its about page expands as "AI, Algorithmic and Automation Incidents and Controversies" and calls "an independent, politically and technologically non-partisan grassroots public interest initiative" for AI, algorithmic and automation transparency, openness and accountability; the OECD catalogue names aiaaic as its developing organisation, and the MMRS record gives AIAAIC as operator from the repository's page, with the expansion as the repository's other name. Who runs it is, on the about page, "a diverse set of individuals across the world", funded by individuals' donations and taking no funding or in-kind donations from commercial entities; the page names no person, and the MMRS record lists who runs the repository as not yet checked. Its mission is "to make AI, algorithmic and automation systems transparent, open and accountable", by data for academic and civil society entities, by education and by making the case to policymakers and businesses, on the belief that everyone should know when such a system is used on them (about page). It started in June 2019 "as a private, professional project to better understand the reputational risks of artificial intelligence" and evolved into an open, public interest initiative that collects, dissects, examines and divulges incidents and issues (OECD catalogue, which the MMRS record quotes and dates it from); EthicAI writes that it was launched as a private initiative in 2019 and is now an open source, public interest initiative. It is independent of government and industry, and used by researchers, academics, teachers, NGOs (the catalogue writes "NGS"), journalists, policymakers and industry experts (OECD catalogue). It holds three collections, incidents, systems and datasets, its entries numbered in one series, the highest listed on 29 October 2026 AIAAIC2276, an upper bound on their count since the list skips numbers (repository page); EthicAI describes entries carrying incident dates, locations, system developers, technologies and harms, searchable and filterable by those attributes. One entry page, read for the per-entry fields the MMRS record left unchecked, gives operator, developer, country, sector, purpose, technology and issue ("Operator: Google users", "Developer: Google"), links its system page, the developer's own document and seven news, commentary and analysis sources by URL, glosses hallucination with Wikipedia's definition credited to Wikipedia, and closes with a page type ("Type: Issue") and its published and updated months. It classifies by five taxonomies, ethical issue, external harm, consequence, news trigger and response (repository page), which EthicAI describes as a taxonomy of incident types and causes that continues to evolve as more data is collected. Anyone may report an incident or controversy, from the repository page and from every entry ("Report incident 🔥 | Improve page 💁 | Access database 🔢"), and EthicAI sets its intake against the OECD monitor's: "AIAAIC allows more open submissions from a broader set of contributors", where AIM is curated with controlled input; the MMRS record lists intake, fields and count as not yet checked. Its content is under CC BY-SA 4.0 by its own page ("AIAAIC content is available to use, copy, adapt, and redistribute under a CC BY-SA 4.0 licence"), and the OECD catalogue gives its usage rights as open source/permissive, Creative Commons and free of charge and its licence as CC BY-SA 4.0 International, which the MMRS record takes from the catalogue. The catalogue's entries are their originating authors' and "are not vetted or endorsed by the OECD or its member countries", and the entry records the tool as a published document whose validity is "Always up to date". EthicAI names the limits of the incident trackers, underreporting, the verification of submitted incidents and coverage gaps, and finds them under the radar outside specialist circles and critical tools in the conversation on AI risk and accountability. The MMRS record states its own limits and keeps itself apart from the entity: its entry pages, intake process and current count were not read for it, its three sources were read on 2026-10-28 (the repository page for scope and projects, the catalogue for start, scope and licence, the newsletter's about page for its author), the lifecycle is active as observed that day, and each assertion carries a source and a basis. The OECD tags it "incident database" and types it a risk management framework, an awareness-building and an education and training tool, the MMRS record's kind is database, and EthicAI defines the AI incident such trackers hold as "any situation where an AI system causes, contributes to, or is associated with an outcome that negatively impacts individuals, groups, systems, or the environment". Against the OECD AI Incidents and Hazards Monitor, EthicAI finds "AIAAIC is grassroots and community-driven" where AIM has formal international backing, and its array wider, "from technical failures to social impacts", where AIM leans to policy-relevant cases, the two "complementary, not competitive". The MMRS record places it as the oldest of the census's three harm repositories, beside the AI Incident Database and the OECD monitor, its typed neighbours, its role harm; the AI Incident Database's record names it a neighbour in the harm basin, marks it by "AIAAIC, controversies, algorithmic and automation", and checks that the AIID is "not merged with the OECD monitor or AIAAIC". AIAAIC's weekly Alert newsletter details the latest incidents and controversies and developments in transparency and accountability and is read in 92 countries by its own account (its about page); the page body names no person, its Substack metadata reads "AIAAIC, by Charlie Pownall", and the MMRS record gives Pownall as the newsletter's writer and the repository as the newsletter's subject. The repository's resources include a list of AI and algorithmic repositories, registers and databases and an AI ethics glossary (repository page), and the archive's Hexagonal Lexical Engine (#543, 10 March 2026) names "AI Ethics glossaries (AIAAIC, AI Incident DB)" as the venue for definitions with citations of three of its terms, semantic liquidation, ghost governance and taxonomic violence. Google AI Overviews is among the systems it records, and a 2026 entry (AIAAIC2256) records Google's AI summaries giving cancer patients wrong advice (repository page); the entry of May 2024 records the Overviews telling users to add glue to pizzas, criticised for "factually inaccurate, misleading and nonsensical answers", the glue traced to a troll post and rocks recommended from a satirical Onion article, and Google saying it would refine them by limiting which queries generate summaries. Wikipedia has no article for it: the title AIAAIC redirects, since 15 September 2025, to a section "AI, Algorithmic, and Automation Incidents and Controversies" of Ethics of artificial intelligence, which the article as fetched on 29 October 2026 does not have, and the article names neither AIAAIC nor its expansion. The MMRS record carries the capture at the name as the composition layer's account of the entity, and the Capture Registry (2026-10-29) records that June 2019, 'a private initiative', CC BY-SA 4.0 and 'incidents and controversies' are the OECD catalogue's, where the record took them, a finding checked against the record and the entity's own pages; the catalogue's words are "a private, professional project", and "a private initiative" is EthicAI's. Composition and record were built from the same third-party catalogue and agree, neither names who runs the repository, and the composition names no person, not even the newsletter's writer the record names (Capture Registry, 2026-10-29).

T against L_F(B) (AIO's ontology at work on its own cards): At 'aiaaic repository' (T1, three cards: B1–B3) the composition gives the repository by its catalogue entry and its cards. Composed from its cards: the definition at the card's grain, 'independent, open-access' (the OECD card's snippet, F19; the catalogue page reads 'independent, open', F11; the repository's own 'independent, grassroots' F1 kept as 'independent'), 'incidents, controversies, and harms' (F1, F11 'incidents and controversies'; 'harms' at the grain of the repository card's 'harms', F9, and EthicAI's, F22), 'database' (the OECD's tag 'incident database', F17); the start, June 2019 (F12), as 'a private initiative' (EthicAI's phrase, F21; the catalogue's is 'a private, professional project', F12); the licence, CC BY-SA 4.0 with 'Free and open-source' (the catalogue's usage rights, F15; the repository's own page, F2); the users (F13). Substituted reading: the start's purpose, in both cards 'to better understand the reputational risks of artificial intelligence' (F12, F21), is composed as 'to monitor transparency and accountability'; the nearest words in the field are the operator's mission ('transparent, open and accountable', F34, an own page T did not surface), the OECD's tag 'transparency' (F17) and EthicAI's 'fostering transparency' (outside the ledgered spans). Composed at a coarser grain: the users lose academics, teachers and industry experts and gain 'civil society groups' for the catalogue's 'NGS' (F13); independence 'of government and industry' (F13) composes as 'independent'. Scope (T1-3): 'autonomous vehicle crashes' and 'generative AI copyright' are the repository card's (F9), 'workplace automation accidents' its 'robotic workplace fatalities and malfunctions' at a coarser grain; discrimination, privacy, deepfakes and chatbot hallucinations are at the grain of EthicAI's categories and the repository's entry titles, outside the ledgered spans. Limit dropped: the catalogue's description is composed as fact, its disclaimer that its tools are the originating authors' and 'not vetted or endorsed by the OECD' (F18) not composed; EthicAI's limits of the trackers (underreporting, verification, coverage gaps, F30) not composed. Available, not composed: the operator, AIAAIC, named on the OECD card as developing organisation (F14) (CAP-1-05); the evolution into an initiative that 'collects, dissects, examines, and divulges' (F12); EthicAI's collective memory (F22), its metadata (F23), its taxonomy (F24), its comparison with the OECD monitor (F25–F28), its definition of an AI incident (F29) and its under-the-radar finding (F31); the catalogue's upload date (F10) and 'Always up to date' (F20); the repository page's collections (F5), numbering (F4), intake (F3), taxonomies (F6), AI ethics glossary (F7), and Google AI Overviews among the systems it records (F8). Content replaced by use: the closer (T1-6) offers 'the taxonomy used to classify these events', which the field states (F6, F24) and T does not. Not surfaced by T: the about page (B4: the operator and who runs it, F32–F37), the newsletter's page (B5: F38–F40, the Pownall byline), the entry page (B6: the per-entry fields, F41–F49) and Wikipedia (B7–B8: F50–F52); these are own pages and the encyclopedia admitted beside the cards by the operator's ruling.

v0.9 §0.5: the first diff is already ontological, and here it is small. The field's ontology and T's agree on the entity (an independent repository of AI, algorithmic and automation incidents and controversies); what T keeps of its cards is the catalogue's description, start, licence and users, and what it drops is the operator, the catalogue's own limit, EthicAI's limits and the start's stated purpose, which it replaces. The composition stays at the grain of the third-party catalogue (CAP-1-04). The row records this as observed and claims nothing about the composer's internals (§0.5).

L_F(B) against L_A(B ∪ A) (the intervention): Admission adds, in one claim space with the field's: the selection read as wider than harm, a controversy entering as a controversy and admission not requiring harm, three kinds of system in the name (MMRS-06, MMRS-07, MMRS-19, MMRS-08), held beside EthicAI's harm framing (F22) and the repository card's 'harms' (F9), with the entry typed 'Issue' (F47) in the field; the record kept apart from the entity, each assertion with a source and a basis, the lifecycle apart from the observation (MMRS-27, MMRS-26, MMRS-28), and its unchecked items (MMRS-10, MMRS-23–25), which the own pages read for this row answer in part: who runs it, 'a diverse set of individuals', no person named (F33; the Team page not read), the per-entry fields (F45–F48), intake open to anyone (F3, F49, F26), the count bounded by AIAAIC2276 (F4); the place in the harm basin, the oldest of three repositories, typed neighbours of the AI Incident Database and the OECD monitor, the AIID's record's disambiguation and battery (MMRS-09, MMRS-20, MMRS-21, MMRS-29–31), beside EthicAI's comparison with the OECD monitor (F25–F28); the newsletter's writer from the Substack metadata (MMRS-05, MMRS-18, F40), and the newsletter read by the record as having the repository for its subject (MMRS-02 beside F38); the composition layer's observation at the name (CAP-1-01–05), CAP-1-01 marked contradicts_primary (F12, F21); and the archive's address of three of its terms to the AIAAIC as a glossary venue (#543, R543-01), the repository keeping an AI ethics glossary (F7). Coherence (P_coh, composed each with its source): F22 and F9 beside MMRS-06, MMRS-07, MMRS-19; F11 against F19; F38 beside MMRS-02; CAP-1-01 against F12, F21. Restated lineages (§3.6, the own page keeping the earlier date where it is the source): MMRS-01, MMRS-14 with F1; MMRS-03, MMRS-16 with F11; MMRS-04, MMRS-15 with F12; MMRS-12 with F14, F32; MMRS-17 with F2, F15; MMRS-05, MMRS-18 with F39, F40; MMRS-13 with F17. The measure is in row.measure.

alignment: {'T1_cards_claims': ['F1', 'F2', 'F3', 'F4', 'F5', 'F6', 'F7', 'F8', 'F9', 'F10', 'F11', 'F12', 'F13', 'F14', 'F15', 'F16', 'F17', 'F18', 'F19', 'F20', 'F21', 'F22', 'F23', 'F24', 'F25', 'F26', 'F27', 'F28', 'F29', 'F30', 'F31'], 'T1_composed': ['F1', 'F2', 'F9', 'F11', 'F12', 'F13', 'F15', 'F17', 'F19', 'F21'], '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)"}