The short version
New York’s RAISE Act is now in its registration phase: frontier AI labs with more than $500 million in revenue must sign up by January 1, 2027. The law targets model makers, not their customers. Teams doing generative AI integration on top of these models get new public documents to review and a reason to update vendor contracts.
What did New York launch on November 10?
New York launched a RAISE Act registration portal and a state-run AI testing center on the same day, Governor Kathy Hochul announced. The portal is where frontier AI developers register under the law. It is run by DIGIT, the Office of Digital Innovation, Governance, Integrity and Trust, which sits inside the New York State Department of Financial Services.
The second is the New York Center for AI Responsibility and Research at SUNY Binghamton. According to the governor’s office, more than 80 researchers from New York institutions will use the state’s Empire AI computing initiative to test advanced AI systems and review companies’ safety frameworks together with DIGIT. The center is backed by a $30 million philanthropic gift and a $25 million research investment from SUNY.
Who has to register, and what must they do?
Only large frontier developers carry the full obligations, which makes the list of registrants short. TechCrunch reports the threshold as more than $500 million in annual revenue, and the law defines frontier models by training compute (more than 1026 operations). In practice that means model labs, not the companies that build products on their APIs.
| Obligation | What the law requires |
|---|---|
| Registration | Register with DIGIT through the RAISE portal by January 1, 2027 |
| Safety framework | Publish it on the company website, starting January 2027 |
| Transparency reports | Publish public reports on the models they release |
| Incident reporting | Report critical safety incidents within 72 hours |
| Ongoing filings | Quarterly catastrophic risk assessments and periodic disclosure statements, per the governor’s September release |
| Penalties | Up to $1 million for a first violation, up to $3 million for later ones (TechCrunch) |
How does it compare with California and Illinois?
New York’s law looks a lot like California’s SB 53, which also requires large frontier developers to publish safety frameworks and report incidents. The main gap is timing: California gives 15 days to report a critical incident, New York gives 72 hours. Illinois’ SB 315, which TechCrunch says takes effect in January 2027, adds independent safety compliance audits.
All of this happens without a federal standard. Hochul framed the move as New York acting “while the federal government fails to act,” and the law was signed in December 2025, shortly after a federal executive order told the Justice Department to challenge state AI laws. A legal fight over preemption is still possible.
What it means for US & EU software teams
If you build products on frontier models, the RAISE Act isn’t your compliance problem, but it does change your inputs. From 2027, the labs you depend on will publish safety frameworks and model transparency reports under state law. These are better evidence for vendor reviews than marketing pages, and security or procurement teams can ask for them by name.
The 72-hour clock matters too. A lab will have to tell New York about a critical safety incident within three days, but nothing in the law promises that its customers hear about it at the same speed. Check what your AI vendor contracts say about incident notice, model changes and data handling.
For teams selling into both markets, the US rules now look closer to the EU’s. The EU AI Act already puts documentation and transparency duties on general-purpose AI model providers. One model inventory and one vendor evidence file can serve SB 53, the RAISE Act and the EU AI Act instead of three separate efforts.
Engadget notes one limit: until the law is fully in force, the new Binghamton center can only study what AI companies disclose publicly. Treat the first reports as a starting point, not a guarantee.
What should teams do before January 2027?
- Confirm you are out of scope. If you train your own large models, ask counsel to check the compute and revenue thresholds. Most product teams will not meet them.
- List your model vendors. Record which frontier models each product uses and where they run.
- Collect the disclosures. When labs publish their RAISE safety frameworks and transparency reports, add them to your vendor review file.
- Update contracts. Ask for incident notification terms and advance notice of model changes or retirements.
- Plan for fallback. Keep a second model option tested, so an incident at one lab doesn’t stop your product.
Frequently asked questions
When do AI developers have to register under the New York RAISE Act?
New York opened the RAISE Act registration portal on November 10, 2026. Large frontier AI developers have to register by January 1, 2027, the date the law takes effect. The portal is run by the Office of Digital Innovation, Governance, Integrity and Trust (DIGIT), a new office inside the New York State Department of Financial Services.
Which companies does the RAISE Act apply to?
The heaviest obligations apply to large frontier developers: companies that train frontier models and have more than $500 million in annual revenue, such as OpenAI and Anthropic. The law defines frontier models by training compute, at more than 1026 operations. Companies that only call AI APIs or fine-tune existing models are unlikely to meet these thresholds, but legal counsel should confirm for each case.
What does the RAISE Act require from large AI developers?
Covered developers must publish a safety framework on their website, publish transparency reports for their models, report critical safety incidents to the state within 72 hours, and file periodic catastrophic risk assessments and disclosure statements with DIGIT. TechCrunch reports fines of up to $1 million for a first violation and up to $3 million for later ones.
How is the RAISE Act different from California SB 53?
The two laws are similar in structure: both require large frontier developers to publish safety frameworks and report critical incidents. The biggest difference is speed. New York requires incident reports within 72 hours, while California allows 15 days. New York also created a dedicated oversight office, DIGIT, to implement the law.
Does the RAISE Act affect companies that build products on OpenAI or Anthropic models?
Not directly. The obligations fall on the model developers. Indirectly, it gives product teams new public documents to review, such as safety frameworks and model transparency reports, and a reason to make sure vendor contracts say how and when customers are told about incidents that affect them.
Sources
Office of the Governor of New York — Governor Hochul declares NY will lead the nation in AI science and safety (primary source, November 10, 2026)
Office of the Governor of New York — Next steps to regulate major AI developers (September 21, 2026)
TechCrunch — OpenAI, Anthropic face new oversight as New York’s AI safety law takes effect (November 10, 2026)
Engadget — New York will independently test AI models (November 10, 2026)