New York opens AI safety registry, giving OpenAI and Anthropic a Jan. 1 deadline
New York opened its RAISE Act registration portal, giving the largest frontier AI developers until January 1, 2027 to sign up, and launched a state center to test advanced AI systems.
New York opened its registration portal for the largest AI developers on Tuesday, putting a date on when companies such as OpenAI and Anthropic must sign up with the state. The state's new Office of Digital Innovation, Governance, Integrity, and Trust (DIGIT) launched the RAISE Portal, and large frontier developers have until January 1, 2027 to register through it.
The portal is the first concrete compliance step under the Responsible AI Safety and Education (RAISE) Act, the frontier-model law Governor Kathy Hochul signed. Starting in January, the biggest labs must publish safety frameworks, report serious incidents within 72 hours and file regular risk summaries with a state regulator. Hochul paired the launch with a new state-backed research center that will independently test advanced AI systems.
Who has to register
The law's scope is set by compute and revenue. According to DIGIT's guidance for developers, a frontier model is a foundation model trained with more than 10^26 integer or floating-point operations, counting the original training run plus later fine-tuning, reinforcement learning and other material modifications. A frontier developer is anyone who has trained, or started training, such a model. A large frontier developer is one that, together with its affiliates, had more than $500 million in annual gross revenue in the preceding calendar year.
That revenue line captures the leading labs easily. Anthropic, for example, reported $16.3 billion in first-half revenue in its IPO filing.
Registration happens through an NY.gov ID account. Once DIGIT authorizes an account, a large developer files what the law calls a disclosure statement: its principal place of business, any New York offices, beneficial ownership information and three points of contact. The statement must be renewed every two years, or sooner if model ownership changes or the reported information materially changes. Large developers also have to pay assessments to the state.
What the rules require
Every frontier developer, large or not, must report a qualifying critical safety incident to DIGIT within 72 hours of determining it happened, or of learning enough to reasonably believe it did. If the incident poses an imminent risk of death or serious physical injury, the developer must alert an appropriate authority, such as law enforcement, within 24 hours. All frontier developers must also publish a transparency report when they deploy a new or substantially modified frontier model, covering its release date, supported languages and output modalities, intended uses and restrictions.
Large developers carry more. They must publish and update at least yearly a frontier AI safety framework explaining how they assess and mitigate catastrophic risks, respond to incidents, secure unreleased model weights and govern these processes internally. Every quarter they must send DIGIT a summary of their assessment of catastrophic risk from internal use of their own models, within 30 days of the quarter's end. Their model transparency reports must also summarize catastrophic-risk assessments, their results and how far third-party evaluators were involved.
The internal-use requirement is notable as labs increasingly point their own models at their research. Anthropic's amended IPO filing, for instance, says its AI may soon sharply speed up its own research.
Penalties can reach $1 million for a first violation and $3 million for subsequent ones, TechCrunch reported. The state can also bring enforcement actions against companies that fail to register, though a spokesperson declined to give specifics. Under the law, members of the public can also report suspected critical safety incidents, and DIGIT will publish an annual report summarizing what it receives, according to a September state announcement that also named Marc Gilman as DIGIT's deputy director for the RAISE Act.
A state lab to check the labs
Alongside the portal, Hochul launched the New York Center for AI Responsibility and Research at SUNY's Binghamton University. The center will bring together more than 80 researchers from institutions across the state to test frontier AI systems, using computing power from Empire AI, the state's $500 million public-private research initiative. Greg Morrisett of Cornell Tech and Jeannette Wing of Columbia University will serve as interim co-directors.
The center was first announced in January, funded by a $30 million philanthropic gift from donors led by Bloomberg LP co-founder Tom Secunda and an initial $25 million research investment from SUNY. The governor's office said it will coordinate with DIGIT to begin reviewing developers' published safety frameworks and recommend ways to strengthen the requirements.
"The stakes are too high to simply take AI companies at their word," Hochul said in the announcement.
How New York compares
New York is not the only state writing frontier rules. California's SB 53, in effect since January, requires frontier labs to publish safety frameworks, protect whistleblowers and report critical safety incidents within 15 days, according to TechCrunch. Illinois' SB 315, which takes effect at the start of 2027, has similar requirements plus independent safety compliance audits, the outlet reported. New York's 72-hour window for incident reports is far tighter than California's 15 days.
The disclosure regime arrives as labs keep shipping more capable systems, such as OpenAI's GPT-6 Cyber, which found 81 Firefox flaws. The registration deadline is January 1, 2027, and the governor's office said Hochul will keep exploring ways to build on the RAISE Act in the coming months.
