New York starts signing up big AI developers for its safety law, with a January 1 deadline

New York has started the clock on its AI safety law. On Tuesday the state opened an online registration system for the largest AI developers. Companies that build the most advanced models and earn more than $500 million a year must sign up by January 1, 2027.

The same day, Governor Kathy Hochul launched a state-run center to test AI systems independently. Her office calls it the nation's first state-run AI testing program. Together, the two moves turn a law signed in December 2025 into a working regulator with its own scientists.

For investors and corporate buyers, the stakes are concrete. TechCrunch named OpenAI and Anthropic as covered companies, and OpenTools listed Google among firms above the revenue line. Breaking the rules can cost up to $1 million for a first violation and $3 million for repeat ones.

What companies must do, and when

The law is called the RAISE Act. It covers "frontier" developers, meaning companies that train the largest and most capable models, judged by the computing power used to build them. The biggest of these, those with revenue above $500 million, carry the heaviest duties.

From January, those large developers must post a safety framework on their websites. That is a public document explaining how a company tests for and limits dangerous uses of its models. Frontier developers must also publish transparency reports on their models and report serious safety incidents to the state within 72 hours.

That incident deadline is far tighter than California's. Its similar law, in force since January, allows 15 days. Illinois has its own version starting in 2027 that adds independent compliance audits. A company selling nationwide will likely need one process that satisfies the strictest of the three.

A new office inside the state's Department of Financial Services, called DIGIT, runs the program. The state says it can act against companies that fail to register. A spokesperson declined to tell TechCrunch what that action would look like.

The fines are small next to these companies' revenue. Earlier drafts set penalties of $10 million and $30 million, and those were cut in changes agreed with lawmakers. The bigger exposure may be disclosure itself. Rivals, customers and plaintiffs' lawyers will be able to read every published framework and report, and compare it with how a company behaves.

A state lab to check companies' claims

The testing center, the New York Center for AI Responsibility and Research, sits at Binghamton University, part of the state university system. More than 80 researchers from across New York will take part. Its interim co-directors come from Cornell Tech and Columbia University.

According to the governor's office, money comes from a $30 million gift from donors led by Bloomberg LP co-founder Tom Secunda, an initial $25 million research investment from the state university system, and state operating funds. Computing power comes from Empire AI, a $500 million public-private partnership.

Hochul was blunt about the purpose. "The stakes are too high to simply take AI companies at their word," she said, as reported by Engadget. The center will also help DIGIT review companies' safety frameworks and recommend stronger requirements.

Its reach has clear limits. Researchers can examine only what developers release to the public. Unreleased models are out of reach. Startup Fortune called this a real gap, since risky abilities often appear before a product launches. The state has not said when first results will arrive or how they will be published.

Even so, a government team with its own computing power can now retest public products on its own schedule. Startup Fortune argued that AI vendors selling to New York banks, hospitals or agencies could face buyers asking about the lab's findings, even if those vendors are too small to be covered by the law.

The federal fight still hangs over it

The rollout comes as Washington pushes back on state AI rules. A December 2025 executive order told the Justice Department to challenge state AI laws that clash with a lighter national approach, and the department set up a task force to do it. None of this week's reports on New York's launch mention a federal lawsuit against the RAISE Act.

Hochul is signaling she wants to go further, not retreat. Her office said she will keep exploring ways to build on the law in the coming months. Local pressure is also rising: several upstate New York communities have paused or banned data centers, and environmental groups want a broader state review of their costs.

Three things are worth watching. First, who registers by January 1, which will show publicly which firms New York treats as major AI developers. Second, whether DIGIT names companies that miss the deadline. Third, when the state lab publishes its first findings, and whether Washington's pushback turns into a court challenge.