STATE OF NEW YORK
________________________________________________________________________
10731
IN SENATE
October 28, 2026
___________
Introduced by Sens. GOUNARDES, FAHY -- read twice and ordered printed,
and when printed to be committed to the Committee on Rules
AN ACT to amend the general business law, in relation to the responsible
AI safety and education (RAISE) act
The People of the State of New York, represented in Senate and Assem-bly, do enact as follows:
1 Section 1. Section 1420 of the general business law, as added by chap-
2 ter 96 of the laws of 2026, is amended to read as follows:
3 § 1420. Definitions. As used in this article, the following terms
4 shall have the following meanings:
5 1. "Affiliate" means a person controlling, controlled by, or under
6 common control with a specified person, directly or indirectly, through
7 one or more intermediaries.
8 2. "Artificial intelligence model" means an engineered or machine-
9 based system that varies in its level of autonomy and that can, for
10 explicit or implicit objectives, infer from the input it receives how to
11 generate outputs that can influence physical or virtual environments.
12 3. (a) "Catastrophic risk" means a foreseeable and material risk that
13 a frontier developer's development, storage, use, or deployment of a
14 frontier model will materially contribute to the death of, or serious
15 injury to, more than fifty people or more than one billion dollars in
16 damage to, or loss of, property arising from a single incident or multi-
17 ple incidents involving a frontier model doing any of the following:
18 (i) providing expert-level assistance in the creation or release of a
19 chemical, biological, radiological, or nuclear weapon;
20 (ii) engaging in conduct with [no meaningful] limited human oversight,
21 intervention, or supervision that is either a cyberattack or, if the
22 conduct had been committed by a human, would constitute the crime of
23 murder, assault, extortion, or theft, including theft by false pretense;
24 [or]
25 (iii) attempting to overthrow or seize control of the federal govern-
26 ment or state or local governments in the state of New York; or
27 (iv) evading the control of its frontier developer or user.
EXPLANATION--Matter in italics (underscored) is new; matter in brackets
[] is old law to be omitted.
LBD16516-01-6
S. 10731 2
1 (b) "Catastrophic risk" does not include a foreseeable and material
2 risk from any of the following:
3 (i) information that a frontier model outputs if the information is
4 otherwise publicly accessible in a substantially similar form from a
5 source other than a foundation model;
6 (ii) lawful activity of the federal government; or
7 (iii) harm caused by a frontier model in combination with other soft-
8 ware if the frontier model did not materially contribute to the harm.
9 4. "Covered emergency" means a situation that poses an imminent
10 catastrophic risk or security risk. "Covered emergency" shall be
11 construed to include a situation in which immediate action is necessary
12 to prevent a catastrophic risk or security risk, even if the harm that
13 would be prevented is not proximate in time to the situation.
14 5. "Critical safety incident" means any of the following:
15 (a) unauthorized access to, modification of, or exfiltration of[, the]
16 unreleased model weights of a frontier model [that results in death or
17 bodily injury];
18 (b) harm resulting from the materialization of a catastrophic risk;
19 (c) [loss of control of a frontier model causing death or bodily inju-
20 ry; or
21 (d) a frontier model that uses deceptive techniques against the fron-
22 tier developer to subvert the controls or monitoring of its frontier
23 developer outside of the context of an evaluation designed to elicit
24 this behavior and in a manner that demonstrates materially increased
25 catastrophic risk] conduct by a foundation model of a frontier developer
26 that circumvents, disables, evades, or attempts to circumvent, disable,
27 or evade, a material, technical or organizational control of its fron-
28 tier developer that limits the model's access to systems, resources, or
29 data, provides for its suspension or modification, or monitors its
30 conduct or capabilities, outside of the context of an evaluation
31 designed to elicit such conduct where such conduct remains confined to
32 the evaluation environment;
33 (d) conduct by a frontier model that, without the intent of any user,
34 materially impairs the confidentiality, integrity, or availability of a
35 computer system or computer network which is not owned or operated by
36 the frontier developer or user and where such conduct occurs without the
37 consent of the owner or operator of such system or network;
38 (e) an occurrence in which a safeguard against an incident described
39 in this subdivision failed, was circumvented, or was materially
40 degraded, such that, absent timely intervention, mitigation, or fortui-
41 ty, there was a substantial likelihood that such a critical safety inci-
42 dent would have resulted; or
43 (f) any other occurrence or condition designated by regulation of the
44 office which is related to catastrophic risk, security risk, or the
45 automation of AI research and development.
46 6. "Department" means the department of financial services.
47 [5.] 7. (a) "Deploy" means to make a frontier model available to a
48 third party for use, modification, copying, or combination with other
49 software.
50 (b) "Deploy" does not include making a frontier model available to a
51 third party for the primary purpose of developing or evaluating the
52 frontier model.
53 [6.] 8. "Foundation model" means an artificial intelligence model that
54 is all of the following:
55 (a) trained on a broad data set;
56 (b) designed for generality of output; and
S. 10731 3
1 (c) adaptable to a wide range of distinctive tasks.
2 [7.] 9. "Frontier AI framework" means documented technical and organ-
3 izational protocols to:
4 (a) manage, assess, and mitigate catastrophic risks; and
5 (b) assess the ability of frontier models to automate AI research and
6 development and to manage any catastrophic risks and security risks
7 associated with such automation.
8 [8.] 10. "Frontier developer" means a person who has trained, or
9 initiated the training of, a frontier model, with respect to which the
10 person has used, or intends to use, at least as much computing power to
11 train the frontier model as would meet the technical specifications
12 found in subdivision [nine] twelve of this section.
13 11. "Frontier examiner" means a third party that performs services
14 under section fourteen hundred twenty-four of this article and that is
15 eligible to be a frontier examiner under such section.
16 [9.] 12. (a) "Frontier model" means a foundation model that was
17 trained using a quantity of computing power greater than 10§[26] 25
18 integer or floating-point operations, provided, however, that foundation
19 models shall be considered to be frontier models if the office has
20 determined that the foundation model:
21 (i) is among the most capable foundation models;
22 (ii) possesses capabilities which pose catastrophic risk; and
23 (iii) is reasonably likely to possess the ability to automate AI
24 research and development.
25 (b) The quantity of computing power described in paragraph (a) of this
26 subdivision shall include computing for the original training run and
27 for any subsequent fine-tuning, reinforcement learning, or other materi-
28 al modifications the developer applies to a preceding foundation model.
29 13. "Internal use" means any use of a frontier model by its frontier
30 developer or its affiliates, including, but not limited to, any of the
31 following:
32 (a) the training of the frontier model;
33 (b) modifying or fine-tuning the frontier model;
34 (c) using the frontier model for internal business purposes;
35 (d) evaluating the frontier model; and
36 (e) using the frontier model for AI research and development.
37 [10.] 14. "Large frontier developer" means a frontier developer that
38 together with its affiliates collectively [had annual gross revenues in
39 excess of five hundred million dollars in the preceding calendar year]
40 incurred more than one billion dollars in expenditures for research,
41 development, training, fine-tuning, modification, security, evaluation,
42 or other activities which have contributed to the development, creation,
43 or modification of one or more frontier models in the preceding thirty-
44 six months.
45 [11.] 15. "Model weight" means a numerical parameter in a frontier
46 model that is adjusted through training and that helps determine how
47 inputs are transformed into outputs.
48 [12. "Department" means the department of financial services.
49 13. "Property" means tangible or intangible property.
50 14. "Person" means an individual, proprietorship, firm, partnership,
51 joint venture, syndicate, business trust, company, corporation, limited
52 liability company, association, committee, or any other nongovernmental
53 organization or group of persons acting in concert.
54 15.] 16. "Office" means an office within the department of financial
55 services, which shall report to the superintendent of financial services
56 and is tasked with implementation of this article.
S. 10731 4
1 17. "Open source release" means making the model weights of a founda-
2 tion model publicly available under a license that grants to all
3 persons, without payment of a fee or royalty and without restriction or
4 discrimination based on the identity of any person or group of persons
5 or on the field of endeavor, the rights to access, use, modify, and
6 redistribute the foundation model or modified versions of the foundation
7 model. A license does not fail to satisfy this subdivision solely
8 because it requires attribution to the frontier developer or requires
9 that modified or redistributed versions be made available under identi-
10 cal or comparable terms.
11 18. "Person" means an individual, proprietorship, firm, partnership,
12 joint venture, syndicate, business trust, company, corporation, limited
13 liability company, association, committee, or any other nongovernmental
14 organization or group of persons acting in concert.
15 19. "Property" means tangible or intangible property.
16 20. "Relevant information" means all information reasonably necessary
17 for a frontier examiner to comply with section fourteen hundred twenty-
18 four of this article, including, but not limited to:
19 (a) all unredacted versions of materials published pursuant to this
20 article;
21 (b) access to frontier models sufficient to enable assessments under
22 section fourteen hundred twenty-four of this article;
23 (c) interviews with employees of a large frontier developer;
24 (d) internal documents of a large frontier developer; and
25 (e) information to which an employee of a large frontier developer
26 whose responsibilities include assessing the capabilities of frontier
27 models or managing catastrophic risk or security risk has access in the
28 ordinary course of employment is relevant information unless the large
29 frontier developer establishes that it is not reasonably necessary for
30 the frontier examiner to perform its functions under section fourteen
31 hundred twenty-four of this article.
32 21. "Specification" means a high-level, general description of the
33 principles that the model is intended to follow, how the model is
34 intended to prioritize different principles and user prompts, and the
35 topics for which the model is intended to refuse, redirect, or otherwise
36 subvert user prompts.
37 22. "Security risk" means a foreseeable and material risk of unauthor-
38 ized access to, modification of, or transfer of the unreleased model
39 weights of a frontier model by any person, entity, or model.
40 23. "Superintendent" means the superintendent of financial services.
41 [16. "Office" means an office within the department of financial
42 services, which shall report to the superintendent of financial services
43 and is tasked with implementation of this article.]
44 § 2. Section 1421 of the general business law, as added by chapter 96
45 of the laws of 2026, is amended to read as follows:
46 § 1421. Transparency requirements. 1. A large frontier developer shall
47 write, implement, comply with, and clearly and conspicuously publish on
48 its internet website a frontier AI framework that applies to the large
49 frontier developer's frontier models and describes in detail how the
50 large frontier developer handles all of the following:
51 (a) incorporating national standards, international standards, and
52 industry consensus best practices into its frontier AI framework;
53 (b) defining and assessing thresholds used by the large frontier
54 developer to identify and assess whether a frontier model has capabili-
55 ties that could pose a catastrophic risk or a security risk or whether a
S. 10731 5
1 frontier model has the capability to automate AI research and develop-
2 ment, which may include multiple-tiered thresholds;
3 (c) applying mitigations to address the potential for catastrophic or
4 security risks based on the results of assessments undertaken pursuant
5 to paragraph (b) of this subdivision;
6 (d) reviewing assessments and adequacy of mitigations as part of the
7 decision to deploy a frontier model or use it [extensively] internally;
8 (e) [using third parties to assess the potential for catastrophic
9 risks and the effectiveness of mitigations of catastrophic risks;
10 (f)] revisiting and updating the frontier AI framework, including any
11 criteria that trigger updates and how the large frontier developer
12 determines when its frontier models are substantially modified enough to
13 require disclosures pursuant to subdivision [three] four of this
14 section;
15 [(g)] (f) cybersecurity and information security practices to [secure
16 unreleased model weights from unauthorized modification or transfer by
17 internal or external parties] mitigate security risks based on the
18 results of assessments undertaken pursuant to paragraph (b) of this
19 subdivision;
20 [(h)] (g) identifying and responding to critical safety incidents;
21 [(i)] (h) instituting internal governance practices to ensure imple-
22 mentation of [these processes] the frontier AI frameworks; [and
23 (j)] (i) assessing and managing catastrophic risk resulting from the
24 internal use of its frontier models, including risks resulting from a
25 frontier model circumventing oversight mechanisms and risks related to
26 the frontier model being used to automate AI research and development;
27 (j) assessing frontier models for signs of exploiting, undermining, or
28 evading the large frontier developer's training, assessment, or miti-
29 gation processes or for deceptive or manipulative behavior; and
30 (k) whether and how the large frontier developer uses frontier models
31 to monitor and evaluate other frontier models.
32 2. The office may adopt regulations to add additional elements that a
33 large frontier developer is required to include in its frontier AI
34 framework to account for changes in the scientific understanding of
35 catastrophic risk, changes in AI capabilities or mitigation approaches,
36 or to add additional elements that would effectuate the purposes of this
37 chapter.
38 3. (a) A large frontier developer shall review and, as appropriate,
39 update its frontier AI framework at least once per year.
40 (b) If a large frontier developer makes a material modification to its
41 frontier AI framework, the large frontier developer shall clearly and
42 conspicuously publish the modified frontier AI framework and a justi-
43 fication for that modification within thirty days.
44 [3.] 4. (a) Before, or concurrently with, deploying a new frontier
45 model or a substantially modified version of an existing frontier model,
46 a frontier developer shall clearly and conspicuously publish on its
47 internet website a transparency report containing all of the following:
48 (i) the internet website of the frontier developer;
49 (ii) a mechanism that enables a natural person to communicate with the
50 frontier developer;
51 (iii) the release date of the frontier model;
52 (iv) the languages supported by the frontier model;
53 (v) the modalities of output supported by the frontier model;
54 (vi) the intended uses of the frontier model; [and]
55 (vii) the frontier model specification; and
S. 10731 6
1 (viii) any generally applicable restrictions or conditions on uses of
2 the frontier model.
3 (b) Before, or concurrently with, deploying a new frontier model or a
4 substantially modified version of an existing frontier model, a large
5 frontier developer shall include in the transparency report required by
6 paragraph (a) of this subdivision, summaries of all of the following:
7 (i) assessments of catastrophic risks or security risks from the fron-
8 tier model conducted pursuant to the large frontier developer's frontier
9 AI framework;
10 (ii) the results of [the] such assessments [under subparagraph (i) of
11 this paragraph];
12 (iii) the extent to which third-party evaluators were involved in such
13 assessments; [and]
14 (iv) the extent to which such assessments are reproducible by external
15 entities; and
16 (v) other steps taken to fulfill the requirements of the frontier AI
17 framework with respect to the frontier model.
18 (c) A frontier developer that publishes the information described in
19 paragraph (a) or (b) of this subdivision as part of a larger document,
20 including a system card or model card, shall be deemed in compliance
21 with the applicable paragraph.
22 [4.] 5. (a) (i) A frontier developer shall not make a materially false
23 or misleading statement about catastrophic risk or security risk from
24 its frontier models or its management of catastrophic risk or security
25 risk.
26 (ii) A large frontier developer shall not make a materially false or
27 misleading statement about its implementation of, or compliance with,
28 its frontier AI framework.
29 (iii) A frontier developer shall not make a materially false or
30 misleading statement in any publication, submission, or filing to the
31 office under this article.
32 (iv) A large frontier developer shall designate an accountable senior
33 officer responsible for compliance with this chapter. The accountable
34 senior officer shall certify, after reasonable inquiry and under penalty
35 of perjury, that, to the best of the officer's knowledge, each publica-
36 tion and material submission to the office is accurate, complete, and
37 not materially misleading.
38 (b) [This subdivision] Subparagraphs (i), (ii) and (iii) of paragraph
39 (a) of this subdivision shall not apply to a statement that was made in
40 good faith and was reasonable under the circumstances.
41 [5.] 6. (a) A large frontier developer shall implement technical and
42 administrative measures that are sufficient to allow it to do both of
43 the following:
44 (i) prevent an identified imminent critical safety incident; and
45 (ii) contain a critical safety incident as defined in subdivision five
46 of section fourteen hundred twenty of this article.
47 (b) This subdivision shall not be construed to require technical and
48 administrative measures sufficient to prevent or contain a critical
49 safety incident that arises from a copy of a frontier model that has
50 been distributed in an open source release and that is not in the
51 possession of its frontier developer.
52 7. (a) When a frontier developer publishes documents to comply with
53 this section, such frontier developer may make redactions to such docu-
54 ments that are necessary to protect such frontier developer's trade
55 secrets, such frontier developer's cybersecurity, public safety, or the
S. 10731 7
1 national security of the United States or to comply with any federal or
2 state law.
3 (b) If a frontier developer redacts information in a document pursuant
4 to this subdivision, such frontier developer shall:
5 (i) describe the character and justification of such redaction in any
6 published version of such document to the extent permitted by the
7 concerns that justify redaction [and shall];
8 (ii) retain the unredacted information for five years; and
9 (iii) grant the office access to an unredacted copy of the publication
10 upon request.
11 § 3. Section 1422 of the general business law, as added by chapter 96
12 of the laws of 2026, is renumbered to be section 1423 and a new section
13 1422 is added to read as follows:
14 § 1422. Minimum standards. 1. No later than January first, two thou-
15 sand twenty-nine, the office shall adopt regulations, which shall be
16 binding on large frontier developers, establishing minimum standards for
17 the frontier AI frameworks required in section fourteen hundred twenty-
18 one of this article as well as minimum standards for required miti-
19 gations of catastrophic risk and security risk which shall address:
20 (a) each element required to be included in subdivision one of section
21 fourteen hundred twenty-one of this article;
22 (b) each type of catastrophic risk enumerated in the definition of
23 catastrophic risk in subdivision three of section fourteen hundred twen-
24 ty of this article;
25 (c) security risks; and
26 (d) the automation of AI research and development.
27 2. In implementing this section, the office shall aim to promulgate
28 regulations that, if successfully implemented, would come as close as
29 practicable to preventing catastrophic risks and security risks that are
30 unreasonable in light of the societal benefits of the activities that
31 create those risks.
32 3. In implementing this section, the office shall consider whether,
33 given the current state of knowledge regarding AI safety and security,
34 the subject matter of the proposed regulation is sufficiently understood
35 to support effective regulation, or whether the purposes of this article
36 would be better served through voluntary guidance, monitoring, or
37 further study.
38 4. In implementing this section, the office shall consider relevant
39 national standards, international standards, and industry consensus best
40 practices, including existing frontier AI frameworks published by fron-
41 tier developers.
42 5. The office shall consult with frontier developers, frontier examin-
43 ers, and other technical experts when creating the minimum standards
44 prescribed herein.
45 § 4. Section 1423 of the general business law, as added by chapter 96
46 of the laws of 2026, and as renumbered by section three of this act, is
47 amended to read as follows:
48 § 1423. Reporting. 1. The office shall establish a mechanism to be
49 used by a frontier developer or a member of the public to report a crit-
50 ical safety incident that includes all of the following:
51 (a) the date of the critical safety incident;
52 (b) the reasons the incident qualifies as a critical safety incident;
53 (c) a short and plain statement describing the critical safety inci-
54 dent; [and]
55 (d) whether the incident was associated with internal use of a fron-
56 tier model; and
S. 10731 8
1 (e) any other information the office may deem necessary to better
2 understand such incident or the state of catastrophic risk, security
3 risk, and the automation of AI research and development.
4 2. (a) A large frontier developer shall transmit to the office a
5 summary of any assessment of:
6 (i) catastrophic risk resulting from internal use of its frontier
7 models [every three months or pursuant to another reasonable schedule
8 requested by the large frontier developer, communicated in writing to
9 the office with written updates, as appropriate, and agreed upon by the
10 office.];
11 (ii) the ability of its frontier models to automate AI research and
12 development and their use in AI research and development, including a
13 specific percentage of how much of such research and development is
14 currently automated;
15 (iii) the percentage of computing power used for models that are
16 internally used and/or deployed; and
17 (iv) a general description of how such models are being internally
18 used and/or deployed.
19 (b) A large frontier developer shall transmit the assessments required
20 under paragraph (a) of this subdivision at all of the following times:
21 (i) every three months or pursuant to another reasonable schedule
22 requested by the large frontier developer, communicated in writing to
23 the office, and agreed upon by the office;
24 (ii) within seventy-two hours of any time that there has been a
25 substantial increase in the large frontier developer's assessment of
26 catastrophic risks or security risks from internal use of its frontier
27 models, or capabilities of its frontier models to automate AI research
28 and development, in comparison to the previously reported assessment;
29 and
30 (iii) at any other time that the office may by regulation require to
31 ensure the office can adequately monitor the state of catastrophic
32 risks, security risks, the ability of frontier models to automate AI
33 research and development, and frontier models' use in AI research and
34 development.
35 The office shall establish a mechanism to be used by a large frontier
36 developer to confidentially submit summaries of any assessments [of the
37 potential for catastrophic risk resulting from internal use of its fron-
38 tier models] required under paragraph (a) of this subdivision, provided,
39 however, that such summaries shall be shared with the attorney general.
40 The attorney general shall be empowered to pursue the actions provided
41 in subdivision two of section fourteen hundred thirty of this article as
42 set forth in such subdivision.
43 [(b)] (c) The office shall take all reasonable precautions to limit
44 access to any reports related to internal use of frontier models [to
45 only personnel authorized to know the information] and to protect the
46 reports from unauthorized access.
47 3. (a) [Subject to paragraph (b) of this subdivision, a] A frontier
48 developer shall report any critical safety incident pertaining to one or
49 more of its frontier models to the office within seventy-two hours from
50 a determination that a critical safety incident has occurred or within
51 seventy-two hours of the frontier developer learning facts sufficient to
52 establish a reasonable belief that a critical safety incident has
53 occurred.
54 (b) If a frontier developer discovers that a critical safety incident
55 poses an imminent risk of death or serious physical injury, the frontier
56 developer shall disclose that incident within twenty-four hours to an
S. 10731 9
1 authority, including any law enforcement agency or public safety agency
2 with jurisdiction, that is appropriate based on the nature of that inci-
3 dent and as required by law.
4 (c) A frontier developer that discovers information about a critical
5 safety incident after filing the initial report required by this subdi-
6 vision may file an amended report.
7 4. The office shall review:
8 (a) critical safety incident reports submitted by frontier developers
9 [and];
10 (b) assessments transmitted by frontier developers under subdivision
11 two of this section; and
12 (c) reports, notifications, and other information transmitted to the
13 office by a frontier examiner pursuant to this article.
14 The office may review reports submitted by members of the public.
15 5. (a) The office may transmit reports of critical safety incidents or
16 summaries of any assessments [of catastrophic risk from internal use of
17 frontier models] required under subdivision two of this section to other
18 governmental entities at their discretion, considering for example and
19 without limitation the following: the severity of any such incident,
20 potential ongoing risks, legal or regulatory obligations, the need for
21 coordinating with other governmental agencies or other entities and the
22 availability of information. [The office shall consider transmitting
23 such reports or summaries to the office of the attorney general, as
24 appropriate.] Any report transmitted from the office to another govern-
25 mental entity shall be exempt from disclosure under article six of the
26 public officers law.
27 (b) The office may consider, at its discretion, any risks related to
28 trade secrets, public safety, cybersecurity of a frontier developer, or
29 national security when transmitting reports.
30 6. A report of a critical safety incident submitted to the office
31 pursuant to this section and a report of assessments of catastrophic
32 risk from internal use pursuant to section fourteen hundred twenty-one
33 of this article, are exempt from disclosure under article six of the
34 public officers law.
35 7. (a) Beginning January first, two thousand twenty-eight, and annual-
36 ly thereafter, the office shall produce a report, that includes the
37 following:
38 (i) anonymized and aggregated information about critical safety inci-
39 dents that have been reviewed by the office since the preceding report;
40 (ii) any information that the office deems relevant to frontier model
41 safety;
42 (iii) recommended updates to this article, if any; and
43 (iv) any developments relevant to the purposes of this article.
44 (b) The office shall not include information in a report pursuant to
45 this subdivision that would compromise the trade secrets or cybersecuri-
46 ty of a frontier developer, public safety, or the national security of
47 the United States or that would be prohibited by any federal or state
48 law.
49 (c) The office shall transmit a report pursuant to this subdivision to
50 the governor, the temporary president and minority leader of the senate,
51 the speaker and minority leader of the assembly, the chair and ranking
52 member of the senate committee on internet and technology, and the chair
53 and ranking member of the assembly committee on science and technology.
54 8. The office may adopt regulations designating one or more federal
55 laws, regulations, or guidance documents that meet all of the following
56 conditions for the purposes of subdivision nine of this section:
S. 10731 10
1 (a) (i) the law, regulation, or guidance document imposes or states
2 standards or requirements for critical safety incident reporting that
3 are substantially equivalent to, or stricter than, those required by
4 subdivision three of this section; and
5 (ii) the law, regulation, or guidance document described in subpara-
6 graph (i) of this paragraph does not need to require critical safety
7 incident reporting to the state of New York; and
8 (b) the law, regulation, or guidance document is intended to assess,
9 detect, or mitigate the catastrophic risk.
10 9. (a) A frontier developer that intends to comply with subdivision
11 three of this section by complying with the requirements of, or meeting
12 the standards stated by, a federal law, regulation, or guidance document
13 designated pursuant to subdivision eight of this section shall declare
14 its intent to do so to the office.
15 (b) After a frontier developer has declared its intent pursuant to
16 paragraph (a) of this subdivision, the following shall apply:
17 (i) the frontier developer shall be deemed in compliance with subdivi-
18 sion three of this section to the extent that the frontier developer
19 meets the standards of, or complies with the requirements imposed or
20 stated by, the designated federal law, regulation, or guidance document
21 until the frontier developer declares the revocation of that intent to
22 the office or the office revokes a relevant regulation pursuant to
23 subdivision ten of this section;
24 (ii) the failure by a frontier developer to meet the standards of, or
25 comply with the requirements stated by, the federal law, regulation, or
26 guidance document designated pursuant to subdivision eight of this
27 section shall constitute a violation of this article; and
28 (iii) frontier developers who comply with subdivision three of this
29 section by meeting such federal standards shall send copies of any crit-
30 ical safety incident reports required by such federal standards to the
31 office concurrently with sending them to federal authorities.
32 10. The office shall revoke a regulation adopted under subdivision
33 eight of this section if the requirements of subdivision eight of this
34 section are no longer met.
35 § 5. Sections 1423, 1424, 1425, 1426, 1427, 1428 and 1429 of the
36 general business law, as added by chapter 96 of the laws of 2026 are
37 renumbered sections 1427, 1428, 1429, 1430, 1431, 1432 and 1433, respec-
38 tively, and three new sections 1424, 1425 and 1426 are added to read as
39 follows:
40 § 1424. Frontier examiners. 1. (a) No later than sixty days after the
41 effective date of this article or sixty days after the date on which a
42 person first qualifies as a large frontier developer, whichever is
43 later, and thereafter on an annual basis, a large frontier developer
44 shall retain a frontier examiner to produce a report which assesses the
45 following:
46 (i) whether a large frontier developer has complied with its frontier
47 AI framework as well as any instances of noncompliance with such frame-
48 work;
49 (ii) whether such AI framework may require additional detail in order
50 to better assess such compliance; and
51 (iii) whether any statements made by the large frontier developer to
52 the public, office, attorney general, or other governmental entities
53 regarding catastrophic risks or security risks, such developer's manage-
54 ment of such risks, and/or such developer's compliance with its frontier
55 AI framework are inconsistent with the findings of the frontier examin-
56 er's report.
S. 10731 11
1 (b) A frontier examiner shall transmit an unredacted version of the
2 final report to the office and the large frontier developer within seven
3 days of completing such report; provided, however, that such unredacted
4 report shall be exempt from disclosure under article six of the public
5 officers law.
6 (c) No less than twenty-one days after transmitting the final report
7 under paragraph (b) of this subdivision, the frontier examiner shall
8 conspicuously publish such report with appropriate redactions as
9 provided in paragraph (e) of this subdivision.
10 (d) Within seven days of a frontier examiner publishing a report
11 pursuant to paragraph (c) of this subdivision, the large frontier devel-
12 oper shall conspicuously publish a link to such report on such develop-
13 er's public website.
14 (e) Any public version of a frontier examiner's report under this
15 subdivision may contain redactions to the extent necessary to protect
16 the frontier examiner or large frontier developer's trade secrets, the
17 frontier examiner's or frontier developer's cybersecurity, public safe-
18 ty, or the national security of the United States, or to comply with any
19 federal or state law.
20 2. (a) On or before January first, two thousand twenty-eight, a large
21 frontier developer shall retain a frontier examiner to perform ongoing,
22 continuous assessment of the large frontier developer for the purposes
23 of producing the outputs required by this subdivision.
24 (b) On or before April first, two thousand twenty-eight, and every
25 ninety days thereafter, a frontier examiner retained by a larger fron-
26 tier developer under this subdivision shall produce a report that
27 includes:
28 (i) an assessment of each type of catastrophic risk posed by the large
29 frontier developer's frontier models, taking into account model capabil-
30 ities, applicable threat models, and the large frontier developer's
31 mitigations;
32 (ii) an assessment of security risks related to the large frontier
33 developer's frontier model;
34 (iii) an assessment of the ability of the large frontier developer's
35 frontier models to automate artificial intelligence research and devel-
36 opment and any increased potential for catastrophic risks or challenges
37 to risk monitoring, assessment or mitigation resulting from such abili-
38 ty;
39 (iv) an assessment of whether the large frontier developer's activ-
40 ities would pose a substantial level of catastrophic risk over the next
41 ninety day period, if additional mitigations were not put in place; and
42 (v) recommended mitigations, if any, that would reduce the level of
43 catastrophic risk.
44 (c) A frontier examiner shall further include in an appendix to the
45 report described in this subdivision:
46 (i) a list of the personnel involved in the report and the processes
47 involved in creating such report;
48 (ii) the frontier examiner's procedures for managing conflicts of
49 interest and any specific conflicts of interest of any personnel
50 involved in the report;
51 (iii) the methodology of the process used to produce the report and
52 the nature of the information reviewed by the frontier examiner to
53 produce the report; and
54 (iv) the signature of the lead examiner certifying the results of the
55 report.
S. 10731 12
1 (d) A frontier examiner shall transmit an unredacted version of the
2 final report and its appendix to the office and the large frontier
3 developer within seven days of completing such report, provided, howev-
4 er, that such unredacted report shall be exempt from disclosure under
5 article six of the public officers law.
6 (e) No less than twenty-one days after transmitting the final report
7 under paragraph (d) of this subdivision, the frontier examiner shall
8 conspicuously publish such report with appropriate redactions as
9 provided in paragraph (g) of this subdivision.
10 (f) Within seven days of a frontier examiner publishing a report
11 pursuant to paragraph (e) of this subdivision, the large frontier devel-
12 oper shall conspicuously publish a link to such report on such develop-
13 er's public website.
14 (g) Any public version of a frontier examiner's report under this
15 subdivision may contain redactions to the extent necessary to protect
16 the frontier examiner or large frontier developer's trade secrets, the
17 frontier examiner's or frontier developer's cybersecurity, public safe-
18 ty, or the national security of the United States, or to comply with any
19 federal or state law.
20 (h) If a redaction is made pursuant to this subdivision, the person
21 making or requesting the redaction shall provide a written statement
22 with the character and justification of the redaction that is included
23 in any published version of the document to the extent permitted by the
24 concerns that justify redaction, and shall retain the unredacted infor-
25 mation for five years.
26 (i) If a large frontier developer requests a redaction and a frontier
27 examiner believes that such redaction does not comply with paragraph (g)
28 of this subdivision, the frontier examiner shall notify the office in
29 writing within fourteen days of publishing the examiner's report. Upon
30 receipt of such notification or upon the office's own determination that
31 a redaction is not in compliance with this subdivision the office shall:
32 (i) notify the large frontier developer and provide the large frontier
33 developer at least ten days to submit a written justification for such
34 redaction; and
35 (ii) within fifteen days of receiving such justification from the
36 developer, determine in writing whether such redaction complies with
37 paragraph (g) of this subdivision and provide a copy of such determi-
38 nation to the frontier examiner and large frontier developer.
39 (j) If the office determines pursuant to paragraph (i) of this subdi-
40 vision that redaction is not in compliance with paragraph (g) of this
41 subdivision, the frontier examiner shall publish its report and appendix
42 without such redaction within ten days of such office's written determi-
43 nation.
44 (k) A large frontier developer shall grant its frontier examiner
45 access to all relevant information as defined in subdivision twenty of
46 section fourteen hundred twenty of this article.
47 (l) To protect a large frontier developer's trade secrets, confiden-
48 tial business information, cybersecurity, national security of the
49 United States, or public safety, a large frontier developer may impose
50 reasonable security protocols on the frontier examiner, including, but
51 not limited to restrictions on note taking, copying, retaining, or
52 removing materials, requirements for on-premise review, and confiden-
53 tiality requirements.
54 (m) Notwithstanding paragraph (l) of this subdivision, a large fron-
55 tier developer shall not impose security protocols on the frontier exam-
56 iner that would prevent the frontier examiner from reasonably carrying
S. 10731 13
1 out its responsibilities under this section, and a frontier examiner
2 which believes such security protocols have been imposed shall notify
3 the large frontier developer and the office in writing.
4 (n) Upon receipt of a notification pursuant to paragraph (m) of this
5 subdivision, the office shall:
6 (i) provide the developer at least ten days to either submit a written
7 justification for withholding access or information from the frontier
8 examiner, or to provide such access or information;
9 (ii) determine within ten days of receiving the information in subpar-
10 agraph (i) of this paragraph whether information or access is required
11 to be provided by the developer under this section; and
12 (iii) if the office determines that such information or access is so
13 required, provide written notice of such determination to the frontier
14 examiner and large frontier developer.
15 (o) The large frontier developer shall provide such information or
16 access required under subparagraph (iii) of paragraph (n) of this subdi-
17 vision to the frontier examiner within ten days of receiving notice of
18 such requirement from the office.
19 (p) A large frontier developer shall retain for five years each docu-
20 ment created, published, or submitted pursuant to this article alongside
21 records sufficient to permit a frontier examiner to verify: the develop-
22 er's assessments of frontier models; the developer's decisions to deploy
23 or use frontier models, and the developer's handling of critical safety
24 incidents. Upon written request by the frontier examiner or the office,
25 a large frontier developer shall preserve all relevant information iden-
26 tified in such request for five years.
27 3. (a) Any entity that seeks to perform the duties of a frontier exam-
28 iner under this article shall:
29 (i) employ or retain personnel for its duties as a frontier examiner
30 who have technical expertise in catastrophic risks, security risks, and
31 the automation of AI research and development;
32 (ii) have and implement policies to manage conflicts of interest that
33 could impair the independence or integrity of the examiner's work; and
34 (iii) not have financial, operational, or managerial dependence on any
35 large frontier developer or affiliate of a large frontier developer;
36 provided, however, that reasonable compensation for services performed
37 as a frontier examiner pursuant to this article shall not constitute
38 financial dependence for purposes of this subparagraph if the compen-
39 sation is not contingent on the content of any report, finding, or
40 recommendation of the frontier examiner.
41 (b) Beginning January first, two thousand twenty-nine, only entities
42 licensed as a frontier examiner under subdivision four of this section
43 shall perform the duties of a frontier examiner under this article.
44 4. (a) No later than January first two thousand twenty-eight, the
45 office shall adopt regulations establishing a procedure for the licens-
46 ing of frontier examiners who meet the requirements of paragraph (a) of
47 subdivision three of this section and any other requirements the office
48 determines would further the purposes of this article.
49 (b) Such regulations shall detail the procedure by which an entity may
50 apply to be a frontier developer and the procedure by which the office
51 shall approve or deny such application within ninety days.
52 (c) The regulations shall detail conditions under which and the proce-
53 dure by which the office shall revoke the license of a frontier develop-
54 er.
55 (d) The regulations shall ensure the integrity of examinations
56 conducted by licensed examiners by:
S. 10731 14
1 (i) including conflict of interest rules for a frontier examiner, its
2 employees, and its contractors; and
3 (ii) including data security and handling rules for frontier examin-
4 ers.
5 (e) The office may in its discretion require a large frontier develop-
6 er to retain an additional specified frontier examiner:
7 (i) In cases where a critical safety incident has been reported to the
8 office and a frontier examiner is needed to investigate such critical
9 safety incident; and/or
10 (ii) In cases where the office finds that a large frontier developer's
11 original frontier examiner has not performed its duties in a manner that
12 permits the office to have satisfactory visibility into the level of
13 catastrophic risk and security risk posed by the large frontier develop-
14 er's activities.
15 (f) When requiring a large frontier developer to retain an additional
16 frontier developer under paragraph (e) of this subdivision, the office
17 shall describe in detail the identity of the designated additional fron-
18 tier examiner, the reasons for such designation, the form and timing of
19 reports that the additional frontier examiner is expected to produce in
20 the event of a critical safety incident investigation, and the duration
21 of such additional examination, which shall expire after one year absent
22 redesignation by the office.
23 5. A large frontier developer shall not retain an entity as a frontier
24 examiner if the entity is not eligible to be a frontier examiner under
25 this section. Beginning January first, two thousand twenty-nine, a large
26 frontier developer may only retain entities licensed as frontier examin-
27 ers under subdivision four of this section.
28 6. Nothing in this section shall authorize the imposition of a civil
29 penalty solely for a large frontier developer's failure to adopt a
30 recommendation of a frontier examiner where no court order requiring the
31 recommended action has been issued in an action brought pursuant to
32 section fourteen hundred thirty-one of this article.
33 7. (a) If, at any time, a frontier examiner believes that there is a
34 covered emergency related to a large frontier developer's activities,
35 the frontier examiner shall submit a report to the large frontier devel-
36 oper, the office, and the attorney general as soon as practicable, but
37 no later than seventy-two hours after forming such belief.
38 (b) The frontier examiner shall also submit a report containing the
39 matters required by this subdivision to the large frontier developer,
40 the office, and the attorney general within seventy-two hours after the
41 attorney general or the office requests such a report in connection with
42 an action brought or under consideration pursuant to section fourteen
43 hundred thirty-one of this article, whether or not the frontier examiner
44 has independently formed the belief described in paragraph (a) of this
45 subdivision.
46 (c) A report under this subdivision shall state the frontier examin-
47 er's assessment of whether there is a covered emergency, the factual and
48 technical basis for that assessment, any recommendations for actions the
49 large frontier developer could take to prevent or materially mitigate
50 the covered emergency, and any material information the frontier examin-
51 er was unable to obtain.
52 § 1425. Computing power limitations. 1. The office shall be empowered
53 to adopt regulations prescribing the circumstances under which, and the
54 procedure by which, it may order that large frontier developers limit
55 computing power as described in this section. Such order shall apply to
56 all large frontier developers simultaneously, provided however that the
S. 10731 15
1 office shall retain the discretion to order different limitations of
2 computing power per large frontier developer as the office deems neces-
3 sary and in order to achieve the objective of subdivision two of this
4 section.
5 2. Such order may only be issued in order to reduce catastrophic risk,
6 prevent a critical safety incident, or avert a covered emergency. Such
7 order may be informed by reports or assessments submitted or published
8 under this article or by information otherwise available to the office.
9 3. Such order may limit the amount of computing power used to train
10 new models of a large frontier developer and/or the amount of computing
11 power dedicated to automated research and development.
12 4. The office shall retain a frontier examiner as described in section
13 fourteen hundred twenty-four of this article to verify compliance with
14 such order.
15 § 1426. Whistleblower protections. 1. A large frontier developer who
16 is an employer shall be bound by all of the provisions of section seven
17 hundred forty of the labor law.
18 2. A large frontier developer shall not make, adopt, enforce, or enter
19 into any rule, regulation, policy, or contract that prevents an employee
20 from engaging in any of the protected activities provided in paragraph
21 (a), (b) or (c) of subdivision two of section seven hundred forty of the
22 labor law.
23 3. An employee may use the website created under subdivision three of
24 section fourteen hundred thirty-one of this article to disclose any
25 activity under this section.
26 4. (a) In addition to the notice requirements provided in subdivision
27 eight of section seven hundred forty of the labor law, a large frontier
28 developer shall ensure that any new employees receive equivalent notice
29 of such employee's protections, rights, and obligations under such
30 section within ten days of the initial date of such employee's employ-
31 ment.
32 (b) A large frontier developer shall ensure that employees who work
33 remotely receive equivalent notice of their protections, rights, and
34 obligations under section seven hundred forty of the labor law.
35 (c) A large frontier developer shall provide written notice to each
36 covered employee of such employee's rights under section seven hundred
37 forty of the labor law and under this section at least once each year. A
38 large frontier developer shall ensure that the notice is received and
39 acknowledged by all employees.
40 § 6. Section 1431 of the general business law, as added by chapter 96
41 of the laws of 2026 and as renumbered by section five of this act, is
42 amended to read as follows:
43 § 1431. Violations. 1. The attorney general may bring a civil action
44 to recover a civil penalty in an amount not to exceed one million
45 dollars [for a first violation] per day that a first violation occurs
46 and in an amount not to exceed three million dollars per [subsequent
47 violation, determined based on the severity of the violation] day that a
48 subsequent violation occurs where a large frontier developer fails to
49 [publish or transmit a compliant document required to be published or
50 transmitted under this article, makes a statement in violation of subdi-
51 vision four of section fourteen hundred twenty-one of this article,
52 fails to report an incident as required by section fourteen hundred
53 twenty-two of this article, or fails to comply with its own frontier AI
54 framework] comply with this chapter, a regulation adopted pursuant to
55 this article, or a court order issued pursuant to this section.
S. 10731 16
1 2. (a) The attorney general may further bring an action in a court of
2 competent jurisdiction for injunctive relief, including a temporary
3 restraining order, against a large frontier developer where:
4 (i) The large frontier developer violates this article or fails to
5 comply with a regulation adopted pursuant to this article;
6 (ii) A frontier examiner reports pursuant to subdivision seven of
7 section fourteen hundred twenty-four of this article that there is a
8 covered emergency related to the large frontier developer's activities;
9 or
10 (iii) The attorney general otherwise has reason to believe, including
11 on the basis of a report or assessment submitted or published under this
12 article, that there is a covered emergency related to the large frontier
13 developer's activities.
14 (b) In an action under this section, the court may order any relief
15 reasonably necessary to prevent or materially mitigate the covered emer-
16 gency or remedy the violation, including corrective actions, the cessa-
17 tion of any activity, and/or an order restricting the deployment, use,
18 or availability of a model within the state. The court may modify or
19 dissolve an order as circumstances change.
20 (c) The court shall preserve the confidentiality of information exempt
21 from disclosure under this article and/or redacted under this article by
22 reasonable means, including by sealing portions of the record, while
23 making its decision and the basis for the decision public to the maximum
24 extent consistent with that protection.
25 (d) The office shall promptly refer to the attorney general any infor-
26 mation indicating a violation of this chapter or a covered emergency,
27 shall provide to the attorney general on request any record submitted to
28 the office under this article, and shall provide technical assistance in
29 an action under this section at the attorney general's request.
30 3. The attorney general shall maintain a website to receive
31 complaints, information, and/or referrals from persons regarding an
32 operator's or covered platform's alleged compliance or noncompliance
33 with the provisions of this article.
34 4. Nothing in this article shall be construed to establish, authorize
35 or create a private right of action associated with violations of this
36 article.
37 [3.] 5. Nothing in this article shall be construed to prevent a large
38 frontier developer from asserting that another person, entity, or
39 factor, may be responsible for any alleged harm, injury or damage
40 resulting from a catastrophic risk or critical safety incident.
41 § 7. Severability. If any clause, sentence, paragraph, subdivision,
42 section or part of this act shall be adjudged by any court of competent
43 jurisdiction to be invalid, such judgment shall not affect, impair, or
44 invalidate the remainder thereof, but shall be confined in its operation
45 to the clause, sentence, paragraph, subdivision, section, or part there-
46 of directly involved in the controversy in which such judgment shall
47 have been made.
48 § 8. This act shall take effect on the on the same date and in the
49 same manner as chapter 699 of the laws of 2025, takes effect.
