New York opens RAISE Act registration for OpenAI, Anthropic and others
New York opened registration under its RAISE Act on November 10, 2026. The law is one of the strictest frontier AI safety laws in the US. Large frontier AI developers, including OpenAI and Anthropic, must register with the state by January 1, 2027, the day the law takes effect. After that they must publish safety frameworks, report critical safety incidents within 72 hours and publish transparency reports for their models. Fines go up to $3 million. Governor Kathy Hochul announced the launch together with a new state-backed AI safety research center, and TechCrunch reported which labs are affected.
What’s new
The RAISE Act passed in 2025 and was amended this spring to bring it closer to California’s frontier AI law. Hochul signed the amendment on March 27, 2026. On Tuesday the state opened the RAISE Portal, where companies register. A new office runs it: the Office of Digital Innovation, Governance, Integrity, and Trust (DIGIT), part of the Department of Financial Services.
| What | Details |
|---|---|
| Who must register | “Large frontier developers”: over $500 million in annual gross revenue in the prior year, counting affiliates |
| What counts as a frontier model | A foundation model trained with more than 10^26 operations, including later fine-tuning and reinforcement learning |
| Registration deadline | January 1, 2027, through the RAISE Portal |
| Law takes effect | January 1, 2027 |
| Incident reporting | Critical safety incidents within 72 hours; 24 hours when there is an imminent risk of death or serious injury |
| Publishing rules | Safety frameworks on the developer’s website, plus model transparency reports |
| Penalties | Up to $1 million for a first violation, $3 million for later ones |
| Regulator | DIGIT, inside the Department of Financial Services |
The thresholds, the 24-hour rule and the penalties come from Cooley’s analysis of the amended law. The registration deadline and DIGIT’s role are in the governor’s announcement. According to TechCrunch, a spokesperson said the state can take enforcement action against companies that don’t register, but would not say what that would look like.
On the same day, New York launched the New York Center for AI Responsibility and Research at SUNY Binghamton. More than 80 researchers from across the state will independently test frontier AI systems, and the center will help DIGIT review companies’ safety frameworks. Its funding is a $30 million gift from donors led by Bloomberg LP co-founder Tom Secunda, $25 million from SUNY, and state operating money. Its compute comes from Empire AI, New York’s $500 million public-private computing consortium.
“While the federal government fails to act, New York is leading the way on AI safety and holding big tech accountable,” Hochul said in the announcement.
What it means for developers
Most developers are not covered. The revenue threshold of $500 million limits registration to the biggest labs, and the compute threshold means only very large training runs make a model a frontier model. A startup fine-tuning an open-weight model, or an app built on the OpenAI or Claude APIs, has nothing to register.
What changes for everyone else is what the labs have to publish. From January, the companies behind the models most developers build on (OpenAI, Anthropic, Google and others) must post safety frameworks and model transparency reports that New York can check. When you pick a model for a product, that gives you documents to read on how it was tested and what risks the lab found. Our recent post on GPT-6 Astra’s safety research shows how much of that has so far been up to the labs themselves.
Teams that train their own large models should check two things. First, the compute count includes fine-tuning and reinforcement learning done after the first training run. Second, the law applies to models developed, deployed or operating in New York, so a company doesn’t need an office there to be covered.
How it compares
California’s Transparency in Frontier Artificial Intelligence Act (SB 53) took effect on January 1, 2026, and the amended RAISE Act follows it closely. The main differences, according to Cooley and TechCrunch:
- Incident reporting: California gives 15 days; New York gives 72 hours.
- Penalties: California caps a violation at $1 million; New York allows $3 million for repeat violations.
- Disclosure: New York adds a filing for large frontier developers, renewed every two years, that names beneficial owners and comes with fees.
- Whistleblowers: California has whistleblower protections for frontier lab staff; the RAISE Act doesn’t.
TechCrunch also notes that Illinois has a similar law, SB 315, which takes effect at the start of 2027 and adds independent safety compliance audits.
Sources
- Governor Hochul Declares NY Will Lead the Nation in AI Science and Safety; Launches New Research Institute (New York State, November 10, 2026)
- OpenAI, Anthropic face new oversight as New York’s AI safety law takes effect (TechCrunch)
- New York’s Frontier AI Law Gets a California Makeover, With Some Key Differences (Cooley)