New York RAISE Act registration portal opens for AI firms

Three dates matter here, and they’re easy to mix up. New York’s RAISE Act became law in 2025. Its registration portal opened this week. Its actual safety rules don’t bind anyone until January 1, 2027. I had to read the state’s own page twice to keep those straight, so here’s the order.

According to a Department of Financial Services (DFS) release, Governor Kathy Hochul announced in September that New York would “direct large frontier AI developers to register with the State” starting in November, ahead of full compliance beginning January 2027. TechCrunch reported this week that the registration portal itself opened on Tuesday, November 10, and that large frontier developers like OpenAI and Anthropic now have until January 1, 2027 to register. I couldn’t find a state page naming that exact portal-opening date, so that specific detail traces to TechCrunch’s reporting, not a government document I read directly.

What isn’t in dispute, because it’s on the state’s own site, is who runs this. DIGIT, the Office of Digital Innovation, Governance, Integrity and Trust, sits inside DFS and oversees the RAISE Act. Per dfs.ny.gov, large frontier developers operating in New York must register by January 1, 2027, and all frontier developers, large or not, have to report critical safety incidents and publish model transparency reports once the law takes effect. The public can file suspected safety-incident reports with DIGIT too, starting the same day.

What “large” means, and what it costs to get it wrong

TechCrunch’s reporting puts the large-developer threshold at more than $500 million in annual revenue, a bar that OpenAI and Anthropic clear easily. Above that line, a company has to publish its safety framework on its own website, report critical safety incidents to DIGIT within 72 hours, and file quarterly assessments of catastrophic risk. TechCrunch also reported the penalty structure: up to $1 million for a first violation, $3 million for repeat ones. I read DFS’s own page and couldn’t confirm either the $500 million figure or the specific dollar penalties there, so those two numbers are TechCrunch’s reporting, not something I verified on a .gov page myself.

New York isn’t alone, and it isn’t even first with a bill, though it may be first with real enforcement. California’s SB 53 already requires frontier labs to publish safety frameworks and report incidents, and TechCrunch’s report says Illinois has a similar bill, SB 315, moving as well. Here’s how the three compare on the one thing a company actually has to act on fast, the incident-reporting clock:

State	Law	Incident report window	Status as of Nov 10
New York	RAISE Act	72 hours	Registration open, enforcement from Jan 1, 2027
California	SB 53	15 days	In effect since January 2026
Illinois	SB 315	Not yet finalized, per TechCrunch	Moving through the legislature

New York’s window is the tightest of the three. If the RAISE Act’s structure spreads to a handful of states, companies serving the whole country functionally comply nationwide, which is close to what Hochul said she wants.

My read: the deadline date is doing less work than the registration date

The headline framing, that OpenAI and Anthropic “face new oversight,” slightly overstates what changed this week. The oversight itself, the 72-hour reporting and the published frameworks, doesn’t start until January. What changed is paperwork: companies now have a place to register and a countdown to do it. That’s not nothing. A state agency having a named deputy director, an open portal and a public complaint channel before the compliance date is a sign New York actually built the machine instead of just passing a bill and walking away, which is the fate of a lot of state AI legislation.

What I’m less sure about is enforcement bite. A spokesperson told TechCrunch the state can bring enforcement actions against companies that don’t register, but declined to say what that looks like in practice. Fines on paper mean little until DIGIT actually issues one, and that test doesn’t come until well into 2027 at the earliest, once incidents start getting reported and someone decides whether a report was late or a framework was thin.

Hochul also tied this to a new research arm: the New York Center for AI Responsibility and Research at Binghamton University, funded with $55 million she first announced in January, which is supposed to independently evaluate frontier model safety and feed findings back to DIGIT. Whether that center has the staff and access to actually audit a frontier model, rather than just read the safety reports companies hand over, is the kind of thing that only shows up once it tries.

Related: see how eight states are already suing OpenAI over kids and ChatGPT testing and how the White House’s AI accord leans toward self-regulation instead of state-style mandates like this one.

Note on sources: I read the DFS press release and the DIGIT program page in full, plus the governor's rush transcript. I read TechCrunch's report in full for the portal-opening date, the $500 million threshold and the penalty figures, none of which I could confirm on a state government page myself. I did not read the full RAISE Act statute text or Illinois SB 315's bill text.

New York opened an AI registration portal. OpenAI and Anthropic have until January to use it.

What “large” means, and what it costs to get it wrong

My read: the deadline date is doing less work than the registration date

Sources