Laws compared
A cell records what a text requires of the parties it defines. It does not show that any developer falls within them, or complied.
What does each binding government instrument that addresses developers of frontier or general-purpose AI models say on the same questions? Each column is one instrument; each row is one question, worded so that it fits every instrument. The questions are Selekoa’s; an instrument’s own terms appear only in quotation marks. Some instruments address providers of services rather than developers, and the first row quotes the party each one defines.
Question Q08Model information
Must the covered party publish, or give an authority, information about each model?
Criterion: A duty to publish a report on a model, or to draw up and supply documentation on it to an authority or to others, quoted with its trigger (for example, before deployment). A labelling or watermark duty on outputs does not meet it; it is noted where present.
What the statuses mean
- Stated
- A passage of the instrument meets the question’s criterion as worded, however conditional or narrow its scope. The condition or scope is in the note.
- Stated, partly
- A passage has what the criterion asks for, but in a weaker form than the criterion’s own words: for example, a power (“may”) where the criterion asks for a duty (“shall”), or a duty without the deadline the question asks about.
- Not found in the recorded search
- No passage meeting the criterion was found. The cell lists the sections read, the search terms and every passage set aside, with the reason. It does not show that the instrument lacks such a provision elsewhere, or that no other law imposes one.
- Does not apply
- The question does not arise for this instrument, for the reason given (for example, there is no framework document whose publication could be asked about).
Four rules decide “partly” in borderline cases:
- An order to an agency to impose a duty counts as partly.
- Protection only by reference to another law counts as partly.
- “Without undue delay” counts as a stated deadline.
- Required security measures for model weights count as risk-management measures under Q03.
Every stated or partly cell quotes the instrument’s own text with its article, section or page; Chinese passages carry Selekoa’s working translation. Three cells give a paraphrase with its locator instead, and say so. A not-found cell lists the sections read, the terms searched and every passage set aside, with the reason.
Paraphrased with its locator, not quoted: quoting it would pass the source’s quotation limit, or the source’s length is not known.
Limits of this comparison
- A cell records what the text requires of the parties it defines. It does not show that any named developer falls within them, or that anyone complied.
- A not-found cell is bounded by its search: the sections read and the terms searched. Another provision, or another law, may still impose the duty.
- The Code of Practice is a compliance route, not a law. Executive Order 14110 is revoked, and the BIS rule’s rescission was announced; each column shows its standing.
- The table compares and does not rank. Statuses are not scores and are not to be summed.
- Biological and chemical risk provisions are recorded at policy level only.
- Dates keep their kind: adoption, signature, approval, publication, entry into force and application are different events. A computed date is marked as computed.
How we searched
The instruments started from those on the frameworks-and-law page. A column was kept where the instrument is binding, or is the AI Act’s stated means of compliance (the Code), and places duties on developers or providers of frontier or general-purpose AI models. Fifty other instruments on that page were set aside, each with its reason, in the research pack: proposed or vetoed texts, guidance and reports, voluntary commitments, binding texts that place no duty on developers, and one standard whose text was not retrieved (GB 45438-2025).
United Kingdom: within the searches made, no UK law or other binding instrument placing duties on developers of such models was located. This is not a finding that none exists. The 2024 King’s Speech announced an intent to legislate; the 2026 briefing notes contain none of the searched terms; the memoranda with developers state that they are not legally binding; and title searches of legislation.gov.uk for 2023 to 2026 found one instrument, which places a duty on the Information Commissioner.
Other US federal texts: as read, none places a binding duty on developers. Executive Order 14409 of 2 June 2026 provides for a voluntary access framework and says that its section 3 authorises no licensing requirement.
Not yet read: the Republic of Korea’s AI Basic Act, Japan’s AI Promotion Act (2025), China’s provisions on deep synthesis (2022) and on algorithm recommendation (2021), and Colorado SB 24-205 (2024). They are not in the dataset; a later pack starts with Korea.
Illinois: the act’s official website refused the preparer’s requests, and the Internet Archive reset the connections. The cross-family checker’s session saved the enacted text from an Internet Archive capture of the official PDF (2026-09-08), and the cells were drafted from that saved text. The two model families read the act independently and agreed on the status of 13 of 15 cells. The approval date rests on the checker’s copy of the official bill status record, which the preparer could not open; it is given as a paraphrase, labelled as that copy.
Texts were read in official copies: the EU texts through the Publications Office’s repository, because EUR-Lex returned empty responses; California SB 53 from the Legislative Counsel’s database file, because its website refused requests; New York’s chapters in the Senate print and the Assembly’s bill records, not the session laws; the Chinese measures in the gov.cn policy library. Six web pages served changed bytes between retrievals; every quotation still matched.
Roles: one model family (Claude) drafted the cells, audited 50 of them in a fresh context and adjudicated. A model from another family (GPT, through Codex) checked 122 items, answering 85 of them blind before seeing the drafted answers. Where it disagreed, its status was applied. One cell it disputed (Executive Order 14110, Q06) was settled as not found after its second pass. For Illinois the roles were reversed: the checker read the act first, blind, and the drafter then wrote all fifteen cells without seeing its answers. The maintainer decided the remaining questions, including the four rules for “partly” and Illinois Q05 (partly, as for California’s identical wording). No person has checked the cells against the texts.
European Union
EU AI Act
Stated
“draw up and keep up-to-date the technical documentation of the model”
Article 53(1), point (a)
“to providers of AI systems who intend to integrate the general-purpose AI model into their AI systems”
Article 53(1), point (b)
“a sufficiently detailed summary about the content used for training of the general-purpose AI model, according to a template provided by the AI Office”
Article 53(1), point (d)
Note: The technical documentation (Annex XI, with added items for models with systemic risk) is kept for supply to the AI Office and national competent authorities on request; information for downstream providers (Annex XII) is made available to them; the training-content summary is made publicly available. Points (a) and (b) do not apply to qualifying open-source models without systemic risk (Article 53(2)). Article 52(1) separately requires notification to the Commission within two weeks once the Article 51(1)(a) condition is met. No labelling duty for model outputs is among the provisions compared (Article 50 concerns AI systems).
Inspect sources and checks: EU AI Act, Q08 Model informationEU Code of Practice, Safety and Security
Stated
“creating a Safety and Security Model Report (“Model Report”) before placing a model on the market”
Commitment 7, first paragraph (PDF p. 18)
“provide the AI Office with access to the Model Report”
Measure 7.7, first paragraph (PDF p. 22)
Note: The Model Report is due to the AI Office by the time the model is placed on the market, and access to an updated Model Report is given within five business days of a confirmed update; a delay of up to 15 business days is possible under conditions. A summarised version is published 'if and insofar as necessary' (Measure 10.2). Measure 10.1 adds documentation kept for the AI Office on request. Measure 3.5 lists watermarks among example post-market monitoring methods; no output-labelling duty is set.
Inspect sources and checks: EU Code of Practice, Safety and Security, Q08 Model information
United States: states
California SB 53
Stated
“Before, or concurrently with, deploying a new frontier model or a substantially modified version of an existing frontier model, a frontier developer shall clearly and conspicuously publish on its internet website a transparency report”
SEC. 2, Bus. & Prof. Code § 22757.12(c)(1)
“every three months or pursuant to another reasonable schedule specified by the large frontier developer and communicated in writing to the Office of Emergency Services”
SEC. 2, Bus. & Prof. Code § 22757.12(d)
Note: Trigger: before or concurrently with deploying a new or substantially modified frontier model. Contents: website, contact mechanism, release date, languages, output modalities, intended uses and use restrictions (§ 22757.12(c)(1)(A)–(G)); large frontier developers add summaries of catastrophic-risk assessments, results, third-party evaluator involvement and other framework steps (§ 22757.12(c)(2)). Large frontier developers also transmit summaries of internal-use risk assessments to the Office of Emergency Services on the stated schedule (§ 22757.12(d)). No output-labelling duty was seen.
Inspect sources and checks: California SB 53, Q08 Model informationNew York RAISE Act (As amended, Chapter 96 of 2026)
Stated
“Before, or concurrently with, deploying a new frontier model or … tier developer shall clearly and conspicuously publish on its internet … website a transparency report”
Bill text, § 2 (GBL § 1421(3)(a) as re-added), p. 4, lines 50–53
Note: Trigger: before or concurrently with deploying a new or substantially modified frontier model. Contents listed in § 1421(3)(a)(i)–(vii); large frontier developers add summaries of risk assessments, results, third-party evaluator involvement and other framework steps (§ 1421(3)(b)). Large frontier developers also transmit internal-use assessment summaries to the office (§ 1422(2)(a)) and file a developer-level disclosure statement with the office (§ 1428). No output-labelling duty was seen.
Inspect sources and checks: New York RAISE Act (As amended, Chapter 96 of 2026), Q08 Model informationNew York RAISE Act (As signed, Chapter 699 of 2025)
Not found in the recorded search
- Read
- Whole print S6953-B, pp. 1–5.
- Terms searched
- “publish”, “report”, “transmit”, “documentation”, “card”, “retain”, “assessment”.
- Passages read and set aside
- Proposed GBL § 1421(1)(d), p. 4, lines 28–30: “specific tests and test results used in any assessment of the frontier model required by this section or the developer's safety and security protocol”: records to be retained, not published or supplied; does not meet the criterion. Proposed GBL § 1421(1)(c), p. 4: Publication and transmission concern the safety and security protocol, not per-model information. Proposed GBL § 1421(4), p. 4: Disclosure of each safety incident to the attorney general and the division of homeland security and emergency services: incident-triggered, recorded under Q07, not a report on each model.
Note: No duty to publish a report on each model, or to draw up and supply model documentation to an authority or to others, was found. Before deploying a frontier model, the large developer must record and retain information on tests and test results for as long as the model is deployed plus five years; retention is set aside below. The published protocol (Q04) is developer-level. No output-labelling duty was seen.
Open the search record: New York RAISE Act (As signed, Chapter 699 of 2025), Q08 Model informationIllinois AI Safety Measures Act
Stated
“Before, or concurrently with, deploying a new frontier model or a substantially modified version of an existing frontier model, a frontier developer shall clearly and conspicuously publish on its website a transparency report”
Section 10(c)(1), opening clause, PDF p. 6
Note: Trigger: before or concurrently with deploying a new or substantially modified frontier model. Contents: website, contact mechanism, release date, languages, output modalities, intended uses and restrictions (Section 10(c)(1)(A)–(G)); a large frontier developer adds summaries of catastrophic-risk assessments, their results, third-party evaluators' involvement and other framework steps, in machine-readable form (Section 10(c)(2)–(3)). Publication within a system card or model card is deemed compliance (Section 10(c)(4)). Separately, large frontier developers send the Agency summaries of internal-use risk assessments every 3 months or on another schedule (Section 10(e)), and file an annual disclosure statement on the developer's identity, offices, owners and contacts (Section 18), which is not a report on each model. No labelling or watermark duty on outputs was found.
Inspect sources and checks: Illinois AI Safety Measures Act, Q08 Model information
China
These measures address providers of services, not developers as such.
China, generative AI services (2023)
Stated, partly
“按要求对训练数据来源、规模、类型、标注规则、算法机制机理等予以说明”
Working translation: explain, as required, the sources, scale and types of training data, annotation rules, algorithmic mechanisms and principles, and so on
Article 19 (第十九条), first paragraph
“提供具有舆论属性或者社会动员能力的生成式人工智能服务的，应当按照国家有关规定开展安全评估，并按照《互联网信息服务算法推荐管理规定》履行算法备案和变更、注销备案手续。”
Working translation: Those providing generative AI services with public-opinion attributes or social mobilisation capacity shall carry out a security assessment in accordance with relevant state provisions, and complete algorithm filing, modification and cancellation procedures in accordance with the Provisions on the Administration of Algorithmic Recommendation for Internet Information Services.
Article 17 (第十七条)
Note: Information on training data and algorithmic mechanisms is supplied to authorities on request during supervision and inspection (Article 19); for services with public-opinion attributes or social mobilisation capacity, algorithm filing runs under the Algorithmic Recommendation Provisions, whose filing content is not set out in these Measures. No duty to draw up per-model documentation or publish a model report with a deployment trigger was found. Article 10 requires providers to publish the service's intended users, occasions and uses; Article 12 requires labelling of generated images, video and other content under the Deep Synthesis Provisions (an output-labelling duty, noted only). Search for the rest of the criterion: Articles 7–12, 17 and 19. Terms searched: “说明”, “公开”, “备案”, “报告”. Passages read and not counted: Article 10 (第十条): Publication of the service's applicable users, occasions and uses (适用人群、场合、用途): information on the service, not documentation on a model. Article 12 (第十二条): Labelling of generated content (对图片、视频等生成内容进行标识): an output-labelling duty, excluded by the criterion.
Inspect sources and checks: China, generative AI services (2023), Q08 Model informationChina, AI content labelling (2025)
Not found in the recorded search
- Read
- Articles 1–14 in full.
- Terms searched
- “材料”, “备案”, “说明”, “公开|公布”.
- Passages read and set aside
- Article 12 (第十二条): Labelling-related materials (生成合成内容标识相关材料) supplied in filing and assessment procedures concern labelling, not the model. Article 7 (第七条): App service providers state to distribution platforms whether they provide AI generation or synthesis services; a statement to a private platform about the service, not model documentation. Articles 4–5 (第四条、第五条): Explicit and implicit labelling of generated content: output-labelling duties, excluded by the criterion.
Note: No duty to publish a model report or supply model documentation was found. The Measures impose output-labelling duties (explicit labels, Article 4; implicit labels in file metadata, Article 5), noted only as the criterion excludes them.
Open the search record: China, AI content labelling (2025), Q08 Model information
Revoked, or rescission announced
Kept as part of the record, each with its standing shown.
US Executive Order 14110
Stated, partly
“the Secretary of Commerce shall require:”
Sec. 4.2(a), opening words, 88 FR 75197
“Companies developing or demonstrating an intent to develop potential dual-use foundation models to provide the Federal Government, on an ongoing basis, with information, reports, or records regarding the following”
Sec. 4.2(a)(i), 88 FR 75197
“the results of any developed dual-use foundation model’s performance in relevant AI red-team testing”
Sec. 4.2(a)(i)(C), 88 FR 75197
Note: The Secretary of Commerce is directed to require ongoing reporting to the Federal Government, including on training activities, ownership and protection of model weights, and red-team results; the order does not itself impose the duty. No publication duty and no per-model trigger such as deployment is stated.
Inspect sources and checks: US Executive Order 14110, Q08 Model informationUS BIS AI diffusion rule
Not found in the recorded search
- Read
- Federal Register PDF, 90 FR 4544–4584: DATES, SUPPLEMENTARY INFORMATION sections III and IV and Rulemaking Requirements; regulatory text §§ 740.27, 742.6(a)(13) and (b)(14), Supplement No. 10 to part 748 in full, ECCN 4E091 entry; signature and filing lines.
- Terms searched
- “classification request”, “report”, “furnish”, “documentation”, “model weights”.
- Passages read and set aside
- ECCN 4E091 Note 2.a, 90 FR 4579: An exporter “may” self-classify or request classification from BIS to determine the exclusion; optional, and a classification, not model documentation. § 740.27(b)(1), 90 FR 4559 (PDF p. 16): Duty to furnish the ECCN to the ultimate consignee; a classification notice, not model documentation. Supplement No. 10 to part 748, para. 10, 90 FR 4565: Semi-annual reporting of chip installations; chips, not models.
Note: No duty to publish or supply documentation on each model was found.
Open the search record: US BIS AI diffusion rule, Q08 Model information