New York Opens RAISE Act Registration — Frontier AI Labs Must Register by 1 January 2027
New York has opened registration under its frontier AI safety law, giving the largest AI developers until 1 January 2027 to register with the state. The registration portal opened on 10 November 2026, run by a new office inside the New York Department of Financial Services. That date is not the day the law takes effect: the Responsible AI Safety and Education Act, known as the RAISE Act, takes effect on 1 January 2027, and that same date is the deadline for large frontier developers to register. A frontier AI developer is a company that builds the most advanced, general-purpose AI models, the kind that can be adapted to many different tasks and that carry the highest potential for large-scale harm if misused. New York's law is the first of its kind in the United States to put registration, public safety frameworks, and incident reporting on that class of company.
What happened on 10 November
On 10 November 2026, the Office of Digital Innovation, Governance, Integrity, and Trust, known as DIGIT, opened the RAISE registration portal. DIGIT is a new office inside the New York Department of Financial Services, the state regulator better known for overseeing banks and insurers. Until 1 January 2027, large frontier developers must file their registrations through the portal. Failure to register can draw enforcement action, so the portal is not voluntary. The opening of the portal is the practical start of the regime, even though the legal obligations themselves begin on 1 January.
The road to the RAISE Act
The law has had a long and contested path. Governor Kathy Hochul signed the RAISE Act on 19 December 2025, making New York the first US state to pass a frontier-model safety law. An amending bill, S8828, was signed on 27 March 2026, and it moved the law's effective date to 1 January 2027, giving developers and the new regulator more time to prepare. The amendments also softened the penalties. The original bill set fines of up to $10 million for a first violation and up to $30 million for subsequent violations; the negotiated version brings those down to up to $1 million for a first violation and up to $3 million for subsequent violations.
| Date | Event |
|---|---|
| 19 December 2025 | Governor Hochul signs the RAISE Act |
| 27 March 2026 | Amending bill S8828 signed; effective date moved to 1 January 2027 |
| 10 November 2026 | DIGIT opens the RAISE registration portal |
| 1 January 2027 | Law takes effect; registration deadline for large frontier developers |
| From 1 January 2027 | Safety frameworks, 72-hour incident reporting, and transparency reports required |
Who has to register
The law covers large frontier developers, defined in part by revenue of more than $500 million a year. That threshold is designed to catch the companies building the most capable models, not every startup that uses AI. TechCrunch names OpenAI and Anthropic as examples of the companies that would be covered, and both are the obvious candidates given their size and their frontier models. But the state has not publicly named its covered-registrant list, so no company should be stated as registered or as definitively covered until it files or the state says so. The revenue test is only part of the definition; the full test depends on statute text not retrieved here. What is clear is that the registration duty falls on the developers themselves, not on the companies that merely deploy or fine-tune someone else's model.
 |
| The New York State Capitol building in Albany on a clear day |
What registered developers must do
From 1 January 2027, covered developers must publish safety frameworks on their own websites, setting out how they assess and manage the risks of their most capable models. They must report critical safety incidents to the state within 72 hours, a deadline that puts AI safety incidents on a footing similar to data-breach notification rules. They must also issue public model transparency reports. Secondary summaries add that developers must produce quarterly catastrophic-risk assessments, but that detail is attributed to those summaries and should not be presented as verbatim text from the statute, since the statute itself was not retrieved. Taken together, the obligations push frontier labs towards public, written, and regularly updated accounts of their safety practices, a sharp change from the current position, where safety policies are often published voluntarily and on each company's own timetable. The industry's own voluntary commitments, such as the OpenAI Model Spec, set out how one leading lab says it will behave, while the OpenAI safety researchers letter showed how contested the question of who judges AI safety has become.
Penalties and enforcement
The penalties for breaking the law are up to $1 million for a first violation and up to $3 million for subsequent violations. Those figures are far lower than the $10 million and $30 million in the original bill, a reduction that was part of the negotiated amendments. The state can bring enforcement actions for failure to register, and the registration duty is the first obligation to bite, since the deadline and the effective date coincide on 1 January 2027. A company that misses the deadline is therefore exposed to enforcement from the very first day the law operates.
NYCAIRR: a research centre at SUNY Binghamton
On the same day the portal opened, Governor Hochul launched the New York Center for AI Responsibility and Research, known as NYCAIRR, at the State University of New York at Binghamton. The centre is backed by $55 million in donations and capital announced in January, and its job is to independently evaluate advanced AI systems and help DIGIT review the safety frameworks that developers publish. That gives the regulator an independent source of technical expertise, rather than leaving it to judge safety claims on the developers' own word. NYCAIRR's launch on the same day as the portal opening signals that New York intends the registration regime and the independent evaluation capacity to grow together.
 |
| A government office interior with stacks of documents, a folder and a laptop on a wooden desk |
How this fits New York's other AI laws
The RAISE Act is not New York's first AI law. The state has already moved on AI in consumer-facing areas, and our coverage of the New York AI chatbot law set out the rules for AI companions and chatbots that interact directly with the public. The RAISE Act operates at a different level: it targets the developers of the underlying frontier models, not the apps built on top of the m. The two laws are separate instruments with separate duties, and they should not be conflated. One governs how chatbots behave towards users; the other governs how the most capable models are developed, tested, and reported on. A company can be affected by both, but for different reasons and under different rules.
What to watch before 1 January
 |
| The New York City skyline at dusk with the Empire State Building and warm city lights |
The practical question between now and 1 January 2027 is which companies register, and when. The portal is open, the deadline is fixed, and the state has not published a list of covered developers, so the first registrations will be the first public signal of who the law actually catches. Watch for OpenAI and Anthropic, the two names most often cited, and for whether any company challenges the law or its scope. One reporting trap to avoid: some syndicated coverage carried a headline saying the law "takes effect" on 10 November, even though the body of the same report said it "kicks in at the start of the year". The body is right and the headline is wrong. The law takes effect on 1 January 2027.
Key Takeaways
- New York opened the RAISE Act registration portal on 10 November 2026, run by DIGIT inside the Department of Financial Services — the law itself takes effect on 1 January 2027.
- Large frontier developers, over $500 million in annual revenue, must register by 1 January 2027; OpenAI and Anthropic are named as examples, but the state has not published a covered-registrant list.
- From 1 January 2027, covered developers must publish safety frameworks, report critical safety incidents within 72 hours, and issue public model transparency reports.
- Penalties run to $1 million for a first violation and $3 million for subsequent violations, down from $10 million and $30 million in the original bill.
- Governor Hochul also launched NYCAIRR at SUNY Binghamton, backed by $55 million, to independently evaluate advanced AI systems and help DIGIT review safety frameworks.
Conclusion
New York's RAISE Act registration portal is open, and the clock is now running to 1 January 2027. That date is both the day the law takes effect and the deadline for the largest frontier AI developers to register. The obligations that follow — public safety frameworks, 72-hour incident reporting, and transparency reports — are the first of their kind in the United States, enforced by a new office with a new research centre behind it. The next two months will show who registers, and whether other states follow New York's lead.
Comments
Post a Comment