STATE OF NEW YORK ________________________________________________________________________ 6453 2025-2026 Regular Sessions IN ASSEMBLY March 5, 2025 ___________ Introduced by M. of A. BORES -- read once and referred to the Committee on Science and Technology AN ACT to amend the general business law, in relation to the training and use of artificial intelligence frontier models The People of the State of New York, represented in Senate and Assem- bly, do enact as follows: 1 Section 1. Short title. This act shall be known and may be cited as 2 the "Responsible AI safety and education act" or "RAISE act". 3 § 2. The general business law is amended by adding a new article 44-B 4 to read as follows: 5 ARTICLE 44-B 6 RESPONSIBLE AI SAFETY AND EDUCATION (RAISE) ACT 7 Section 1420. Definitions. 8 1421. Transparency requirements regarding frontier model train- 9 ing and use. 10 1422. Protections, rights and obligations of employees. 11 1423. Violations. 12 1424. Duties and obligations. 13 1425. Preemption. 14 § 1420. Definitions. As used in this article, the following terms 15 shall have the following meanings: 16 1. "Appropriate redactions" means redactions to a safety and security 17 protocol or audit report that are reasonably necessary to protect any of 18 the following: 19 (a) Public safety or United States national security; 20 (b) Trade secrets; or 21 (c) Confidential information pursuant to state and federal law. 22 2. "Artificial intelligence" means a machine-based system that can, 23 for a given set of human-defined objectives, make predictions, recommen- 24 dations, or decisions influencing real or virtual environments, and that 25 uses machine- and human-based inputs to perceive real and virtual envi- EXPLANATION--Matter in italics (underscored) is new; matter in brackets [] is old law to be omitted. LBD00047-07-5A. 6453 2 1 ronments, abstract such perceptions into models through analysis in an 2 automated manner, and use model inference to formulate options for 3 information or action. 4 3. "Artificial intelligence model" means an information system or 5 component of an information system that implements artificial intelli- 6 gence technology and uses computational, statistical, or machine-learn- 7 ing techniques to produce outputs from a given set of inputs. 8 4. "Compute cost" means the cost incurred to pay for compute used in 9 training a model when calculated using the average market prices of 10 cloud compute in the United States at the start of training such model 11 as reasonably assessed by the person doing the training. 12 5. "Deploy" means to use a frontier model or to make a frontier model 13 foreseeably available to one or more third parties for use, modifica- 14 tion, copying, or a combination thereof with other software, except as 15 reasonably necessary for training or developing the frontier model, 16 evaluating the frontier model or other frontier models, or complying 17 with federal or state laws. 18 6. "Frontier model" means either of the following: 19 (a) an artificial intelligence model trained using greater than 10�26 20 computational operations (e.g., integer or floating-point operations), 21 the compute cost of which exceeds one hundred million dollars. 22 (b) an artificial intelligence model produced by applying knowledge 23 distillation to a frontier model as defined in paragraph (a) of this 24 subdivision. 25 7. "Critical harm" means the death or serious injury of one hundred or 26 more people or at least one billion dollars of damages to rights in 27 money or property caused or materially enabled by a large developer's 28 creation, use, storage, or release of a frontier model, through either 29 of the following: 30 (a) The creation or use of a chemical, biological, radiological, or 31 nuclear weapon; or 32 (b) An artificial intelligence model engaging in conduct that does 33 both of the following: 34 (i) Acts with limited human intervention; and 35 (ii) Would, if committed by a human, constitute a crime specified in 36 the penal law that requires intent, recklessness, or gross negligence, 37 or the solicitation or aiding and abetting of such a crime. 38 A harm inflicted by an intervening human actor shall not be deemed to 39 result from a developer's activities unless such activities made it 40 substantially easier or more likely for the actor to inflict such harm. 41 8. "Knowledge distillation" means any supervised learning technique 42 that uses a larger artificial intelligence model or the output of a 43 larger artificial intelligence model to train a smaller artificial 44 intelligence model with similar or equivalent capabilities as the larger 45 artificial intelligence model. 46 9. "Large developer" means a person that has trained at least one 47 frontier model, the compute cost of which exceeds five million dollars, 48 and has spent over one hundred million dollars in compute costs in 49 aggregate in training frontier models. Accredited colleges and univer- 50 sities shall not be considered large developers under this article to 51 the extent that such colleges and universities are engaging in academic 52 research. If a person subsequently transfers full intellectual property 53 rights of the frontier model to another person (including the right to 54 resell the model) and retains none of those rights for themself, then 55 the receiving person shall be considered the large developer and shallA. 6453 3 1 be subject to the responsibilities and requirements of this article 2 after such transfer. 3 10. "Model weight" means a numerical parameter in an artificial intel- 4 ligence model that is adjusted through training and that helps determine 5 how inputs are transformed into outputs. 6 11. "Person" means an individual, proprietorship, firm, partnership, 7 joint venture, syndicate, business trust, company, corporation, limited 8 liability company, association, committee, or any other nongovernmental 9 organization or group of persons acting in concert. 10 12. "Safety and security protocol" means documented technical and 11 organizational protocols that: 12 (a) Specify reasonable protections and procedures that, if successful- 13 ly implemented would appropriately reduce the risk of critical harm; 14 (b) Describe reasonable administrative, technical, and physical 15 cybersecurity protections for frontier models within the large develop- 16 er's control that, if successfully implemented, appropriately reduce the 17 risk of unauthorized access to, or misuse of, the frontier model leading 18 to critical harm, including by sophisticated actors; 19 (c) Describe in detail the testing procedure to evaluate if the fron- 20 tier model poses an unreasonable risk of critical harm; 21 (d) Describe in detail how the testing procedure assesses whether the 22 frontier model could be misused, modified, combined with other software 23 or be used to create another foundation model in a manner that would 24 produce critical risk; 25 (e) State compliance requirements with sufficient detail and specif- 26 icity to allow the large developer or a third party to readily ascertain 27 whether the requirements of the safety and security protocol have been 28 followed; 29 (f) Describe how the large developer will fulfill their obligations 30 under this article, including with respect to any requirements, safe- 31 guards, or modifications; and 32 (g) Designate senior personnel to be responsible for ensuring compli- 33 ance. 34 13. "Safety incident" means an incident of the following kinds that 35 occurs in such a way that it provides demonstrable evidence of an 36 increased risk of critical harm: 37 (a) A frontier model autonomously engaging in behavior other than at 38 the request of a user; 39 (b) Theft, misappropriation, malicious use, inadvertent release, unau- 40 thorized access, or escape of the model weights of a frontier model; 41 (c) The critical failure of any technical or administrative controls, 42 including controls limiting the ability to modify a frontier model; or 43 (d) Unauthorized use of a frontier model. 44 14. "Trade secret" means any form and type of financial, business, 45 scientific, technical, economic, or engineering information, including a 46 pattern, plan, compilation, program device, formula, design, prototype, 47 method, technique, process, procedure, program, or code, whether tangi- 48 ble or intangible, and whether or how stored, compiled, or memorialized 49 physically, electronically, graphically, photographically or in writing, 50 that: 51 (a) Derives independent economic value, actual or potential, from not 52 being generally known to, and not being readily ascertainable by proper 53 means by, other persons who can obtain economic value from its disclo- 54 sure or use; and 55 (b) Is the subject of efforts that are reasonable under the circum- 56 stances to maintain its secrecy.A. 6453 4 1 § 1421. Transparency requirements regarding frontier model training 2 and use. 1. Before deploying a frontier model, the large developer of 3 such frontier model shall do all of the following: 4 (a) Implement a written safety and security protocol; 5 (b) Retain an unredacted copy of the safety and security protocol, 6 including records and dates of any updates or revisions. Such unredacted 7 copy of the safety and security protocol, including records and dates of 8 any updates or revisions, shall be retained for as long as the frontier 9 model is deployed plus five years; 10 (c) (i) Conspicuously publish a copy of the safety and security proto- 11 col with appropriate redactions and transmit a copy of such redacted 12 safety and security protocol to the attorney general; 13 (ii) Grant the attorney general access to the unredacted safety and 14 security protocol upon request; 15 (d) Record, as and when reasonably possible, and retain for as long as 16 the frontier model is deployed plus five years information on the 17 specific tests and test results used in any assessment of the frontier 18 model that provides sufficient detail for third parties to replicate the 19 testing procedure; and 20 (e) Implement appropriate safeguards to prevent unreasonable risk of 21 critical harm. 22 2. A large developer shall not deploy a frontier model if doing so 23 would create an unreasonable risk of critical harm. 24 3. A large developer shall conduct an annual review of any safety and 25 security protocol required by this section to account for any changes 26 to the capabilities of their frontier models and industry best practices 27 and, if necessary, make modifications to such safety and security proto- 28 col. If any modifications are made, the large developer shall publish 29 the safety and security protocol in the same manner as required pursuant 30 to paragraph (c) of subdivision one of this section. 31 4. (a) Beginning on the effective date of this article, or ninety days 32 after a developer first qualifies as a large developer, whichever is 33 later, a large developer shall annually retain a third party to perform 34 an independent audit of compliance with the requirements of this 35 section. Such third party shall conduct audits consistent with best 36 practices. 37 (b) The third party shall be granted access to unredacted materials as 38 necessary to comply with the third party's obligations under this subdi- 39 vision. 40 (c) The third party shall produce a report including all of the 41 following: 42 (i) A detailed assessment of the large developer's steps to comply 43 with the requirements of this section; 44 (ii) If applicable, any identified instances of noncompliance with the 45 requirements of this section, and any recommendations for how the devel- 46 oper can improve its policies and processes for ensuring compliance with 47 the requirements of this section; 48 (iii) A detailed assessment of the large developer's internal 49 controls, including its designation and empowerment of senior personnel 50 responsible for ensuring compliance by the large developer, its employ- 51 ees, and its contractors; and 52 (iv) The signature of the lead auditor certifying the results of the 53 audit. 54 (d) The large developer shall retain an unredacted copy of the report 55 for as long as the frontier model is deployed plus five years.A. 6453 5 1 (e) (i) The large developer shall conspicuously publish a copy of the 2 third party's report with appropriate redactions and transmit a copy of 3 such redacted report to the attorney general. 4 (ii) The large developer shall grant the attorney general access to 5 the unredacted third party's report upon request. 6 5. A large developer of a frontier model shall submit to the attorney 7 general the updated total compute cost used to train their frontier 8 models simultaneously with their annual third party's audit report. 9 6. A large developer shall disclose each safety incident affecting the 10 frontier model to the attorney general within seventy-two hours of the 11 large developer learning of the safety incident or within seventy-two 12 hours of the large developer learning facts sufficient to establish a 13 reasonable belief that a safety incident has occurred. Such disclosure 14 shall include: (a) the date of the safety incident; (b) the reasons the 15 incident qualifies as a safety incident as defined in subdivision thir- 16 teen of section fourteen hundred twenty of this article; and (c) a short 17 and plain statement describing the safety incident. 18 7. (a) This section shall not apply to products or services to the 19 extent that the requirements would strictly conflict with the terms of a 20 contract with a federal government entity and a large developer. 21 (b) This section applies to the development or deployment of a fron- 22 tier model for any use that is not the subject of a contract with a 23 federal government entity, even if such frontier model has already been 24 developed, trained, or used by a federal government entity. 25 8. A large developer shall not knowingly make false or materially 26 misleading statements or omissions in or regarding documents produced 27 pursuant to this section. 28 9. Any person who is not a large developer, but who sets out to train 29 a frontier model that if completed as planned would qualify such person 30 as a large developer (i.e. at the end of the training, such person will 31 have spent five million dollars in compute costs on one frontier model 32 and one hundred million dollars in compute costs in aggregate in train- 33 ing frontier models, excluding accredited colleges and universities to 34 the extent such colleges and universities are engaging in academic 35 research) shall, before training such model: 36 (a) Implement a written safety and security protocol, excluding the 37 requirements described in paragraphs (c) and (d) of subdivision twelve 38 of section fourteen hundred twenty of this article; and 39 (b) Conspicuously publish a copy of the safety and security protocol, 40 with appropriate redactions, and transmit a copy of such redacted safety 41 and security protocol to the attorney general. 42 § 1422. Protections, rights and obligations of employees. 1. A large 43 developer or a contractor or subcontractor of a large developer shall 44 not prevent an employee from disclosing, or threatening to disclose, or 45 retaliate against an employee for disclosing or threatening to disclose, 46 information to the large developer or the attorney general, if the 47 employee has reasonable cause to believe that the large developer's 48 activities pose an unreasonable or substantial risk of critical harm, 49 regardless of the employer's compliance with applicable law. 50 2. An employee harmed by a violation of this section may petition a 51 court for appropriate temporary or preliminary injunctive relief. 52 3. A large developer shall inform employees of their protections, 53 rights and obligations under this article by posting a notice thereof. 54 Such notice shall be posted conspicuously in easily accessible and well- 55 lighted places customarily frequented by employees.A. 6453 6 1 4. Nothing in this section shall be deemed to diminish the rights, 2 privileges, or remedies of any employee under any other law or regu- 3 lation or under any collective bargaining agreement or employment 4 contract. 5 5. As used in this section, the following terms shall have the follow- 6 ing meanings: 7 (a) "Employee" has the same meaning as defined in subdivision five of 8 section two of the labor law and includes both of the following: 9 (i) Contractors or subcontractors and unpaid advisors involved with 10 assessing, managing, or addressing the risk of critical harm from fron- 11 tier models; and 12 (ii) Corporate officers. 13 (b) "Contractor or subcontractor" means any person, sole proprietor, 14 partnership, firm, corporation, limited liability company, association 15 or other legal entity who by oneself or through others offers to under- 16 take, or holds oneself out as being able to undertake, or does undertake 17 work assessing, managing, or addressing the risk of critical harm from 18 frontier models on behalf of the large developer. 19 § 1423. Violations. 1. The attorney general may bring a civil action 20 for a violation of this article and to recover all of the following: 21 (a) For a violation of section fourteen hundred twenty-one of this 22 article, a civil penalty in an amount not exceeding five percent of the 23 total compute cost used to train the large developer's frontier models 24 as reported by the large developer pursuant to this article for a first 25 violation and in an amount not exceeding fifteen percent of that value 26 for any subsequent violation. 27 (b) For a violation of section fourteen hundred twenty-two of this 28 article, a civil penalty in an amount not exceeding ten thousand dollars 29 per employee for each violation of such section to be awarded to the 30 employee who was retaliated against. 31 (c) For a violation of section fourteen hundred twenty-one or fourteen 32 hundred twenty-two of this article, injunctive or declaratory relief. 33 2. (a) A provision within a contract or agreement that seeks to waive, 34 preclude, or burden the enforcement of a liability arising from a 35 violation of this article, or to shift that liability to any person or 36 entity in exchange for their use or access of, or right to use or 37 access, a large developer's products or services, including by means of 38 a contract of adhesion, is void as a matter of public policy. 39 (b) A court shall disregard corporate formalities and impose joint and 40 several liability on affiliated entities for purposes of effectuating 41 the intent of this section to the maximum extent allowed by law if the 42 court concludes that both of the following are true: 43 (i) The affiliated entities, in the development of the corporate 44 structure among the affiliated entities, took steps to purposely and 45 unreasonably limit or avoid liability; and 46 (ii) As the result of the steps described in subparagraph (i) of this 47 paragraph, the corporate structure of the large developer or affiliated 48 entities would frustrate recovery of penalties, damages, or injunctive 49 relief under this section. 50 3. This section does not limit the application of other laws. 51 § 1424. Duties and obligations. The duties and obligations imposed by 52 this article are cumulative with any other duties or obligations imposed 53 under other law and shall not be construed to relieve any party from any 54 duties or obligations imposed under other law and do not limit any 55 rights or remedies under existing law.A. 6453 7 1 § 1425. Preemption. This article shall not apply to the extent that it 2 is preempted by federal law. 3 § 3. This act shall take effect on the ninetieth day after it shall 4 have become a law.
Policy Tracker
Relates to the training and use of artificial intelligence frontier models; defines terms; establishes remedies for violations.
NY · Legislation · 2025 · A06453
LegislationAI
Introduced
Record updated Jun 12, 2025
Summary
Relates to the training and use of artificial intelligence frontier models; defines terms; establishes remedies for violations.
Timeline
2025-06-12
A
reported referred to rules
2025-06-12
A
reported
2025-06-12
A
rules report cal.656
2025-06-12
A
ordered to third reading rules cal.656
2025-06-12
A
substituted by s6953b
2025-06-11
A
reference changed to ways and means
2025-06-09
A
amend and recommit to codes
2025-06-09
A
print number 6453b
Bill Text
Rendered HTML Filing
Official document markup is preserved and restyled to match the active site theme.