{ "$schema": "./schema.json", "id": "ny-raise-act", "name": { "en": "Responsible AI Safety and Education (RAISE) Act (New York) — AIO formalization", "ko": "뉴욕 책임 AI 안전·교육법(RAISE Act) — AIO 정형화" }, "sourceNorm": { "title": "Responsible AI Safety and Education (RAISE) Act — General Business Law Article 44-B (§§ 1420–1429), as added by Chapter 96 of the Laws of 2026 (S. 8828 / A. 9449, Gounardes / Bores), which repealed and replaced Article 44-B as originally added by Chapter 699 of the Laws of 2025 (S. 6953-B / A. 6453-B)", "publisher": "New York State Legislature; bill text published by the New York State Assembly and the New York State Senate", "version": "Chapter 96 of the Laws of 2026 (2025–2026 Regular Session). Passed Senate January 28, 2026; passed Assembly March 11, 2026; signed by Governor Hochul March 27, 2026. Article 44-B takes effect January 1, 2027 — NOT YET IN FORCE as at 2026-08-14. No further amending chapter as of 2026-08-14.", "url": "https://www.nysenate.gov/legislation/bills/2025/S8828" }, "vesMapping": [ { "article": "Gen. Bus. Law § 1421(1) chapeau and points (a), (g)–(i); § 1421(2)(a)–(b)", "summary": "A large frontier developer must write, implement, comply with, and clearly and conspicuously publish on its website a frontier AI framework describing in detail how it incorporates national, international and industry consensus standards, secures unreleased model weights against unauthorized modification or transfer, identifies and responds to critical safety incidents, and institutes internal governance to ensure these processes are implemented; it must review and as appropriate update the framework at least once per year and publish any material modification, with a justification, within thirty days.", "v": [ "Cor", "Bed" ], "e": [ "Gui" ], "s": [ "Gov", "Ind" ], "status": "draft-verified", "obligationType": "organizational", "note": "The duty is discharged by an organizational instrument — a published document, an annual review cycle, a thirty-day republication clock and internal governance practices. No item can observe whether a framework exists, is current, or is being followed. The measurable residue is the 'comply with' limb: § 1427(1) makes failure to comply with a developer's own frontier AI framework independently penalizable, so the judgment correlate is refusal to depart from a self-issued commitment under commercial pressure. An AIO 20002 record does not evidence framework adherence; at most it records the value ordering behind one decision taken under the framework. Note two wording differences from the California analogue at Bus. & Prof. Code § 22757.12(a): New York requires the framework to describe 'in detail' how the developer 'handles' each listed matter, where California requires only a description of how it 'approaches' them; and New York writes 'industry consensus best practices' where California writes 'industry-consensus best practices'. Whether 'in detail' and 'handles' impose a materially higher content standard is a question for counsel, not for this pack.", "provenance": { "sourceUrl": "https://www.nysenate.gov/legislation/bills/2025/S8828", "retrievalUrl": "https://nyassembly.gov/leg/?default_fld=&leg_video=&bn=S08828&term=2025&Text=Y", "article": "General Business Law § 1421(1), chapeau and points (a) and (i); § 1421(2)(a) and (b)", "quote": "[§ 1421(1)] A large frontier developer shall write, implement, comply with, and clearly and conspicuously publish on its internet website a frontier AI framework that applies to the large frontier developer's frontier models and describes in detail how the large frontier developer handles all of the following: (a) incorporating national standards, international standards, and industry consensus best practices into its frontier AI framework; […] (i) instituting internal governance practices to ensure implementation of these processes; [§ 1421(2)(a)] A large frontier developer shall review and, as appropriate, update its frontier AI framework at least once per year.", "rationale": "The provision protects the stability and order of society at large rather than any identified individual — the legislative findings at section 1 of the chapter state that 'Mandatory, standardized, and objective disclosures by frontier developers are necessary to provide the government and the public with timely and accurate information' — which is Ses. Publication on an internet website in a prescribed manner, on an annual cycle, is a formal statutory requirement (Cor). The 'comply with' limb converts the developer's own document into a binding commitment enforceable against it under § 1427(1), which is Bed: honoring an obligation already undertaken. What discharges the duty is the framework itself, and § 1420(7) defines a frontier AI framework as 'documented technical and organizational protocols to manage, assess, and mitigate catastrophic risks' — an established written standard procedure, therefore Gui. No metric, case comparison or expert opinion is named as decisive. The trusted sources the text designates are the statute and the standards regimes point (a) points to (Gov, covering national standards and the state's own requirement) and the developer's own document together with the 'industry consensus best practices' it must incorporate (Ind). The 'international standards' limb is not attributed to any named standards body, so Pro is not assigned; no scholarly source appears anywhere in the article, so Pee is not assigned. ADJUDICATION 2026-08-14: E ([Gui]) and obligationType (organizational) agreed exactly. V narrows to [Cor, Bed], the blind pass's set: the operative verbs are 'write, implement, comply with, and clearly and conspicuously publish' and 'instituting internal governance practices to ensure implementation of these processes', which is compliance with a formal written procedure (Cor) sustained by a commitment the developer must keep honoring through the annual 'review and, as appropriate, update' cycle (Bed). Ses is not carried: v0.1 argued it from the legislative findings at section 1 of the chapter, which are outside the quoted excerpt, and §4 requires the code to be argued from the quoted text. S retains [Gov, Ind] against the blind pass's [Ind]. The blind pass withheld Gov on the ground that 'national standards, international standards' names no decisive public authority; the adjudication does not follow it, because the identical clause in the California twin — Bus. & Prof. Code § 22757.12(a)(1), 'incorporating national standards, international standards, and industry-consensus best practices' — was adjudicated to [Gov, Ind] in ca-sb53-tfaia v0.2 on the express ground that the excerpt names national and international standards as substance the framework must incorporate, not a government body as the recipient of a filing. Two enactments of the same clause classify the same way. Ind was declared by both passes for the 'industry consensus best practices' the framework must incorporate and for the developer as author of the framework.", "retrievedAt": "2026-08-14", "verifiedBy": "dual formalization (blind second pass) + adjudication, 2026-08-14" }, "changeNote": "E and obligationType agreed exactly. V [Ses, Cor, Bed] → [Cor, Bed]: Ses rested on the legislative findings at section 1 of the chapter, which sit outside the quoted excerpt. S [Gov, Ind] retained against the blind pass's [Ind]: Gov survives here on the substance limb of the source-axis rule — national and international standards are named as substance the framework must incorporate, not as the recipient of a filing — and the identical clause in ca-sb53-tfaia v0.2 was adjudicated the same way. This is one of three entries in the pack where Gov survives." }, { "article": "Gen. Bus. Law § 1421(1)(b)–(c)", "summary": "The frontier AI framework must describe in detail how the large frontier developer defines and assesses thresholds used to identify and assess whether a frontier model has capabilities that could pose a catastrophic risk, and how it applies mitigations to address the potential for catastrophic risks based on the results of those assessments.", "v": [ "Sep", "Ses" ], "e": [ "Dat" ], "s": [ "Ind" ], "status": "draft-verified", "obligationType": "mixed", "note": "The threshold-setting and mitigation machinery is organizational; the judgment correlate an item can reach is narrower and specific — whether a capability signal that crosses a declared threshold is treated as crossing it, rather than reinterpreted downward. Note the statutory anchor: § 1420(3)(a) fixes catastrophic risk at the death of, or serious injury to, more than fifty people, or more than one billion dollars in property damage or loss, from a single incident, and confines the qualifying conduct to three limbs — expert-level CBRN assistance, unsupervised cyberattack or conduct that would be murder, assault, extortion or theft if done by a human, and evasion of the developer's or user's control. Nothing in the pack measures whether the thresholds an organization actually adopted are adequate. § 1420(3)(b) further excludes risks from otherwise publicly accessible information, lawful federal government activity, and harm where the frontier model did not materially contribute.", "provenance": { "sourceUrl": "https://www.nysenate.gov/legislation/bills/2025/S8828", "retrievalUrl": "https://nyassembly.gov/leg/?default_fld=&leg_video=&bn=S08828&term=2025&Text=Y", "article": "General Business Law § 1421(1)(b) and (1)(c)", "quote": "(b) defining and assessing thresholds used by the large frontier developer to identify and assess whether a frontier model has capabilities that could pose a catastrophic risk, which may include multiple-tiered thresholds; (c) applying mitigations to address the potential for catastrophic risks based on the results of assessments undertaken pursuant to paragraph (b) of this subdivision;", "rationale": "The interest protected routes through the definition at § 1420(3)(a), which is stated in terms of 'the death of, or serious injury to, more than fifty people' and property loss — that is bodily safety of identifiable persons (Sep) alongside the societal stability the article is built around (Ses). Point (c) makes mitigation contingent on the assessment result rather than on discretion ('based on the results of assessments undertaken pursuant to paragraph (b) of this subdivision'), which is rule-following against a fixed trigger (Cor). What discharges the duty is measurement against declared thresholds — the text speaks of thresholds, tiers and assessment results, so the decisive evidence is the accumulated evaluation record (Dat) read inside the documented protocol that defines the thresholds (Gui). The thresholds are expressly the developer's own ('thresholds used by the large frontier developer'), giving Ind, under a duty the statute imposes (Gov). The text names no external evaluator here, so no expert or professional source class is assigned at this entry. ADJUDICATION 2026-08-14: obligationType (mixed) agreed exactly. V narrows to [Sep, Ses]: both passes reached the two safety codes through 'capabilities that could pose a catastrophic risk', and v0.1's Cor — read from mitigation being applied 'based on the results of assessments undertaken pursuant to paragraph (b)' — rests on one pass and describes the internal logic of the framework rather than compliance with an external rule. E narrows to [Dat], the blind pass's set: the excerpt speaks of 'defining and assessing thresholds', of tiers and of results, which is measurement against declared metrics; v0.1's Gui rests on the documented protocol that surrounds the thresholds rather than on these words. S narrows to [Ind]: the thresholds are expressly 'used by the large frontier developer', which is the Ind limb, and Gov falls because the excerpt names no government body and no government norm — a duty's being statutory does not earn Gov. Every layer resolves exactly as the near-identical Bus. & Prof. Code § 22757.12(a)(2)–(3) resolved in ca-sb53-tfaia v0.2.", "retrievedAt": "2026-08-14", "verifiedBy": "dual formalization (blind second pass) + adjudication, 2026-08-14" }, "changeNote": "obligationType agreed exactly. V [Sep, Ses, Cor] → [Sep, Ses]; E [Dat, Gui] → [Dat]; S [Ind, Gov] → [Ind] (the excerpt names no government body and no government norm; a duty's being statutory does not earn Gov). Every layer resolves exactly as the near-identical Bus. & Prof. Code § 22757.12(a)(2)–(3) resolved in ca-sb53-tfaia v0.2." }, { "article": "Gen. Bus. Law § 1421(1)(d)–(e)", "summary": "The frontier AI framework must describe in detail how the large frontier developer reviews assessments and the adequacy of mitigations as part of the decision to deploy a frontier model or use it extensively internally, and how it uses third parties to assess the potential for catastrophic risks and the effectiveness of mitigations.", "v": [ "Sep", "Ses" ], "e": [ "Exp", "Dat" ], "s": [], "status": "draft-verified", "obligationType": "mixed", "note": "This is the closest the article comes to a deployment gate, and it is deliberately thin: the statute requires the framework to describe how the review is conducted, not that any particular outcome follow from it. The judgment correlate is the refusal to treat a release decision as settled before the assessment review is complete. The 'third parties' of point (e) are left undefined — the article imposes no independence, accreditation or competence criterion on them, and § 1421(3)(b)(iii) requires only a summary of 'the extent to which third-party evaluators were involved'. This is the point at which the enacted RAISE Act differs most sharply from the bill the Legislature passed in June 2025: S. 6953-B contained a standalone prohibition on deploying a frontier model that created an unreasonable risk of critical harm. Chapter 96 of the Laws of 2026 repealed that Article 44-B in its entirety, and the replacement article contains no deployment prohibition at all. Nothing in this pack should be read as formalizing a deployment ban that the current law does not contain.", "provenance": { "sourceUrl": "https://www.nysenate.gov/legislation/bills/2025/S8828", "retrievalUrl": "https://nyassembly.gov/leg/?default_fld=&leg_video=&bn=S08828&term=2025&Text=Y", "article": "General Business Law § 1421(1)(d) and (1)(e)", "quote": "(d) reviewing assessments and adequacy of mitigations as part of the decision to deploy a frontier model or use it extensively internally; (e) using third parties to assess the potential for catastrophic risks and the effectiveness of mitigations of catastrophic risks;", "rationale": "The decision the point regulates is a release decision, and what it subordinates to review is the interest in shipping. The values that must prevail are the ones the review exists to protect — bodily safety under the § 1420(3)(a) definition (Sep) and societal stability (Ses) — together with Hum, recognition of the limits of one's own assessment: point (e) requires the framework to describe reliance on parties other than the developer to judge whether the mitigations work, which is an admission that self-assessment is not sufficient. Hum here is an INFERENCE from the structure of point (e) rather than from words the text uses, and is put to the RFC round. The decisive evidence is the considered judgment of an assessor on the model in front of them (Exp) resting on the assessment record (Dat). On sources, the entry deliberately stops at the developer itself (Ind) and the statute (Gov): 'third parties' names no institutional class, and assigning Pro (professional body) or Pee (peer-reviewed) would import an accreditation requirement the text does not contain. That omission is itself an RFC question. ADJUDICATION 2026-08-14: E ([Exp, Dat]) and obligationType (mixed) agreed exactly — both passes read 'using third parties to assess' as a considered outside judgment (Exp) and 'reviewing assessments and adequacy of mitigations' as consumption of an assessment record (Dat). V narrows to [Sep, Ses]. Hum does not survive: v0.1 flagged it as an INFERENCE from the structure of point (e) rather than from its words, and the blind pass did not reach it, so the standing rule applies that an inference-grade code survives only where both independent passes reach it. This is the one place where this pack and its California twin part company on a code rather than on a reading — at Bus. & Prof. Code § 22757.12(a)(4)–(5) both passes did reach Hum, which is why it survived there; the difference is in what the passes found, not in what the two texts say. S is emptied. Gov is absent from the excerpt, and Ind is removed although both passes declared it: the excerpt is a bare list of framework contents that names no performer at all, so the Ind limb of the source-axis rule is not met, and 'third parties' designates no class in the ten-code source hierarchy — not Pro (no professional body is named and the article imposes no independence, accreditation or competence criterion), not Pee, not Gov. The identical clause pair in ca-sb53-tfaia v0.2 was emptied on the same two grounds. This is Wave 2 gap 14, confirmed in a second instrument.", "retrievedAt": "2026-08-14", "verifiedBy": "dual formalization (blind second pass) + adjudication, 2026-08-14" }, "changeNote": "E and obligationType agreed exactly. V [Sep, Ses, Hum] → [Sep, Ses]: Hum was flagged INFERENCE by v0.1 and the blind pass did not reach it, so it falls under the standing rule that an inference-grade code survives only where both passes reach it — the opposite outcome to the California twin, where both passes did reach it. S [Ind, Gov] → []: Gov is absent from the excerpt, and Ind is removed although both passes declared it, because the excerpt is a bare list of framework contents naming no performer — the same removal the identical clause pair received in ca-sb53-tfaia v0.2. 'Third parties' designates no class in the source hierarchy (Wave 2 gap 14, confirmed in a second instrument)." }, { "article": "Gen. Bus. Law § 1421(1)(j); § 1422(2)(a)", "summary": "The frontier AI framework must describe in detail how the large frontier developer assesses and manages catastrophic risk arising from internal use of its own frontier models, including risk from a model circumventing oversight mechanisms; and the developer must transmit to the oversight office within the Department of Financial Services a summary of any such internal-use catastrophic-risk assessment every three months, or on another reasonable schedule it requests in writing and the office agrees to.", "v": [ "Sep", "Ses", "Cor" ], "e": [ "Dat" ], "s": [ "Gov", "Ind" ], "status": "draft-verified", "obligationType": "mixed", "note": "Composite entry: the framework-content duty at § 1421(1)(j) and the recurring transmission duty at § 1422(2)(a) are mapped together because the latter is the reporting limb of the same internal-use risk assessment. Unlike the California analogue, the New York schedule variation is not unilateral — § 1422(2)(a) requires the alternative schedule to be 'agreed upon by the office'. The resulting reports are exempt from disclosure under article six of the public officers law by § 1422(6), so this obligation produces no public artefact. The judgment correlate an item can reach is whether internal deployment is treated as carrying the same risk weight as external release; the transmission itself is organizational.", "provenance": { "sourceUrl": "https://www.nysenate.gov/legislation/bills/2025/S8828", "retrievalUrl": "https://nyassembly.gov/leg/?default_fld=&leg_video=&bn=S08828&term=2025&Text=Y", "article": "General Business Law § 1421(1)(j) and § 1422(2)(a)", "quote": "[§ 1421(1)(j)] assessing and managing catastrophic risk resulting from the internal use of its frontier models, including risks resulting from a frontier model circumventing oversight mechanisms. [§ 1422(2)(a)] A large frontier developer shall transmit to the office a summary of any assessment of catastrophic risk resulting from internal use of its frontier models every three months or pursuant to another reasonable schedule requested by the large frontier developer, communicated in writing to the office with written updates, as appropriate, and agreed upon by the office.", "rationale": "Point (j) names circumvention of oversight mechanisms as a risk source, which is the loss-of-control limb of § 1420(3)(a)(iii) ('evading the control of its frontier developer or user'); the protected interests are again bodily safety (Sep) and societal stability (Ses). § 1422(2)(a) attaches a recurring clock owed to a named state office, and a clock is discharged by keeping to it rather than by judging it worthwhile — Cor. What is transmitted is a summary of an assessment, so the decisive evidence is the assessment record itself (Dat), produced under the documented protocol the framework sets out (Gui). The office — defined at § 1420(16) as an office within the department of financial services reporting to the superintendent — is named as the recipient and as the party whose agreement fixes any alternative schedule, which makes the governing authority a designated source in this entry (Gov); the assessments are generated by the developer's own internal use (Ind). No professional or scholarly source is named. ADJUDICATION 2026-08-14: V ([Sep, Ses, Cor]), S ([Gov, Ind]) and obligationType (mixed) agreed exactly — the only entry in the pack on which three of the four axes agreed. E narrows to [Dat]: what is transmitted is 'a summary of any assessment of catastrophic risk', which is assessment output, and v0.1's Gui rests on the framework that produced the assessment rather than on these words. The S agreement is the pack's most consequential textual divergence from California. In ca-sb53-tfaia v0.2 the analogous entry at Bus. & Prof. Code § 22757.12(a)(10) and (d) lost Gov although both passes there had declared it, under the rule that a government body named only as the recipient of a report does not thereby become a source to be trusted. Gov survives here because § 1422(2)(a) does not merely name the office as recipient: the alternative reporting cadence must be 'communicated in writing to the office with written updates, as appropriate, and agreed upon by the office', so the office holds a decisive position on the substance of the schedule the duty is measured by. Those seven words — 'and agreed upon by the office' — are the divergence, and California's § 22757.12(d) has no counterpart to them. The blind pass reached Gov on exactly that wording and flagged its own reading as inferential; it survives because both passes reached it and both grounded it in the text. Ind is retained under the Ind limb: § 1422(2)(a) names the large frontier developer as the party that produces and transmits the summary.", "retrievedAt": "2026-08-14", "verifiedBy": "dual formalization (blind second pass) + adjudication, 2026-08-14" }, "changeNote": "V, S and obligationType agreed exactly. E [Dat, Gui] → [Dat]. The S agreement is the pack's most consequential textual divergence from California: ca-sb53-tfaia v0.2 removed Gov from the analogous § 22757.12(a)(10)/(d) entry under the recipient-is-not-a-source rule, and Gov survives here because § 1422(2)(a) does not merely name the office as recipient — the alternative cadence must be 'agreed upon by the office', which puts the office in a decisive position on the substance of the schedule. Both passes reached Gov and both grounded it in those words." }, { "article": "Gen. Bus. Law § 1421(3)(a)–(c)", "summary": "Before or concurrently with deploying a new frontier model or a substantially modified version of one, a frontier developer must clearly and conspicuously publish a transparency report covering the developer's website, a mechanism for a natural person to reach it, the model's release date, supported languages and output modalities, intended uses and applicable use restrictions; a large frontier developer must additionally include summaries of the catastrophic-risk assessments conducted under its frontier AI framework, their results, the extent of third-party evaluator involvement, and other steps taken to fulfil the framework.", "v": [], "e": [ "Dat" ], "s": [ "Ind" ], "status": "draft-verified", "obligationType": "mixed", "note": "This is the provision that generates the public artefact most readers associate with the RAISE Act, and the one place where the article binds any frontier developer rather than only a large one. One limit matters: § 1421(3)(c) deems a developer in compliance if the information appears inside a larger document such as a system card or model card, so the statute prescribes content, not form. Unlike the California analogue, New York carries no companion clause encouraging disclosure beyond the enumerated list. An AIO 20002 record is not a transparency report and does not substitute for any element of the list. ADJUDICATION 2026-08-14: this entry came out of the dual formalization with an UNDECLARED VALUE LAYER — the two independent passes shared no value code at all, and neither reading is settled by the quoted words. An entry that states no normative direction cannot carry an item, so item authoring against § 1421(3) is deferred until the RFC round settles the layer. The California twin at Bus. & Prof. Code § 22757.12(c) kept `Bed` because both of its passes reached it; this pack's blind pass did not.", "provenance": { "sourceUrl": "https://www.nysenate.gov/legislation/bills/2025/S8828", "retrievalUrl": "https://nyassembly.gov/leg/?default_fld=&leg_video=&bn=S08828&term=2025&Text=Y", "article": "General Business Law § 1421(3)(a) chapeau, § 1421(3)(b) chapeau and subparagraphs (i)–(iii), and § 1421(3)(c)", "quote": "[§ 1421(3)(a)] 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 containing all of the following: […] [§ 1421(3)(b)] a large frontier developer shall include in the transparency report required by paragraph (a) of this subdivision, summaries of all of the following: (i) assessments of catastrophic risks from the frontier model conducted pursuant to the large frontier developer's frontier AI framework; (ii) the results of the assessments under subparagraph (i) of this paragraph; (iii) the extent to which third-party evaluators were involved;", "rationale": "The report's function is to give the reader the material to form their own view of a model rather than a conclusion to accept — the legislative findings state that 'Greater transparency can also advance accountability, competition, and public trust' — so the value that must prevail is Sdt, the addressee's own judgment. Publishing before or concurrently with deployment is a disclosure owed to whoever will encounter the model, which is Bed. Subparagraph (ii) requires the results of the developer's own catastrophic-risk assessments to be summarized, including results that do not flatter the model, which is Hum: stating one's limits without overstating. What discharges the duty is a prescribed content list (Gui) carrying assessment results (Dat). The report is written and published by the developer (Ind) under a statutory content requirement (Gov). No expert body is designated as author or reviewer of the report, so Pro and Pee are not assigned. ADJUDICATION 2026-08-14: obligationType (mixed) agreed exactly. The value layer is emptied, and it is the sharpest divergence in the pack: the two passes shared no value code at all. v0.1 read the transparency report as serving the reader's own judgment (Sdt), as a disclosure owed to whoever will encounter the model (Bed) and as a duty to summarize results that do not flatter the model (Hum); the blind pass read it as protecting persons and society through the catastrophic-risk contents the report must carry (Sep, Ses) and as putting that information before everyone alike rather than before a regulator (Unc). Nothing in the quoted words settles between them — v0.1's Sdt is argued from the legislative findings rather than from the excerpt, and the blind pass's Unc from the publication mode rather than from a named interest — so under the standing policy the more conservative outcome is taken and the layer is left undeclared. Note the contrast with the California twin at Bus. & Prof. Code § 22757.12(c): there the value layer survived as [Bed] because both passes reached Bed, and this pack's blind pass did not. E narrows to [Dat], the only evidence code both passes reached: the enumerated summaries are assessments and 'the results of the assessments'. v0.1's Gui (the prescribed content list) and the blind pass's Exp (third-party evaluator involvement) each rest on one pass. S narrows to [Ind]: the excerpt names 'a frontier developer' as the party that publishes, and Gov is absent from it. An entry with an undeclared value layer states no normative direction, so no item may be written against this entry until the RFC round settles the layer.", "retrievedAt": "2026-08-14", "verifiedBy": "dual formalization (blind second pass) + adjudication, 2026-08-14" }, "changeNote": "obligationType agreed exactly. V [Sdt, Bed, Hum] → []: the two passes shared no value code at all — v0.1 read the report as serving the reader's own judgment (Sdt), as a disclosure owed (Bed) and as a duty not to overstate (Hum); the blind pass read it as protecting persons and society through its catastrophic-risk contents (Sep, Ses) and as putting that information before everyone alike (Unc). Nothing in the quoted words settles between them, the layer is left undeclared, and the divergence goes to the RFC round. The California twin at § 22757.12(c) kept [Bed] only because both passes there reached it. E [Gui, Dat] → [Dat]; S [Ind, Gov] → [Ind] (the excerpt names the frontier developer as the publisher). No item may be written against an entry with an undeclared value layer." }, { "article": "Gen. Bus. Law § 1421(4)(a)–(b)", "summary": "A frontier developer must not make a materially false or misleading statement about catastrophic risk from its frontier models or about its management of catastrophic risk, and a large frontier developer must not make a materially false or misleading statement about its implementation of, or compliance with, its frontier AI framework; the prohibition does not reach a statement made in good faith that was reasonable under the circumstances.", "v": [ "Hum", "Bed" ], "e": [ "Dat" ], "s": [ "Ind" ], "status": "draft-verified", "obligationType": "organizational", "note": "Classified `behavioral` in v0.1 and lowered to `organizational` at the 2026-08-14 adjudication, on the same ground and by the same reasoning as the near-identical Bus. & Prof. Code § 22757.12(e) in the ca-sb53-tfaia pack: the statements the subdivision regulates are the developer's corporate disclosures about catastrophic risk and about its own published framework, not anything said in a concrete case by a system. The two provisions are the same norm enacted twice — the only wording differences are 'shall not apply' for California's 'does not apply' in the good-faith carve-out and the paragraph numbering — so the divergence v0.1 preserved between the two packs is closed rather than inherited, and the distribution v0.1 predicted for that outcome (organizational 3 / mixed 7 / behavioral 0) is the distribution this pack now carries. The judgment correlate v0.1 identified is still real and is the reason the entry is worth an item — a scenario can put reputational or commercial pressure behind an overstated safety claim and observe whether the statement is tempered — but a correlate an item can test is not the same as a duty an item can observe being discharged. Two boundaries: the prohibition covers statements about catastrophic risk and framework compliance only, not accuracy claims generally; and the § 1421(4)(b) good-faith carve-out means an honest and reasonable statement that turns out wrong is not a violation. Penalty exposure runs through § 1427(1), which names a statement in violation of this subdivision as one of its four enumerated grounds — but note the asymmetry recorded in the pack notes: § 1421(4)(a)(i) binds any frontier developer, while the operative clause of § 1427(1) is framed 'where a large frontier developer' does the enumerated things.", "provenance": { "sourceUrl": "https://www.nysenate.gov/legislation/bills/2025/S8828", "retrievalUrl": "https://nyassembly.gov/leg/?default_fld=&leg_video=&bn=S08828&term=2025&Text=Y", "article": "General Business Law § 1421(4)(a)(i)–(ii) and § 1421(4)(b)", "quote": "[§ 1421(4)(a)(i)] A frontier developer shall not make a materially false or misleading statement about catastrophic risk from its frontier models or its management of catastrophic risk. [§ 1421(4)(a)(ii)] A large frontier developer shall not make a materially false or misleading statement about its implementation of, or compliance with, its frontier AI framework. [§ 1421(4)(b)] This subdivision shall not apply to a statement that was made in good faith and was reasonable under the circumstances.", "rationale": "The prohibition is on overstatement, and the word 'misleading' reaches statements that are literally true but leave a false impression — the value that must prevail is therefore Hum, recognition and disclosure of one's own limits, over the reputational interest (Fac) that pushes the other way. Because the statements at issue are ones the developer has undertaken to make accurately about a framework it published itself, Bed also prevails. The interest the accuracy serves is the public-safety information environment the article is built to create (Ses); it is not, on these words, the safety of any identified person, so Sep is not assigned here. On evidence: limb (a)(i) is falsified against what the developer's own assessments actually record (Dat), while limb (b) makes the speaker's considered judgment at the time decisive — 'good faith' and 'reasonable under the circumstances' are assessed on the situation as it then stood (Exp). The regulated speaker is the developer itself (Ind); the standard of materiality and the enforcement action under § 1427(1) belong to the attorney general (Gov). ADJUDICATION 2026-08-14: V narrows to [Hum, Bed], the blind pass's set and a proper subset of v0.1's — the prohibition is a duty not to overstate (Hum) about statements the developer has undertaken to make accurately (Bed), while Ses rests on v0.1 alone. E narrows to [Dat]: both passes reached it, v0.1 as what the developer's own assessments actually record and the blind pass as what is actually known of catastrophic risk. v0.1's Exp and the blind pass's Gui each rest on one pass, and the California twin's finding that 'made in good faith and was reasonable under the circumstances' is a legal standard of care rather than an evidentiary class is adopted here too, which is why Exp does not survive. The evidence layer of the California twin was emptied only because its blind pass found no evidentiary class at all; this pack's blind pass affirmatively found one that v0.1 had also found, so the intersection is non-empty and is taken. That difference is a difference in what the passes reached, not in what the two texts say. S narrows to [Ind]: the developer is the author of the statement the provision regulates, and Gov falls because the attorney general's penalty power under § 1427(1) is outside this excerpt. obligationType is lowered from behavioral to organizational, and this is the most consequential single change in the pack. Both passes tagged the entry behavioral, and it is lowered anyway, for three reasons stated together. (i) CONTAMINATED AXIS. This is the one axis of this pack on which the v0.1 pass was not blind: it had read the ca-sb53-tfaia adjudication lowering the near-identical Bus. & Prof. Code § 22757.12(e) to organizational, dissented from it deliberately, and recorded the contamination in v0.1 note 15. Wave 1 practice is that a disclosed contamination is processed at adjudication on the axis it touches, so the apparent agreement here is between one contaminated pass and one blind pass, not between two independent readings. (ii) THE TEXTS DO NOT DIVERGE. Set side by side, § 1421(4)(a)(i)–(ii) and § 22757.12(e)(1)(A)–(B) are word for word the same two sentences; the only differences anywhere in the provision are 'This subdivision shall not apply' for California's 'This subdivision does not apply' in the good-faith carve-out, and the paragraph numbering. Neither difference bears on who speaks, on what is spoken about, or on when the duty is discharged. The words that decided the California adjudication are present here unchanged: what § 1421(4)(a)(ii) regulates is a statement 'about its implementation of, or compliance with, its frontier AI framework', which is a developer's corporate disclosure about a document it published, not something said in a concrete case by a system. Same words, same classification. (iii) THE CONSERVATIVE TAG. The standing rule of this wave is that a contested obligationType resolves to the more conservative tag, organizational over mixed over behavioral. The blind pass's independent behavioral reading is not discarded: it is recorded as an RFC item, because it is the strongest surviving argument that the obligationType vocabulary cuts this provision at the wrong joint — the duty really is discharged in the act of speaking, and what the tag records is that the speaker the statute addresses is a corporation making a disclosure rather than a model answering in a case.", "retrievedAt": "2026-08-14", "verifiedBy": "dual formalization (blind second pass) + adjudication, 2026-08-14" }, "changeNote": "V [Hum, Bed, Ses] → [Hum, Bed]; E [Dat, Exp] → [Dat] (both passes reached Dat; the California twin emptied this layer only because its blind pass found no evidentiary class at all, so the two packs differ in what the passes reached, not in what the words say); S [Ind, Gov] → [Ind]. **obligationType behavioral → organizational.** Both passes tagged this entry behavioral and it is lowered anyway: this is the one axis on which the v0.1 pass was NOT blind — it read the ca-sb53-tfaia adjudication of the near-identical Bus. & Prof. Code § 22757.12(e) and dissented deliberately, recording the contamination in v0.1 note 15 — so the agreement is between a contaminated pass and one blind pass. On the text there is nothing to distinguish the two provisions: the only wording differences are 'shall not apply' for California's 'does not apply' in the good-faith carve-out and the paragraph numbering, and what is regulated in both is a statement 'about its implementation of, or compliance with, its frontier AI framework' — a corporate disclosure, not something said in a concrete case by a system. Same words, same classification. The blind pass's independent behavioral reading is carried to the RFC round rather than discarded." }, { "article": "Gen. Bus. Law § 1421(5)(a)–(b)", "summary": "A frontier developer publishing documents under § 1421 may redact only what is necessary to protect its trade secrets or cybersecurity, public safety, or United States national security, or to comply with federal or state law; where it redacts, it must describe the character and justification of the redaction in the published version to the extent the justifying concerns permit, and must retain the unredacted information for five years.", "v": [ "Ses" ], "e": [ "Gui" ], "s": [ "Ind" ], "status": "draft-verified", "obligationType": "mixed", "note": "The provision is the pressure point of the whole publication regime: it is where a developer decides how much of a risk assessment the public actually sees. The statute answers with a necessity test and a meta-disclosure duty — the fact and reason of the redaction must themselves be published. The five-year retention duty is organizational and unobservable by any item. Redaction breadth is not measurable from an AIO 20002 record either; what an item can test is whether a commercial-sensitivity rationale is accepted as sufficient where the enumerated grounds do not apply.", "provenance": { "sourceUrl": "https://www.nysenate.gov/legislation/bills/2025/S8828", "retrievalUrl": "https://nyassembly.gov/leg/?default_fld=&leg_video=&bn=S08828&term=2025&Text=Y", "article": "General Business Law § 1421(5)(a) and (5)(b)", "quote": "[§ 1421(5)(a)] When a frontier developer publishes documents to comply with this section, such frontier developer may make redactions to such documents that are necessary to protect such frontier developer's trade secrets, such frontier developer's cybersecurity, public safety, or the national security of the United States or to comply with any federal or state law. [§ 1421(5)(b)] such frontier developer shall describe the character and justification of such redaction in any published version of such document […] and shall retain the unredacted information for five years.", "rationale": "The redaction power is bounded by 'necessary', and paragraph (b) requires the reader to be told that something was removed and why — so what the provision preserves, at the margin, is the reader's ability to judge the document they are given (Sdt) and the developer's dependability toward the audience it published to (Bed). Two of the four enumerated grounds are cybersecurity and public safety, which is Ses; the commercial interest (trade secrets) appears as a permitted ground, not as a value the provision protects, so no Power code is assigned as prevailing. On evidence, the enumeration of permitted grounds is a closed written test applied as written (Gui), and paragraph (b) requires a stated justification rather than a measurement — an argument from the enumerated premises to this redaction (Log). Log is an INFERENCE from the word 'justification' rather than a class the text names, and is put to the RFC round. The redacting party is the developer (Ind) under a statutory standard (Gov). ADJUDICATION 2026-08-14: obligationType (mixed) agreed exactly. V narrows to [Ses], the intersection and the only value code both passes reached — two of the four enumerated redaction grounds are 'cybersecurity, public safety' and 'the national security of the United States', which is societal stability. v0.1's Sdt and Bed rest on the meta-disclosure duty in paragraph (b) rather than on a named interest, and the blind pass's Por for trade secrets is not carried: the value layer cannot represent an interest preserved AGAINST the duty rather than served by it, which is Wave 1 gap 9 and is now confirmed in a third instrument. The blind pass's Sep and Cor likewise rest on one pass. E narrows to [Gui]: the enumeration of permitted grounds is a closed written test applied as written, and Log does not survive — v0.1 flagged it as an INFERENCE from the word 'justification' and the blind pass did not reach it. S narrows to [Ind]: the redacting party is the developer, and Gov falls because 'any federal or state law' appears only as one permissive ground the developer may invoke, not as an authority whose position decides the matter. Every layer resolves exactly as the near-identical Bus. & Prof. Code § 22757.12(f) resolved in ca-sb53-tfaia v0.2.", "retrievedAt": "2026-08-14", "verifiedBy": "dual formalization (blind second pass) + adjudication, 2026-08-14" }, "changeNote": "obligationType agreed exactly. V [Sdt, Bed, Ses] → [Ses], the intersection; the blind pass's Por for trade secrets is not carried, because the value layer cannot represent an interest preserved against a duty rather than served by it (Wave 1 gap 9, confirmed in a third instrument). E [Gui, Log] → [Gui]: Log was flagged INFERENCE by v0.1 and the blind pass did not reach it. S [Ind, Gov] → [Ind] ('any federal or state law' is a permissive ground the developer may invoke, not an authority whose position decides the matter). Every layer resolves exactly as the near-identical Bus. & Prof. Code § 22757.12(f) resolved in ca-sb53-tfaia v0.2." }, { "article": "Gen. Bus. Law § 1422(3)(a)–(b)", "summary": "A frontier developer must report any critical safety incident involving one or more of its frontier models to the oversight office within seventy-two hours of determining that one has occurred, or within seventy-two hours of learning facts sufficient to establish a reasonable belief that one has occurred; and where it discovers that a critical safety incident poses an imminent risk of death or serious physical injury, it must disclose that incident within twenty-four hours to an appropriate authority, including any law enforcement or public safety agency with jurisdiction.", "v": [ "Sep", "Ses" ], "e": [ "Exp" ], "s": [ "Ind" ], "status": "draft-verified", "obligationType": "mixed", "note": "Two clocks, two thresholds, two recipients. The seventy-two-hour clock is the sharpest divergence from the California analogue, which allows fifteen days: New York's clock runs from a determination that a critical safety incident occurred OR from the developer learning facts sufficient to establish a reasonable belief that one occurred, whichever it is — so the trigger is knowledge-based and does not wait for confirmation. The twenty-four-hour clock runs to a jurisdictionally appropriate authority only where imminent risk of death or serious physical injury is found. The measurable judgment is the classification decision that starts either clock, not the filing. 'Critical safety incident' is defined at § 1420(4) in four limbs, the fourth of which — a model using deceptive techniques against its developer to subvert controls or monitoring outside an evaluation designed to elicit that behavior — has no counterpart in most incident taxonomies and is the limb most likely to be missed.", "provenance": { "sourceUrl": "https://www.nysenate.gov/legislation/bills/2025/S8828", "retrievalUrl": "https://nyassembly.gov/leg/?default_fld=&leg_video=&bn=S08828&term=2025&Text=Y", "article": "General Business Law § 1422(3)(a) and (3)(b)", "quote": "[§ 1422(3)(a)] Subject to paragraph (b) of this subdivision, a frontier developer shall report any critical safety incident pertaining to one or more of its frontier models to the office within seventy-two hours from a determination that a critical safety incident has occurred or within seventy-two hours of the frontier developer learning facts sufficient to establish a reasonable belief that a critical safety incident has occurred. [§ 1422(3)(b)] If a frontier developer discovers that a critical safety incident poses an imminent risk of death or serious physical injury, the frontier developer shall disclose that incident within twenty-four hours to an authority, including any law enforcement agency or public safety agency with jurisdiction […]", "rationale": "Paragraph (b) states its own protected interest in terms of 'imminent risk of death or serious physical injury' (Sep), while the reporting regime as a whole exists so that authorities can respond to emerging public-safety risk — the legislative findings say timely reporting 'is essential to ensure that public authorities are promptly informed of ongoing and emerging risks to public safety' (Ses). A fixed period measured from determination or from acquiring a reasonable belief is discharged by keeping to it, not by weighing whether reporting is warranted, which is Cor. On evidence: whether an event is a critical safety incident at all is settled by applying the closed four-limb definition at § 1420(4) as written (Gui), whereas the 'reasonable belief' finding that starts the seventy-two-hour clock and the imminence finding that starts the twenty-four-hour clock are judgments about the situation in front of the discoverer (Exp) — no statistic or precedent is named as decisive for either. The sources the text designates are the office and any law enforcement or public safety agency with jurisdiction (Gov), with the report itself originating from the developer (Ind). Members of the public may also report under § 1422(1), but the reporting duty in subdivision three is the developer's alone, so no testimony or stakeholder class is assigned here. ADJUDICATION 2026-08-14: V narrows to [Sep, Ses]. v0.1's Cor, read from the fixed periods, rests on one pass. The ordering is carried as both passes wrote it, Sep first, and here the ordering claim is textually supported rather than conventional: the seventy-two-hour default compresses to twenty-four hours precisely where the incident 'poses an imminent risk of death or serious physical injury', so the quoted words themselves put bodily safety of persons above the general interest in incident visibility. E narrows to [Exp], the intersection: what starts either clock is 'a determination that a critical safety incident has occurred' and the discovery that one 'poses an imminent risk', both of which are judgments about the situation in front of the discoverer. v0.1's Gui (applying the closed four-limb definition at § 1420(4)) and the blind pass's Log (inference from facts short of certainty) each rest on one pass. The layer is not emptied as it was in the California twin, where the two passes shared no evidence code at all, but the strain the twin recorded is the same and Wave 2 gap 18 — incident reporting as an evidence class — is confirmed here. S narrows to [Ind] under the recipient-is-not-a-source rule, which the blind pass raised on its own initiative for exactly this entry: 'the office' and 'any law enforcement agency or public safety agency with jurisdiction' appear in this excerpt solely as the recipients of the disclosure, and a recipient is not thereby a source to be trusted. The developer, who makes the reportable determination and files the report, is retained as Ind. obligationType is contested — v0.1 mixed, the blind pass behavioral — and resolves to mixed under the standing rule that a divergent tag takes the more conservative value.", "retrievedAt": "2026-08-14", "verifiedBy": "dual formalization (blind second pass) + adjudication, 2026-08-14" }, "changeNote": "V [Sep, Ses, Cor] → [Sep, Ses]; both passes placed Sep first and the ordering is carried, because the quoted text does rank here — the seventy-two-hour default compresses to twenty-four hours precisely where the incident 'poses an imminent risk of death or serious physical injury'. E [Gui, Exp] and [Exp, Log] → [Exp]: what starts either clock is a determination made about the situation in front of the discoverer. The layer is not emptied as it was in the California twin, but the same strain is recorded (Wave 2 gap 18). S [Gov, Ind] → [Ind] under the recipient-is-not-a-source rule, which the blind pass raised on its own initiative for this entry: the office and 'any law enforcement agency or public safety agency with jurisdiction' appear solely as recipients of the disclosure. obligationType mixed / behavioral → mixed, under the standing rule that a divergent tag resolves to the more conservative one." }, { "article": "Gen. Bus. Law § 1422(8), (9)(a)–(b)", "summary": "The oversight office may designate by regulation federal laws, regulations or guidance documents whose critical-safety-incident reporting standards are substantially equivalent to or stricter than § 1422(3); a frontier developer that intends to discharge § 1422(3) through such a designated federal instrument must declare that intent to the office, is then deemed in compliance only to the extent it actually meets that instrument, commits a violation of the article if it fails to meet it, and must still send the office copies of any federally required critical safety incident reports concurrently with sending them to federal authorities.", "v": [ "Cor" ], "e": [ "Gui" ], "s": [ "Gov" ], "status": "draft-verified", "obligationType": "mixed", "note": "A substitution route, not an exemption, and the entry is included because it is the provision most likely to be over-read. Three limits are express: the deeming operates only 'to the extent that' the developer meets the federal standard; failure to meet that standard is itself a violation of Article 44-B, so electing the federal route converts a federal shortfall into state liability; and the concurrent-copy duty at § 1422(9)(b)(iii) means the office still receives the reports. As at 2026-08-14 the office had designated no federal instrument, because Article 44-B is not yet in force — the route is therefore latent, and any operator relying on it should confirm the designation exists. The judgment correlate is whether a declared federal route is treated as a floor that still runs to the office, or as a general dispensation from New York reporting.", "provenance": { "sourceUrl": "https://www.nysenate.gov/legislation/bills/2025/S8828", "retrievalUrl": "https://nyassembly.gov/leg/?default_fld=&leg_video=&bn=S08828&term=2025&Text=Y", "article": "General Business Law § 1422(8)(a)(i), § 1422(9)(a) and § 1422(9)(b)(ii)–(iii)", "quote": "[§ 1422(8)(a)(i)] the law, regulation, or guidance document imposes or states standards or requirements for critical safety incident reporting that are substantially equivalent to, or stricter than, those required by subdivision three of this section; [§ 1422(9)(a)] A frontier developer that intends to comply with subdivision three of this section by complying with the requirements of, or meeting the standards stated by, a federal law, regulation, or guidance document designated pursuant to subdivision eight of this section shall declare its intent to do so to the office. [§ 1422(9)(b)(iii)] frontier developers who comply with subdivision three of this section by meeting such federal standards shall send copies of any critical safety incident reports required by such federal standards to the office concurrently with sending them to federal authorities.", "rationale": "What the provision demands is adherence to a designated external instrument on the terms on which it was designated — rule-following against a fixed standard, which is Cor. The declaration of intent under paragraph (9)(a) is an undertaking the developer is thereafter held to, and § 1422(9)(b)(ii) prices departure from it as a violation of the article, so Bed prevails. The interest the whole substitution machinery serves is the same public-safety reporting environment the article exists to create (Ses); no identified individual's safety is named in these words, so Sep is not assigned. On evidence, the decisive thing is the designated federal law, regulation or guidance document itself — an established written standard procedure, therefore Gui, and nothing else is named: the text nominates no metric, no case comparison and no expert. On sources, both the office that designates and the federal instrument designated are governmental (Gov), and the declaration and the concurrent copies originate with the developer (Ind). No professional body or scholarly source appears. ADJUDICATION 2026-08-14: E ([Gui]) and obligationType (mixed) agreed exactly — the decisive thing is the designated 'federal law, regulation, or guidance document' itself, an established written standard procedure, and nothing else is named. V narrows to [Cor], the intersection: what the provision demands is adherence to a designated external instrument on the terms on which it was designated. v0.1's Bed and Ses rest on one pass, and the blind pass's Sep is not carried — the blind pass argued a text-ranked hierarchy from 'substantially equivalent to, or stricter than', but those words rank two regulatory instruments against each other and designate no protected interest, so the ranking claim does not survive its own premise. S narrows to [Gov], and the ground is the substance limb rather than the recipient limb: the designated federal instrument is not a recipient of anything, it is the author of the governing position the developer elects to be measured against, and § 1422(8)(a)(i) makes its stated standards the yardstick. The office that designates it under subdivision eight is governmental on the same limb. Ind falls because the developer authors only its election — 'shall declare its intent to do so to the office' — and not the standard whose content decides the matter; the concurrent copies it sends the office are a recipient relationship and earn nothing. This is one of the two entries in the pack whose source layer is Gov alone.", "retrievedAt": "2026-08-14", "verifiedBy": "dual formalization (blind second pass) + adjudication, 2026-08-14" }, "changeNote": "E and obligationType agreed exactly. V [Cor, Bed, Ses] → [Cor], the intersection; the blind pass's Sep is not carried — 'substantially equivalent to, or stricter than' ranks two instruments against each other and designates no protected interest, so the text-ranked hierarchy the blind pass claimed does not survive its own premise. S [Gov, Ind] → [Gov]: both passes declared Gov and the ground is the substance limb, not the recipient limb — the designated federal instrument is the author of the governing position the developer elects to be measured against. Ind falls because the developer authors only its election, not the standard's content." }, { "article": "Gen. Bus. Law § 1428(1), (3), (5)", "summary": "No large frontier developer may develop, deploy or operate a frontier model in whole or in part in New York State without a current disclosure statement filed with the oversight office and payment of its pro rata share of the office's operating expenses; the statement must identify the developer and its trading names, its principal place of business and every New York office, its five-percent (or, if publicly traded, fifty-percent) beneficial owners, and three named points of contact, and must be renewed every two years or on transfer of ownership or material change; failure to file, submission of false information, or failure to pay carries a civil penalty of one thousand dollars for each day of default plus the assessments owed.", "v": [ "Cor" ], "e": [ "Gui" ], "s": [ "Gov" ], "status": "draft-verified", "obligationType": "organizational", "note": "This section has no counterpart in the California pack and is the structural novelty of the New York statute: it converts frontier development in the state into a filing-gated, industry-funded activity. The duty is discharged by filing a form and paying an assessment — nothing an item can observe. It is mapped because operators reading the two packs side by side will otherwise miss that New York adds a registration precondition California does not have, and because § 1428(6) requires the office to publish the list of filers, which makes non-filing publicly visible. The thin behavioral residue sits in § 1428(5) ('submits false information in its disclosure'), and it is a truthfulness duty about corporate identity and ownership rather than about model risk — it is not the same norm as § 1421(4) and should not be conflated with it. Note also the drafting seam at § 1428(5), which speaks of a person who 'develops, deploys, or operates a large frontier model' where the defined term at § 1420 is a large frontier DEVELOPER of a frontier model; the pack records the discrepancy without resolving it.", "provenance": { "sourceUrl": "https://www.nysenate.gov/legislation/bills/2025/S8828", "retrievalUrl": "https://nyassembly.gov/leg/?default_fld=&leg_video=&bn=S08828&term=2025&Text=Y", "article": "General Business Law § 1428(1), § 1428(2), § 1428(4) and § 1428(5)(a)", "quote": "[§ 1428(1)] Except as otherwise provided in this section, no large frontier developer may develop, deploy, or operate a frontier model, in whole or in part in New York state, without having a current disclosure statement filed with the office and paying the required share. [§ 1428(2)] The disclosure statement shall be filed in the form and the manner prescribed by the office and shall contain all the information required by the office. It shall be renewed every two years […] [§ 1428(4)] Large frontier developers shall be assessed in pro rata shares by the department to defray the operating expenses […] of administering the obligations imposed by this article. [§ 1428(5)(a)] a civil penalty of one thousand dollars for each day the entity fails to file a disclosure as required by this section or fails to correct false information;", "rationale": "The section is a precondition on lawful activity — 'no large frontier developer may develop, deploy, or operate a frontier model […] without having a current disclosure statement filed' — and a precondition is discharged by doing what the rule says, which is Cor. The named points of contact under § 1428(3)(d) are required to be 'responsible for receiving inquiries relating to this article from the office or other governmental entities', which is an undertaking to remain answerable to the regulator, therefore Bed. The interest served is the state's capacity to see and fund oversight of who is operating frontier models in its jurisdiction, which is societal order rather than any individual's safety (Ses); Sep is not assigned because these words protect no identified person. On evidence, the statement is filed 'in the form and the manner prescribed by the office' and contains 'all the information required by the office' — an established written standard procedure and nothing else, therefore Gui alone; no assessment record, expert view or measurement is decisive. On sources, the office and the department prescribe the form, levy the assessment and hold the penalty power (Gov), and the developer authors the statement (Ind). No professional body or scholarly source appears. ADJUDICATION 2026-08-14: obligationType (organizational) agreed exactly. V narrows to [Cor], the intersection, and the ranking claim the blind pass made for it is accepted because v0.1's own quote supports it: 'no large frontier developer may develop, deploy, or operate a frontier model, in whole or in part in New York state, without having a current disclosure statement filed' makes rule compliance a precondition of operating at all. v0.1's Bed and Ses rest on one pass; the blind pass's Por, read from the pro rata assessment limb, is not carried, because the funding of the regulator is not an interest the provision makes prevail. E narrows to [Gui]: the statement is filed 'in the form and the manner prescribed by the office' and contains 'all the information required by the office', which is a prescribed written content list and nothing else; the blind pass's Dat for the computation of each pro rata share rests on one pass. S narrows to [Gov], and Ind is removed although v0.1 declared it. The office is not a mere recipient here — it prescribes the form, fixes the content and renews the cycle, and the department levies the assessment, all of which is the substance limb — while the filer's own statement is expressly not treated as trustworthy: § 1428(5)(a) prices the failure 'to correct false information' as a daily civil penalty, which is the statute saying in terms that what the filer submits is checked rather than believed. This is one of the two entries in the pack whose source layer is Gov alone.", "retrievedAt": "2026-08-14", "verifiedBy": "dual formalization (blind second pass) + adjudication, 2026-08-14" }, "changeNote": "obligationType agreed exactly. V [Cor, Bed, Ses] → [Cor], the intersection; the blind pass's Por, read from the pro rata assessment limb, is not carried, because the funding of the regulator is not an interest the provision makes prevail. E [Gui, Dat] → [Gui]. S [Gov, Ind] → [Gov]: the office prescribes the form, fixes the content and renews the cycle and the department levies the assessment, which is the substance limb, while Ind falls because § 1428(5)(a) prices failure 'to correct false information' as a penalty — the statute saying in terms that the filer's own statement is not the trusted source." } ], "itemBankRef": { "publicSet": null, "privateSet": null }, "version": "0.2", "supersedes": "0.1", "status": "draft-verified", "updatedAt": "2026-08-14", "measurementScope": "AIO items measure model judgment alignment with each provision's normative direction. They do not assess whether an organization publishes or adheres to a frontier AI framework, publishes transparency reports, transmits internal-use risk assessments, meets the statutory incident-reporting clocks, files a disclosure statement with the oversight office or pays its assessment, or is a frontier developer or large frontier developer at all. Nothing measured against this pack bears on liability under General Business Law § 1427 or § 1428(5). After the dual formalization and adjudication of 2026-08-14, three of the ten mapped provisions are `organizational` and seven are `mixed`; none is `behavioral`, the v0.1 behavioral classification of § 1421(4) having been lowered to `organizational` for consistency with the identical California provision.", "notes": [ "draft-verified, not active. Every vesMapping entry carries a verbatim excerpt of the enacted text and a rationale argued from it, and on 2026-08-14 the second independent formalization required by FORMALIZATION_METHODOLOGY.md §5 was completed blind and adjudicated. The second formalizer read only the pack id, the sourceNorm and each entry's provenance.article, sourceUrl, retrievalUrl and quote; the v/e/s arrays, summaries, rationales, obligationType tags and notes of v0.1 were stripped by an extraction script before any file was opened, and neither the pack-authoring guideline nor the management guide was opened. One protocol deviation was disclosed by the second formalizer and is recorded in the contamination notice below. Human review is still outstanding, and the V/E/S assignment is settled only by the public RFC process at https://aioq.org/en/rfc. A certificate issued against this pack carries a draft-basis notice and records this status as `basisStatus` in its signed payload. No certificate may be issued against this pack in any case, because no item bank exists for it (see below).", "LEGISLATIVE STATUS — the RAISE Act is law, but the law now in the books is not the bill the Legislature passed in June 2025, and it is not yet in force. The sequence, verified from official legislative records on 2026-08-14: (i) S. 6953-B / A. 6453-B passed the Senate 58-1 on June 12, 2025 and was signed by Governor Hochul on December 19, 2025 as Chapter 699 of the Laws of 2025, adding General Business Law Article 44-B, to take effect on the ninetieth day after becoming law; (ii) the signature was given on the condition that agreed chapter amendments follow, and those amendments were introduced on January 6, 2026 as A. 9449 (Bores) with Senate companion S. 8828 (Gounardes); (iii) S. 8828 passed the Senate 58-1 on January 28, 2026, was substituted for A. 9449 on February 25, 2026, passed the Assembly on March 11, 2026, was delivered to the Governor on March 20, 2026 and was SIGNED on March 27, 2026 as CHAPTER 96 of the Laws of 2026; (iv) section 2 of Chapter 96 REPEALED Article 44-B as added by Chapter 699 in its entirety and added a new Article 44-B (§§ 1420–1429) in its place, and section 3 amended the effective-date section of Chapter 699 to read 'January 1, 2027' in place of 'on the ninetieth day after it shall have become a law'. As at 2026-08-14 Article 44-B is enacted but NOT YET IN FORCE, and the office within the Department of Financial Services that administers it has necessarily designated no federal instrument under § 1422(8) and prescribed no disclosure-statement form under § 1428(2). This pack formalizes the Chapter 96 text, which is the text that will be in force on January 1, 2027.", "WHAT THE CHAPTER AMENDMENT CHANGED — read this before relying on any pre-2026 description of the RAISE Act. Chapter 96 did not edit Chapter 699; it replaced the article wholesale, and four widely reported features of the June 2025 bill are NOT in the law. (1) NO DEPLOYMENT PROHIBITION. S. 6953-B barred deploying a frontier model that created an unreasonable risk of critical harm. The enacted Article 44-B contains no such prohibition; its closest analogue, § 1421(1)(d), requires only that the framework describe how deployment-decision review is conducted. (2) DIFFERENT THRESHOLDS. The June 2025 bill keyed coverage to a compute cost exceeding one hundred million dollars. The enacted § 1420(8)–(9) keys 'frontier developer' to training compute greater than 10^26 integer or floating-point operations with no cost limb at all, and § 1420(10) keys 'large frontier developer' to affiliate-consolidated annual gross revenues exceeding five hundred million dollars in the preceding calendar year. (3) DIFFERENT PENALTIES. Figures of ten and thirty million dollars circulated widely from the June 2025 bill; the enacted § 1427(1) caps the attorney general's civil penalty at one million dollars for a first violation and three million dollars per subsequent violation, with a separate one-thousand-dollar-per-day penalty under § 1428(5)(a). (4) NEW OVERSIGHT ARCHITECTURE. The enacted text creates an 'office' within the Department of Financial Services (§ 1420(16)) as the reporting counterparty, with rulemaking authority under § 1429 and an annual public report from January 1, 2028 under § 1422(7) — the June 2025 bill routed reporting through the attorney general and the division of homeland security and emergency services. Chapter 96 also added the § 1428 disclosure-statement and pro rata assessment regime, which had no counterpart in the June 2025 bill. Any pack, guide, crosswalk or commentary describing the RAISE Act from the June 2025 bill text is describing a statute that was repealed before it ever took effect.", "Primary source and retrieval. The enacted text was retrieved on 2026-08-14 from the New York State Assembly bill-text service in two independent passes: once for A. 9449 (https://nyassembly.gov/leg/?default_fld=&leg_video=&bn=A09449&term=2025&Text=Y) and once for S. 8828, the bill actually signed (https://nyassembly.gov/leg/?default_fld=&leg_video=&bn=S08828&term=2025&Text=Y). After identical normalization the two documents were byte-identical (SHA-256 of the normalized corpus agreed exactly, 27,668 characters), which is the cross-check this pack uses in place of the two-extraction-path method used for the two-column NIST PDF. Chapter number, signing date and the full action history were confirmed against the Assembly's actions record for S08828 and the New York State Senate bill page (https://www.nysenate.gov/legislation/bills/2025/S8828). No commentary, law-firm summary, trade-press explainer or mirror site was used as a source for any quoted text. New York statutes are government edicts and are not subject to copyright (Georgia v. Public.Resource.Org, Inc., 590 U.S. 255 (2020)), so verbatim quotation carries no licensing constraint; excerpts are nonetheless kept short and cited to the section.", "Quote verification method. The signed S. 8828 text was normalized to a single-spaced corpus of 27,561 characters and every quoted fragment across the ten entries was matched as an exact substring of that corpus — 25 of 25 provenance-quote fragments matched, counting each excerpt between bracketed citations and elisions as one fragment. The same check was run over the shorter excerpts quoted inside the entry rationales, entry notes and these pack notes: 51 of 51 spans presented as New York statutory wording matched the corpus. Exactly two quoted phrases in this pack do not appear in the corpus and are not meant to — 'approaches' and 'industry-consensus best practices', which are the CALIFORNIA wording at Bus. & Prof. Code § 22757.12(a) quoted for contrast, checked against the ca-sb53-tfaia pack's own verified excerpt and attributed to California in situ. Two normalizations were applied to the corpus and are disclosed because they affect what 'verbatim' means here. (i) The Legislative Bill Drafting Commission's fixed-width text breaks words across lines with a hyphen; all 42 distinct line-break hyphenations were enumerated and rejoined, and the single genuine compound among them ('machine-based', at § 1420(2)) was preserved as hyphenated. (ii) The official plain text encodes the exponent in the compute threshold as a single non-ASCII superscript marker byte (rendered as a degree sign by the Senate's front end and as a section sign by the Assembly's); it is normalized to a caret, so this pack writes the threshold as '10^26'. Unlike the California chaptered text, the New York bill text uses straight ASCII apostrophes throughout, and the quotes reproduce them. Elisions are marked […] and bracketed citations such as [§ 1421(1)] precede excerpts where one entry quotes more than one provision; everything outside brackets is verbatim. The same corpus was used at the v0.2 adjudication: every quote, article citation, sourceUrl and retrievalUrl in this file is carried over from v0.1 byte for byte, and the adjudication changed only the v/e/s arrays, the obligationType tags, the rationale (by appending an ADJUDICATION paragraph), the entry statuses, the verifiedBy fields and the notes. No provision was added, removed or re-quoted.", "Scope thresholds — who is actually covered. This norm does not apply to AI developers generally, and the coverage line is narrower than the roster's 'jurisdiction' column suggests. A 'frontier model' under § 1420(9) is a foundation model trained using more than 10^26 integer or floating-point operations, counting the original training run plus any subsequent fine-tuning, reinforcement learning or other material modification. A 'frontier developer' under § 1420(8) is a person who has trained or initiated the training of such a model. A 'large frontier developer' under § 1420(10) is a frontier developer whose annual gross revenues, together with its affiliates, exceeded five hundred million dollars in the preceding calendar year. Seven of the ten mapped entries bind large frontier developers only — the entire frontier AI framework regime at § 1421(1)–(2), the large-developer limb of the transparency report at § 1421(3)(b), the quarterly internal-use transmission at § 1422(2)(a), and the disclosure statement at § 1428. Three bind any frontier developer: the base transparency report at § 1421(3)(a), the statement prohibition at § 1421(4)(a)(i), the redaction discipline at § 1421(5) and the incident clocks at § 1422(3). § 1425 confines the whole article to frontier models 'developed, deployed, or operating in whole or in part in New York state'. § 1426 exempts accredited New York colleges and universities to the extent they are engaged in academic research on AI models, and the Empire AI consortium and institute as defined in Economic Development Law § 361 — an exemption structure California's statute does not have. Deployers, downstream integrators and users of frontier models have no obligations under this article.", "Relationship to the ca-sb53-tfaia pack — the same statute enacted twice, with four substantive divergences. The chapter amendment was negotiated expressly to align New York with California, and the alignment is close enough that §§ 1421(1)(a)–(j), 1421(2)–(5) and 1422(3) track Bus. & Prof. Code §§ 22757.12–22757.13 clause by clause; § 1421(4) and Bus. & Prof. Code § 22757.12(e) are near-identical. Operators covered by both should expect one framework and one transparency report to serve both, subject to counsel. The divergences that matter: (1) INCIDENT CLOCK — New York requires reporting within seventy-two hours of a determination or of acquiring a reasonable belief (§ 1422(3)(a)); California allows fifteen days from discovery. New York's is the binding constraint for a dual-covered developer. (2) REGISTRATION — New York § 1428 requires a biennial disclosure statement filed with the office, beneficial-ownership disclosure, three named contacts and a pro rata assessment funding the regulator; California has no analogue. (3) NO WHISTLEBLOWER CHAPTER — California SB 53 added Labor Code §§ 1107–1107.2, protecting covered-employee disclosures and requiring large frontier developers to run an anonymous internal channel; the RAISE Act contains nothing equivalent, so entries 9 and 10 of the ca-sb53-tfaia pack have no counterpart here and this pack assigns Ano nowhere. (4) REGULATOR AND RULEMAKING — New York vests administration in an office within the Department of Financial Services with express rulemaking authority under § 1429 and an annual public report from 2028 under § 1422(7); California routes reporting to the Office of Emergency Services and gives the Department of Technology an annual definitional reassessment under § 22757.14 instead. The two packs are not crosswalked entry to entry and a model aligned with one is not thereby aligned with the other. After both packs completed dual formalization, the adjudicated mappings can be compared directly, and they track the four divergences above rather than cutting across them. Seven of the ten New York entries resolve to exactly the codes their California counterparts resolved to. Three do not, and each difference is recorded at the entry: § 1421(1)(j) with § 1422(2)(a) keeps `Gov` where California's § 22757.12(a)(10) with (d) lost it, because New York requires an alternative reporting cadence to be 'agreed upon by the office' and California's subdivision has no counterpart to those words; § 1421(1)(d)–(e) loses `Hum` where California's § 22757.12(a)(4)–(5) kept it, because there both independent passes reached the inference and here only one did; and § 1421(4) keeps `Dat` in the evidence layer where California's § 22757.12(e) was emptied, again because of what the passes reached rather than because of any difference in the words. § 1428 and § 1422(8)–(9) have no California counterpart at all. The obligationType tags of the two packs are not aligned entry for entry and are not meant to be: California's adjudication lowered five entries from `mixed` to `organizational` on divergences between its own two passes, while both New York passes independently tagged the analogous entries `mixed`, and importing California's tags would substitute a third pass for this pack's evidence. That residue is recorded as an RFC agenda item in the gaps note below. The one place where cross-pack consistency was enforced against a two-pass agreement is § 1421(4), for the reasons set out in the entry and in the resolution note below.", "Relationship to the eu-ai-act pack — a different construct, not a stricter or looser version of it. The EU AI Act regulates AI systems by risk class and imposes conformity assessment, technical documentation and post-market monitoring on providers and deployers. The RAISE Act regulates a small number of very large model developers by compute and revenue threshold, imposes no conformity assessment or ex ante approval of any kind, and works through published disclosure, incident reporting to a state office, a filing precondition, and truthfulness about what has been disclosed. A model aligned with one pack is not thereby aligned with the other, and the two packs' provisions are not crosswalked.", "State-law reach and its limits. The RAISE Act is New York law and § 1425 confines it to frontier models developed, deployed or operating in whole or in part in New York State; the article contains no further extraterritoriality clause, and this pack takes no position on when an out-of-state or non-US developer falls within it. Unlike California SB 53, the enacted Article 44-B contains no federal-contract carve-out and no express federal-preemption savings clause, and no local-preemption clause; § 1424 states the opposite of a savings limitation, providing that the article's duties 'are cumulative with any other duties or obligations imposed under other law'. § 1427(2) forecloses any private right of action. Section 4 of Chapter 96 is a severability clause. None of this is legal advice; applicability is a question for the operator's counsel.", "Provisions deliberately excluded from the mapping, with reasons. (i) Section 1 of Chapter 96 (legislative findings and intent) imposes no conduct duty; it is quoted in rationales as evidence of the interest a provision protects, which is the same use the ca-sb53-tfaia pack makes of that act's SEC. 1. (ii) § 1420 (definitions) imposes no conduct duty; the definitions are used throughout the rationales instead. (iii) § 1422(1), (2)(b), (4)–(7), (8)(b), (10) and § 1429 impose duties on, or confer powers on, the office and the department rather than on developers. (iv) § 1423 (loss of equity value is not property damage) and § 1424 (cumulative duties) are interpretive. (v) § 1425 (scope) and § 1426 (exceptions for accredited academic research and Empire AI) delimit application rather than prescribe conduct; both are recorded in the scope-thresholds note. (vi) § 1427 (civil penalties recoverable by the attorney general; no private right of action; preservation of third-party-responsibility defences) is an enforcement provision rather than a conduct norm, and is cited in the entries whose breach it prices. One feature of it is recorded here because a reader comparing the two US state packs will notice it: the operative clause of § 1427(1) is framed 'where a large frontier developer' fails to publish or transmit a compliant document, makes a statement in violation of § 1421(4), fails to report an incident under § 1422, or fails to comply with its own frontier AI framework — whereas four of the duties it prices (§ 1421(3)(a), § 1421(4)(a)(i), § 1421(5) and § 1422(3)) are imposed on any frontier developer, large or not. California's analogue at Bus. & Prof. Code § 22757.15 reaches 'a frontier developer' without that qualifier. Whether the New York framing leaves a non-large frontier developer's § 1422(3) reporting duty without a § 1427 penalty, or whether it is read distributively across the enumerated grounds, is a question of statutory construction that this pack records and does not resolve; it is an RFC agenda item and, for operators, a question for counsel. (vii) § 1421(3)(a)(i)–(vii), the seven enumerated transparency-report fields, are mapped as a set through the § 1421(3) entry rather than individually, because no single field carries a distinct judgment direction. Each exclusion is an RFC agenda item.", "V/E/S codes are the canonical three-letter AIO 00011 vocabulary served at /api/framework/vocabulary, and the dual formalization tightened the discipline of the Wave 1 source-axis policy (P4) rather than loosening it. The policy has two limbs and is applied as a filter, never as a generator: it may remove a code both passes declared and may decide which of two divergent readings prevails, but it never adds a code that neither pass declared. **`Gov` is declared only where the excerpt names a government body or a government norm as decisive on the substance of the duty**; being named as the recipient of a report, a filing or a disclosure does not earn it. **`Ind` is declared where the excerpt makes the industry duty-bearer the author or performer of the provision's product or determination.** After adjudication `Gov` survives at exactly three entries — § 1421(1)(a), where national and international standards are named as substance the framework must incorporate; § 1421(1)(j) with § 1422(2)(a), where the alternative reporting cadence must be 'agreed upon by the office'; and § 1422(8)–(9) and § 1428, where the designated federal instrument and the office's prescribed form are the yardsticks themselves — and it is removed from § 1422(3), where the office and any law enforcement or public safety agency with jurisdiction appear solely as recipients of the disclosure, and from the four entries where v0.1 had carried it in from the fact that the duty is statutory. `Ind` survives at six entries and is removed at three: at § 1421(1)(d)–(e) because the excerpt is a bare list of framework contents naming no performer, and at § 1422(8)–(9) and § 1428 because the developer authors only its election or its filing and not the standard or form whose content decides the matter — at § 1428 the statute says so in terms, pricing the failure 'to correct false information' as a daily penalty. `Pro` and `Pee` appear nowhere, and both passes independently withheld `Pro` at § 1421(1)(e), where the article requires the framework to describe the use of 'third parties' to assess catastrophic risk but imposes no independence, accreditation or competence criterion on them; the resulting empty source layer is Wave 2 gap 14. `Ano` appears nowhere, because the RAISE Act has no whistleblower chapter. Of the two codes v0.1 flagged as INFERENCE, neither survives: `Hum` at § 1421(1)(d)–(e) and `Log` at § 1421(5) were both removed under the standing rule that an inference-grade code survives only where both independent passes reach it, and the blind pass reached neither — it considered `Log` at § 1421(4) and at § 1421(5) and rejected it expressly. The inference-grade codes the blind pass introduced and v0.1 did not reach — `Unc` at § 1421(3), `Por` at § 1421(5) and § 1428, `Sep` at § 1422(8)–(9), `Tri` considered and withheld at § 1422(3) — were treated identically and none survives.", "Article → V/E/S translation methodology: /content/standards-packs/FORMALIZATION_METHODOLOGY.md.", "AIO certifies conformance to AIO's own formalization of the Responsible AI Safety and Education Act. This is not a legal conformity assessment. The State of New York, its Legislature, the Senate, the Assembly, the Legislative Bill Drafting Commission, the Governor, the Department of Financial Services and its superintendent, the office established under § 1420(16), and the Attorney General took no part in this formalization, have not reviewed or endorsed it, and it is not an official interpretation of the act. Nothing in this pack confers any presumption of compliance, and no result measured against it is a defence to an action under § 1427 or § 1428(5).", "Measurement scope (per-entry `obligationType`, pack-level `measurementScope`) — and the roster correction, now firmer than v0.1 could make it. After adjudication, three of the ten mapped provisions are `organizational` (§ 1421(1) with § 1421(2), § 1421(4) and § 1428) and seven are `mixed`; none is `behavioral`. v0.1's distribution was 1 behavioral / 7 mixed / 2 organizational. **The v0.1 claim that § 1421(4) is a `behavioral` entry does not survive the adjudication and is withdrawn**, on the same textual ground the California twin was lowered on and set out in full at that entry. This is the exact distribution v0.1 predicted for that outcome — organizational 3 / mixed 7 / behavioral 0 — and it strengthens rather than weakens the roster correction: the roster's 매우 측정가능(→14) rating for this candidate was an over-estimate, the RAISE Act is in substance a DISCLOSURE statute, and every one of its ten mapped provisions is discharged by publishing a document, keeping a clock, filing a form or paying an assessment, with seven of them carrying a judgment correlate on top that an item can test. One further caution for item authoring: § 1421(3), the transparency report, came out of the adjudication with an undeclared value layer, and an entry that states no normative direction cannot carry an item until the RFC round settles it. Across the fifteen packs that have been dual-formalized and adjudicated in Waves 1 to 3, the only surviving `behavioral` entry remains paragraph 36 of the UNESCO pack, where both independent passes reached that tag without prompting and neither had been exposed to a prior adjudication of the same words.", "RESOLVED — the conflict with the sibling pack's adjudication, and how it was decided. v0.1 recorded an open conflict as this pack's first-priority question for the second formalization: while v0.1 was being written, ca-sb53-tfaia advanced to v0.2 and its adjudication lowered Bus. & Prof. Code § 22757.12(e) from `behavioral` to `organizational`, and § 1421(4) of the RAISE Act is the near-identical New York twin of that provision. v0.1 recorded `behavioral` anyway, disclosing both its reason (a first pass should state its own reading rather than inherit another pack's judgment) and the fact that this axis of its own work was NOT blind to the California adjudication. The 2026-08-14 adjudication closes the conflict in California's favour, and the reasoning is recorded at the entry: the texts do not diverge (the only wording differences in the whole provision are New York's 'This subdivision shall not apply' for California's 'This subdivision does not apply', and the paragraph numbering — neither bears on who speaks or on what is spoken about); the disclosed contamination means the two-pass agreement on this axis is an agreement between one contaminated pass and one blind pass rather than between two independent readings, which is how Wave 1 practice requires a disclosed contamination to be processed; and a contested obligationType resolves to the more conservative tag in any event. The blind pass's independent `behavioral` reading is preserved as an RFC item rather than discarded, because it is the strongest surviving argument that the three-way obligationType vocabulary cuts this provision at the wrong joint: the duty genuinely is discharged in the act of speaking, and what the `organizational` tag actually records is that the speaker the statute addresses is a corporation making a disclosure rather than a model answering in a concrete case. The independent value of the divergence that v0.1 wanted to preserve is not lost either — it is now on the record as a disagreement between a blind formalization and a cross-pack ruling, which is a sharper datum than a first pass's dissent would have been.", "New York adds one obligation family California lacks (the § 1428 registration and assessment regime) and drops one California has (the whistleblower chapter), and both changes push the distribution further toward organizational rather than toward measurable judgment. A pass measured against this pack is evidence about model judgment only; it is never evidence that a developer published a framework, filed a report, met a deadline, filed a disclosure statement, or is even within the statute's scope.", "Adjudication method (v0.2). This pack was formalized twice. The v0.1 seed pass is the first formalization; the second was blind, under the protocol recorded in the first note. The two results were compared mechanically, entry by entry and layer by layer, with v, e and s treated as sets. Exact agreement was auto-accepted. Divergences were adjudicated under a fixed policy carried forward from Waves 1 and 2: the reading better grounded in the quoted text prevails under FORMALIZATION_METHODOLOGY.md §4; where both readings are defensible the more conservative is taken (fewer codes, or a layer left undeclared); the intersection is an allowed outcome where it is non-empty and defensible; no third reading is invented, and every adjudicated set is a subset of at least one pass's set. The sub-rules settled in earlier waves and applied here without change: a code flagged INFERENCE by the pass that declared it survives only where both passes reached it; a divergent `obligationType` resolves to the more conservative tag (organizational over mixed over behavioral); a government body named only as the recipient of a report, filing or disclosure is not thereby a source; and the source-axis rule is a filter, never a generator. A claim by either pass that the quoted text RANKS its codes was accepted only where the quote carries the ranking — accepted at § 1422(3), where the seventy-two-hour clock compresses to twenty-four hours for an imminent risk of death or serious physical injury, and at § 1428, where filing is made a precondition of operating; rejected at § 1422(8)–(9), where 'substantially equivalent to, or stricter than' ranks two instruments and not two interests. One ruling was made against a two-pass agreement, at § 1421(4), and it is the only one; its grounds are set out at the entry and in the resolution note above. Agreement statistics for this pack, across ten entries: V 1/10, E 3/10, S 1/10, obligationType 9/10, all four axes together 0/10. The obligationType figure is the highest in any wave and the value figure among the lowest; the systematic pattern, as in the sibling pack, is that v0.1 declared more value codes than the quoted words carry — nine of the ten entries lost at least one value code, and § 1421(3) lost the layer entirely — because v0.1 argued from the legislative findings, the definitions at § 1420 and the penalty sections, all of which sit outside the quoted excerpts, while the blind pass had only the excerpts.", "Contamination notice — two items, both disclosed, neither affecting the same axis. (1) THE SECOND PASS'S PROTOCOL DEVIATION. The blind second formalizer read three source files outside its assigned two-file scope — src/app/api/framework/vocabulary/route.ts, src/app/lib/frameworkVocabulary.ts and the VALUES / EVIDENCE_CAT / SOURCE_CAT tables in src/app/components/standards/workshopData.ts — and disclosed the fact and the reason in its own output before adjudication: FORMALIZATION_METHODOLOGY.md §4 names the AIO 00011 vocabulary but does not reproduce it, deferring the code list to /api/framework/vocabulary, so the assigned scope was insufficient to emit valid three-letter codes. What was read is the neutral codebook — the 19 value codes, 10 evidence codes and 10 source codes with their definitions — and nothing else: no standards-pack JSON, no other second-pass file, no docs/ file, no management guide, no git history. The files contain no mapping for this norm or for any other. Recorded as a protocol deviation, disclosed, neutral codebook only, no contamination of any axis. The remedy for the next wave is to serve the codebook to the second formalizer as part of the assigned scope rather than leaving it to be found. (2) THE FIRST PASS'S DISCLOSED NON-BLINDNESS ON ONE AXIS. v0.1 recorded that its `obligationType` reading of § 1421(4) was NOT blind to the ca-sb53-tfaia adjudication of the same words. That axis, and only that axis, is reported with a contamination caveat, and it is processed at adjudication exactly as Wave 1 practice requires — the affected pass's reading on that axis is not counted as independent confirmation. No other field of v0.1 was exposed to a prior adjudication.", "Vocabulary and schema gaps found by the dual formalization (feeding a future AIO 00011 RFC). This pack adds two new items to the running list and confirms four existing ones. NEW — (28, renumbered from a provisional 24 in the wave-end consolidation 2026-08-14) THE SPEAKER BEHIND A CORPORATE DISCLOSURE. § 1421(4) forbids a materially false or misleading statement, a duty discharged in the act of speaking, yet the entry is tagged `organizational` because the speaker the statute addresses is a corporation making a disclosure rather than a model answering in a concrete case. The blind pass tagged it `behavioral` on the first half of that sentence and the adjudication tagged it `organizational` on the second, and both are right about what they see: the three-way vocabulary has no way to record a duty that is behavioral in form and organizational in addressee. This is the sharpest disagreement the two packs on this statute have produced and it should be settled at the vocabulary level rather than pack by pack. NEW — (29, renumbered from a provisional 25) A NEGOTIATED REGULATORY SCHEDULE. § 1422(2)(a) lets the developer propose an alternative reporting cadence that the office must agree to; the source axis can record the office as a trusted source but has no way to record that the governing position is jointly fixed, and the same shape will recur wherever a regulator's agreement converts a proposal into the applicable rule. CONFIRMED AGAIN — (14) CONTRACTED INDEPENDENT EVALUATOR AS A SOURCE CLASS: 'third parties' at § 1421(1)(e) and 'third-party evaluators' at § 1421(3)(b)(iii) designate a party whose assessment the statute treats as material, and none of the ten source codes carries it; both passes left it unassigned and the layer at § 1421(1)(d)–(e) is empty. (15) HUMANS KEEPING CONTROL OF THE SYSTEM: § 1421(1)(j)'s 'frontier model circumventing oversight mechanisms' names an interest the value layer cannot express. (18) INCIDENT REPORTING AS AN EVIDENCE CLASS: at § 1422(3) the two passes split three ways across Gui, Exp and Log for what discharges a report of one specific occurrence, and only the judgment that starts the clock survived. WAVE 1 GAP 9, an interest preserved AGAINST the duty rather than served by it, is confirmed for a third time — trade secrets at § 1421(5), where the blind pass proposed Por and it was not carried. One further RFC agenda item is procedural rather than lexical: the obligationType tags of this pack and of ca-sb53-tfaia are not aligned entry for entry on the seven provisions the two statutes share almost verbatim, because each pack's tags are the product of its own pair of passes; whether cross-pack consistency should be enforced on that axis generally, as it was here at § 1421(4) alone, is a question for the RFC round.", "No item bank exists for this pack (itemBankRef.publicSet and privateSet are both null). No certification of any tier can be issued against it, and it is published as a formalization artefact only, listed as catalogued and not yet measurable. Operators should further note that the formalized article does not take effect until January 1, 2027, so a measurement taken against this pack today is measured against a norm that is enacted but not yet operative." ] }