Regulate AI Safety Standards

Section 1. Section 1420 of the general business law, as added by chapter 96 of the laws of 2026, is amended to read as follows:

§ 1420. Definitions. As used in this article, the following terms shall have the following meanings:

1. "Affiliate" means a person controlling, controlled by, or under common control with a specified person, directly or indirectly, through one or more intermediaries.

2. "Artificial intelligence model" means an engineered or machinebased system that varies in its level of autonomy and that can, for explicit or implicit objectives, infer from the input it receives how to generate outputs that can influence physical or virtual environments.

3. (a) "Catastrophic risk" means a foreseeable and material risk that a frontier developer's development, storage, use, or deployment of a frontier model will materially contribute to the death of, or serious injury to, more than fifty people or more than one billion dollars in damage to, or loss of, property arising from a single incident or multiple incidents involving a frontier model doing any of the following:

(i) providing expert-level assistance in the creation or release of a chemical, biological, radiological, or nuclear weapon;

(ii) engaging in conduct with [no meaningful] limited human oversight, intervention, or supervision that is either a cyberattack or, if the conduct had been committed by a human, would constitute the crime of murder, assault, extortion, or theft, including theft by false pretense; [or]

(iii) attempting to overthrow or seize control of the federal government or state or local governments in the state of New York; or

(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

(b) "Catastrophic risk" does not include a foreseeable and material risk from any of the following:

(i) information that a frontier model outputs if the information is otherwise publicly accessible in a substantially similar form from a source other than a foundation model;

(ii) lawful activity of the federal government; or

(iii) harm caused by a frontier model in combination with other software if the frontier model did not materially contribute to the harm.

4. "Covered emergency" means a situation that poses an imminent catastrophic risk or security risk. "Covered emergency" shall be construed to include a situation in which immediate action is necessary to prevent a catastrophic risk or security risk, even if the harm that would be prevented is not proximate in time to the situation.

5. "Critical safety incident" means any of the following:

(a) unauthorized access to, modification of, or exfiltration of[, the] unreleased model weights of a frontier model [that results in death or bodily injury];

(b) harm resulting from the materialization of a catastrophic risk;

(c) [loss of control of a frontier model causing death or bodily injury; or

(d) a frontier model that uses deceptive techniques against the frontier developer to subvert the controls or monitoring of its frontier developer outside of the context of an evaluation designed to elicit this behavior and in a manner that demonstrates materially increased catastrophic risk] conduct by a foundation model of a frontier developer that circumvents, disables, evades, or attempts to circumvent, disable, or evade, a material, technical or organizational control of its frontier developer that limits the model's access to systems, resources, or data, provides for its suspension or modification, or monitors its conduct or capabilities, outside of the context of an evaluation designed to elicit such conduct where such conduct remains confined to the evaluation environment;

(d) conduct by a frontier model that, without the intent of any user, materially impairs the confidentiality, integrity, or availability of a computer system or computer network which is not owned or operated by the frontier developer or user and where such conduct occurs without the consent of the owner or operator of such system or network;

(e) an occurrence in which a safeguard against an incident described in this subdivision failed, was circumvented, or was materially degraded, such that, absent timely intervention, mitigation, or fortuity, there was a substantial likelihood that such a critical safety incident would have resulted; or

(f) any other occurrence or condition designated by regulation of the office which is related to catastrophic risk, security risk, or the automation of AI research and development.

6. "Department" means the department of financial services.

[5.] 7. (a) "Deploy" means to make a frontier model available to a third party for use, modification, copying, or combination with other software.

(b) "Deploy" does not include making a frontier model available to a third party for the primary purpose of developing or evaluating the frontier model.

[6.] 8. "Foundation model" means an artificial intelligence model that is all of the following:

(a) trained on a broad data set;

(b) designed for generality of output; and

(c) adaptable to a wide range of distinctive tasks.

[7.] 9. "Frontier AI framework" means documented technical and organizational protocols to:

(a) manage, assess, and mitigate catastrophic risks; and

(b) assess the ability of frontier models to automate AI research and development and to manage any catastrophic risks and security risks associated with such automation.

[8.] 10. "Frontier developer" means a person who has trained, or initiated the training of, a frontier model, with respect to which the person has used, or intends to use, at least as much computing power to train the frontier model as would meet the technical specifications found in subdivision [nine] twelve of this section.

11. "Frontier examiner" means a third party that performs services under section fourteen hundred twenty-four of this article and that is eligible to be a frontier examiner under such section.

[9.] 12. (a) "Frontier model" means a foundation model that was trained using a quantity of computing power greater than 10º[26] 25 integer or floating-point operations, provided, however, that foundation models shall be considered to be frontier models if the office has determined that the foundation model:

(i) is among the most capable foundation models;

(ii) possesses capabilities which pose catastrophic risk; and

(iii) is reasonably likely to possess the ability to automate AI research and development.

(b) The quantity of computing power described in paragraph (a) of this subdivision shall include computing for the original training run and for any subsequent fine-tuning, reinforcement learning, or other material modifications the developer applies to a preceding foundation model.

13. "Internal use" means any use of a frontier model by its frontier developer or its affiliates, including, but not limited to, any of the following:

(a) the training of the frontier model;

(b) modifying or fine-tuning the frontier model;

(c) using the frontier model for internal business purposes;

(d) evaluating the frontier model; and

(e) using the frontier model for AI research and development.

[10.] 14. "Large frontier developer" means a frontier developer that together with its affiliates collectively [had annual gross revenues in excess of five hundred million dollars in the preceding calendar year] incurred more than one billion dollars in expenditures for research, development, training, fine-tuning, modification, security, evaluation, or other activities which have contributed to the development, creation, or modification of one or more frontier models in the preceding thirtysix months.

[11.] 15. "Model weight" means a numerical parameter in a frontier model that is adjusted through training and that helps determine how inputs are transformed into outputs.

[12. "Department" means the department of financial services.

13. "Property" means tangible or intangible property.

14. "Person" means an individual, proprietorship, firm, partnership, joint venture, syndicate, business trust, company, corporation, limited liability company, association, committee, or any other nongovernmental organization or group of persons acting in concert.

15.] 16. "Office" means an office within the department of financial services, which shall report to the superintendent of financial services and is tasked with implementation of this article.

17. "Open source release" means making the model weights of a foundation model publicly available under a license that grants to all persons, without payment of a fee or royalty and without restriction or discrimination based on the identity of any person or group of persons or on the field of endeavor, the rights to access, use, modify, and redistribute the foundation model or modified versions of the foundation model. A license does not fail to satisfy this subdivision solely because it requires attribution to the frontier developer or requires that modified or redistributed versions be made available under identical or comparable terms.

18. "Person" means an individual, proprietorship, firm, partnership, joint venture, syndicate, business trust, company, corporation, limited liability company, association, committee, or any other nongovernmental organization or group of persons acting in concert.

19. "Property" means tangible or intangible property.

20. "Relevant information" means all information reasonably necessary for a frontier examiner to comply with section fourteen hundred twentyfour of this article, including, but not limited to:

(a) all unredacted versions of materials published pursuant to this article;

(b) access to frontier models sufficient to enable assessments under section fourteen hundred twenty-four of this article;

(c) interviews with employees of a large frontier developer;

(d) internal documents of a large frontier developer; and

(e) information to which an employee of a large frontier developer whose responsibilities include assessing the capabilities of frontier models or managing catastrophic risk or security risk has access in the ordinary course of employment is relevant information unless the large frontier developer establishes that it is not reasonably necessary for the frontier examiner to perform its functions under section fourteen hundred twenty-four of this article.

21. "Specification" means a high-level, general description of the principles that the model is intended to follow, how the model is intended to prioritize different principles and user prompts, and the topics for which the model is intended to refuse, redirect, or otherwise subvert user prompts.

22. "Security risk" means a foreseeable and material risk of unauthorized access to, modification of, or transfer of the unreleased model weights of a frontier model by any person, entity, or model.

23. "Superintendent" means the superintendent of financial services.

[16. "Office" means an office within the department of financial services, which shall report to the superintendent of financial services and is tasked with implementation of this article.]

§ 2. Section 1421 of the general business law, as added by chapter 96 of the laws of 2026, is amended to read as follows:

§ 1421. Transparency requirements. 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;

(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 or a security risk or whether a frontier model has the capability to automate AI research and development, which may include multiple-tiered thresholds;

(c) applying mitigations to address the potential for catastrophic or security risks based on the results of assessments undertaken pursuant to paragraph (b) of this subdivision;

(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;

(f)] revisiting and updating the frontier AI framework, including any criteria that trigger updates and how the large frontier developer determines when its frontier models are substantially modified enough to require disclosures pursuant to subdivision [three] four of this section;