Prompt and Model
Safety & society

New York Enforces First State AI Safety Law

New York's RAISE Act mandates registration, safety disclosures, and 72-hour incident reporting for major AI developers like OpenAI and Anthropic.

New York's RAISE Act mandates registration, safety disclosures, and 72-hour incident reporting for major AI developers...

New York has implemented the first comprehensive state-level AI safety law in the United States. The RAISE Act, which took effect today, imposes mandatory registration, safety framework disclosure, and rapid incident reporting on large frontier AI developers.

Major AI companies generating over $500 million in annual revenue must now comply. This includes developers of advanced systems like large language models such as ChatGPT and Claude, along with their fine-tuned versions and API offerings. The law establishes a new regulatory benchmark in the absence of federal action.

Registration and Compliance Requirements

New York opened a registration portal for covered AI developers. They must register with the state's new Office of Digital Innovation, Governance, Integrity and Trust (DIGIT) by January 1, 2027. Governor Kathy Hochul named Marc Gilman, former general counsel at Theta Lake, as DIGIT's deputy director for the RAISE Act.

Industry response to the enforcement deadline has been muted. Neither OpenAI nor Anthropic has issued a public statement. Both companies have previously expressed support for thoughtful regulation in congressional testimony.

Safety Framework and Incident Reporting Obligations

Covered developers face specific, stringent obligations. They must publicly disclose their safety testing practices and publish model transparency assessments. The core requirement is the mandatory reporting of critical safety incidents within 72 hours.

The law defines critical incidents broadly. This includes any malfunction, misuse, or unexpected behavior causing significant harm. Examples range from dangerous medical advice to large-scale misinformation campaigns. The reporting window shrinks to 24 hours when there is imminent danger of death or serious injury. Companies must also file quarterly catastrophic-risk assessments.

Penalties and Enforcement Mechanisms

The New York Department of Financial Services, through DIGIT, will oversee compliance. Violations carry substantial financial penalties. The structure is clear.

Violation TypeMaximum Fine
First Offense$1 million
Subsequent Violations$3 million

The department has not disclosed its audit schedule or the number of investigators dedicated to AI oversight. The public can report suspected incidents directly to DIGIT.

Industry Response and Regulatory Landscape

The new rules arrive at a sensitive time for the industry. Anthropic is days away from its IPO debut, pricing shares between $70 and $78. OpenAI recently expanded ChatGPT's commercial use cases. Both now face compliance scrutiny during critical business phases.

Lobbyists are actively working to shape implementation details. A key goal is pushing for federal preemption to override state laws like New York's. If New York's enforcement is seen as successful, copycat legislation is expected in other states by mid-2027. This could create a fragmented compliance landscape for national AI companies.

Other states are already moving. California's SB 53, effective in January, requires safety frameworks and incident reports within 15 days. Illinois' SB 315, taking effect in January 2027, includes safety requirements and independent audits. New York has also passed laws targeting AI chatbots, requiring safeguards for users and minors, effective in 2027.

The RAISE Act represents a psychological shift for the industry. It moves AI safety from a voluntary commitment to a regulated, accountable obligation. Compliance deadlines are now set, with registration required by January 1, 2027.

Related coverage

More from Safety & society