New York RAISE Act: frontier AI developers register by Jan 1, 2027, 72-hour incident reports

New York's RAISE Act, the state's frontier AI safety law, moved from paper to practice this week: TechCrunch reports the state opened a registration portal on 10 November 2026, and large frontier AI developers have until the law takes effect on 1 January 2027 to register. The law makes companies such as OpenAI and Anthropic publish safety frameworks, report critical safety incidents within 72 hours, and answer to a new office inside the state's financial regulator. Most readers are not covered by it, but it is the second US state frontier-model law after California's, and its incident clock is far shorter.

Key points

• New York Governor Kathy Hochul signed the final version of the RAISE Act as a chapter amendment on 27 March 2026; it takes effect on 1 January 2027.
• It covers "large frontier developers": companies with more than $500 million in annual revenue that train a model using more than 10^26 operations of computing power (thresholds copied from California's SB 53).
• Critical safety incidents must be reported within 72 hours, against 15 days under California's law; incidents with an imminent risk of death or serious injury need notice to authorities within 24 hours.
• The state attorney general can seek up to $1 million for a first violation and up to $3 million for later ones; there is no private right of action.
• Oversight sits with DIGIT, the Office of Digital Innovation, Governance, Integrity and Trust, within the Department of Financial Services (DFS).

What happened

On 21 September 2026 Governor Hochul's office announced the next steps for implementing the law, saying New York would direct large frontier AI developers to register with the state starting in November, with disclosure statements and compliance duties beginning in January 2027. The release names DIGIT as the receiving office, says DFS Acting Superintendent Kaitlin Asrow leads implementation, and says Marc Gilman is the Act's Deputy Director and DIGIT's first full-time hire.

On 10 November, TechCrunch reported that the registration portal had opened, with large frontier developers required to register by 1 January 2027. TechCrunch also reports that the New York Center for AI Responsibility and Research, based at SUNY Binghamton, launched the same day. It is meant to evaluate advanced AI systems independently and help DIGIT review companies' safety frameworks, and was announced in January with $55 million in donations and capital. A state spokesperson declined to give TechCrunch specifics about enforcement against companies that fail to register. We could not open a state page confirming the portal's opening date, so the 10 November date rests on TechCrunch.

Background

The law has changed since it first passed the legislature. The original bill was reworked by a chapter amendment on 27 March 2026, which law firm Wiley summarises. The amendment aligned the definitions with California's Transparency in Frontier Artificial Intelligence Act (SB 53), which took effect on 1 January 2026, replaced an earlier compute-cost test with the 10^26 operations measure, cut penalties from $10 million and $30 million to $1 million and $3 million, dropped an annual third-party audit requirement, and moved oversight to a new office in DFS, according to the Cooley analysis. TechCrunch adds that Illinois SB 315, effective at the start of 2027, has similar requirements plus independent safety compliance audits.

How it works

A "frontier model" is a foundation model trained with more than 10^26 integer or floating-point operations, counting later fine-tuning and reinforcement learning (Cooley). A "large frontier developer" is one with more than $500 million in prior-year revenue, including affiliates. Large developers must:

• Publish a frontier AI framework on their website covering risk thresholds, mitigations, third-party assessments, cybersecurity, governance and incident response, updated annually or after material changes.
• Publish a transparency report before deploying a new or substantially modified model. A model or system card can serve if it contains the required information.
• Send DIGIT quarterly summaries of catastrophic-risk assessments for internal use of their models.
• Report critical safety incidents within 72 hours, and notify law enforcement or a public safety agency within 24 hours when an incident poses imminent risk of death or serious injury.
• File a disclosure statement every two years, or when ownership changes, listing business names, New York addresses and owners with 5% or more.

"Catastrophic risk" means a foreseeable, material risk of one incident killing or seriously injuring more than 50 people or causing over $1 billion in property damage, including expert-level help with chemical, biological, radiological or nuclear weapons, unsupervised cyberattacks, or a model evading its developer's control (Wiley). The public can also report suspected critical incidents to DIGIT.

New York RAISE Act72

California SB 53360

Deadline to report a critical safety incident (15 days shown as 360 hours) (hours)

	hours
New York RAISE Act	72
California SB 53	360

Source: Wiley and Cooley summaries of the NY RAISE Act and California SB 53

What it means

For the handful of labs above the thresholds, the work is concrete: a registered contact, a published framework, a 72-hour incident process, and quarterly internal-risk summaries. The short incident clock is the biggest practical difference from California, and it will force labs to decide in advance what counts as "critical", which the law defines around weight theft, loss of control and deceptive behaviour that materially raises catastrophic risk.

For everyone else, there is little direct effect. Smaller developers, application builders and anyone fine-tuning open models below the thresholds are not "large frontier developers". But the law sets a template: because the definitions match California's, one compliance programme can cover both, and other states, such as Illinois, are following. Hochul has said she will explore ways to build on the Act in coming months.

What we don't know yet

• Which companies have registered; TechCrunch names Anthropic and OpenAI as examples of firms covered, not as confirmed registrants.
• How the state will enforce against a company that fails to register. A spokesperson would not say.
• What DIGIT's rulemaking will add; the office has broad rulemaking authority under the amendment.
• How the 72-hour rule will be tested in practice, since no critical incident has yet been reported under it.

What to watch next

Watch DFS and DIGIT for rules and guidance before 1 January 2027, the first published frameworks, the Binghamton centre's early evaluations, and whether Congress or other states move to align or pre-empt.

Sources

• Governor Hochul: AI safety, next steps to regulate major AI developers (21 September 2026)
• TechCrunch: OpenAI, Anthropic face new oversight as New York's AI safety law takes effect
• Wiley: New York finalizes RAISE Act for frontier AI models
• Cooley: New York's frontier AI law gets a California makeover