NY S06953: Relates to the training and use of artificial intelligence frontier models; defines terms; establishes remedies for violations.

In one paragraph

New York's RAISE Act (S 6953-B, Chapter 699 of 2025) as signed covered "large developers" spending over $100 million on compute, required safety protocols and set penalties up to $10 million, then $30 million. Chapter 96 of 2026 (signed March 27, 2026) repealed and replaced Article 44-B and set the effective date at January 1, 2027. From then, developers of frontier models (over 10^26 training operations) must publish transparency reports when deploying and report critical safety incidents to a Department of Financial Services office within 72 hours; those with over $500 million in revenue must also publish a frontier AI framework and file disclosure statements. Signed December 19, 2025.

Plain-English summary, written with AI assistance. The official text controls.

What it means for business

From 2027, developers of frontier models (over 10^26 training operations) developed, deployed or operating in New York must publish transparency reports and report critical safety incidents to DFS; those with over $500 million in revenue must also publish a frontier AI framework and file with DFS.

Key provisions

1. 1As signed, Chapter 699's Article 44-B covered "large developers" (trained a frontier model and spent over $100 million in aggregate compute costs) and required a published safety and security protocol, barred deployment creating an unreasonable risk of critical harm, required 72-hour incident reports to the Attorney General and DHSES, and set penalties up to $10 million, then $30 million
2. 2Chapter 96 of 2026 (signed March 27, 2026) repealed that Article 44-B and enacted a new one (General Business Law §§ 1420-1429), and changed Chapter 699's effective date from the 90th day after signing to January 1, 2027
3. 3Frontier model: a foundation model trained with more than 10^26 integer or floating-point operations, counting later fine-tuning, reinforcement learning and other material modifications; a large frontier developer also had, with affiliates, over $500 million in gross revenue the preceding calendar year
4. 4Large frontier developers must write, follow and publish a frontier AI framework for catastrophic risk (more than 50 deaths or serious injuries, or over $1 billion in property damage, from one incident in which a model gives expert-level help with chemical, biological, radiological or nuclear weapons, carries out a cyberattack or acts like murder, assault, extortion or theft without meaningful human oversight, or evades control), review it yearly and publish material changes within 30 days
5. 5Before or when deploying a new or substantially modified frontier model, a frontier developer must publish a transparency report (contact, release date, languages, output modalities, intended uses, restrictions); large frontier developers add summaries of catastrophic-risk assessments and third-party involvement
6. 6Critical safety incidents must be reported to the DFS office within 72 hours; one posing an imminent risk of death or serious physical injury must be disclosed within 24 hours to an appropriate authority, such as a law enforcement or public safety agency; large frontier developers send internal-use risk assessment summaries every three months or on an agreed schedule
7. 7Large frontier developers need a current disclosure statement on file with the office (renewed every two years or on a change of ownership or material change) and must pay pro rata assessments (the office may impose $1,000 a day, after notice and hearing, for not filing); Attorney General penalties against large frontier developers up to $1 million, then $3 million per violation; no private right of action; academic research by accredited New York colleges and Empire AI exempt
8. 8Signed December 19, 2025 (Chapter 699, Approval Memo 76); as amended by Chapter 96 of 2026, effective January 1, 2027

Latest legislative action

Dec 19, 2025. Signed Chapter 699 of 2025 (Approval Memo 76)

Sponsors (showing 5 of 24)

Roll call votes