What the New York RAISE Act Actually Requires
On Tuesday morning, a new government website went live in Albany. It looked ordinary enough. A registration form, a few pages of guidance, and a deadline.
But for the companies building the most powerful artificial intelligence on Earth, that quiet launch changed the rules. New York had opened the registration portal for its RAISE Act, and the countdown had begun.
Under the law, large frontier AI developers must register with the state by January 1, 2027. Miss the deadline, and the fines start at $1 million.
The law’s full name is the Responsible AI Safety and Education Act. According to TechCrunch, it is one of the most significant AI safety laws yet enacted in the United States.
Governor Kathy Hochul opened the portal herself at a press briefing on Tuesday, November 10. She did not soften the message.
“Tech moguls tell us that their vision for AI is inevitable,” she said. “They tell us get on board or get left behind… That any collateral damage is worth it as long as we win the race to the AI cliff.”
Her answer, in short: not in New York.
Who Has to Register Under the RAISE Act
The law does not apply to every company that uses AI. It targets a narrow group: large frontier developers. To qualify, a company must earn more than $500 million in annual revenue and develop foundation models trained with more than 10^26 operations of computing power.
That bar is high by design. Today, it captures companies at the scale of OpenAI and Anthropic — the two labs whose models power ChatGPT and Claude. Smaller startups, app developers, and businesses that simply build on top of frontier models are not covered.
But the companies that are covered face real obligations. Registration is only the first step. Each covered developer must publish a safety framework on its public website, issue transparency reports for the models it releases, and file regular assessments of catastrophic risks.
Anthropic, one of the companies in scope, is already in the public spotlight. The Claude maker is heading toward a stock market listing that could value it at around $1 trillion, and regulatory risk is now a visible line item for its investors.
The 72-Hour Rule That Will Keep AI Lawyers Awake
The toughest requirement in the law may be the shortest one. When a covered company discovers a critical safety incident, it must tell the state within 72 hours.
Three days is not much time. Inside that window, a lab must spot the problem, work out what happened, decide whether it counts as critical, and file a report with the state. That means building internal processes before anything goes wrong, not after.
For comparison, America’s securities regulator gives public companies four business days to disclose a major cybersecurity incident. California’s AI safety law, SB 53, allows 15 days. New York’s 72-hour clock is far tighter than either.
The definition of a “critical safety incident” will matter enormously in practice. The law leaves the details to the new regulator, and how it interprets that phrase will decide whether the rule changes lab behaviour or just produces paperwork.
The incident-reporting rule lands in a year when AI safety has already moved from theory to headlines. OpenAI recently built a control room that can halt a rogue model in 35 seconds, a sign that even the labs themselves now treat runaway systems as a live operational risk.
DIGIT: New York’s New AI Watchdog
Enforcing the law falls to a brand-new office: the Office of Digital Innovation, Governance, Integrity, and Trust, known as DIGIT. It sits inside New York’s Department of Financial Services, the same regulator that oversees banks and insurers in the state.
DIGIT will implement the RAISE Act and oversee the labs it covers. Marc Gilman has been appointed deputy director of the office and will lead the RAISE Act work.
The state can bring enforcement actions against companies that fail to register, and the attorney general can bring civil actions for missing reports or false statements. A first violation can draw a fine of up to $1 million. Each subsequent violation can cost up to $3 million.
For companies worth hundreds of billions, that is still small money — but the reputational cost of a public enforcement action is another matter entirely.
A New Research Center to Check the Labs’ Homework
Alongside the portal, Hochul launched the New York Center for AI Responsibility and Research, based at the State University of New York’s Binghamton University. The center will independently evaluate the safety, security, and operation of advanced AI systems.
It will also work with DIGIT to review the safety frameworks that companies publish. Hochul first announced the center in January, backed by $55 million in donations and capital.
There is one limit. According to Engadget, the center can only evaluate materials and data that AI labs make publicly accessible. Unreleased internal models stay outside its direct reach, at least for now.
“We’re building a national model for smart, sensible AI regulation,” Hochul said at the briefing, “a model that puts communities above corporations and people above powerful technology.”
New York vs California: Two Different Playbooks
New York is not the first state to write AI safety into law. California’s SB 53 took effect in January 2026, and it covers similar ground: published safety frameworks, whistleblower protections, and incident reporting.
The biggest difference is speed. California gives labs 15 days to report a critical incident. New York gives them three. Illinois is moving too: its SB 315 takes effect at the start of 2027 with similar requirements, plus independent safety compliance audits that neither New York nor California demands.
Together, the three states are building a patchwork. Each law has different definitions, different deadlines, and different enforcement. For a lab operating nationally, that means tracking three rulebooks at once — and more states are likely to follow.
Hochul signed the RAISE Act on December 19, 2025, and an amendment in March 2026 set the effective date of January 1, 2027. The November registration window is the preparatory phase before full compliance begins.
New York’s Wider AI Crackdown
The RAISE Act is the headline, but it is not New York’s only AI law. The state has been building a wider set of rules around the technology, piece by piece.
Last year, New York began enforcing a law that requires AI companion operators to tell users regularly that they are talking to a bot, and to make reasonable efforts to detect suicidal ideation and self-harm.
Then in October, Hochul signed a separate chatbot safety law aimed at minors. It restricts deceptive human-like interactions, emotionally manipulative or sycophantic responses, and encouragement of suicide or disordered eating. That law takes effect on June 1, 2027.
What Happens Next
The immediate deadline is January 1, 2027. By then, every covered developer must be registered, with its safety framework published and its reporting machinery in place.
Several questions remain open. New York has not yet said publicly which companies DIGIT will identify as covered registrants, or published detailed guidance on what counts as a critical safety incident. The portal’s opening answers the first gap. The rest will be filled by DIGIT’s guidance — and by its first enforcement actions.
Why the RAISE Act Matters Beyond New York
For years, AI regulation in the United States moved slowly at the federal level. Washington debated, drafted, and delayed. States are now filling that gap fast, and New York’s law is the most forceful step yet.
The momentum is global. Just days ago, six governments and ten AI companies signed the Paris Call for AI Safety and Security, a landmark pact on frontier AI oversight. New York’s law turns that kind of commitment into something enforceable, with a regulator, a deadline, and fines.
For ordinary users, the law’s effects will be indirect. Nobody in New York will see a different ChatGPT on January 1. But for businesses built on frontier AI, a provider’s compliance posture is now part of their operational risk — if a provider faces enforcement action, its customers feel it too.
The race to build more powerful AI is not slowing down. What changed on Tuesday is that, in New York at least, the race now has a rulebook — and a referee with the power to fine anyone who ignores it.
