New York Enacts AI Safety Rules Requiring 72-Hour Incident Reporting

Original title: 纽约推出 AI 监管新规,OpenAI、Anthropic 需 72 小时内上报安全事故
78AI Score

New York has opened the registration portal for its RAISE Act, requiring frontier AI developers with over $500 million in annual revenue to register by January 1, 2027, publish safety frameworks and model transparency reports, and report major safety incidents within 72 hours. Violations carry fines of up to $1 million per incident, rising to $3 million for repeat offenses, with the new DIGIT office overseeing enforcement. It marks one of the most binding state-level AI safety regimes in the US to date.

Full Report

IT之家 November 11 news — New York State is preparing to implement one of the most significant AI safety regulations in the United States. On Tuesday, the state officially opened a regulatory registration platform, and major frontier AI developers such as Anthropic and OpenAI must complete registration by January 1, 2027.

New York introduces new AI regulations: OpenAI and Anthropic must report safety incidents within 72 hours

The regulation, known as the Responsible AI Safety and Education Act (RAISE Act), is an AI safety and transparency bill introduced by New York State and will officially take effect in early 2027. Under the rules, major frontier AI developers with annual revenue exceeding $500 million (IT之家 note: approximately RMB 3.344 billion at current exchange rates) must publish their safety frameworks on their websites, report major safety incidents to regulators within 72 hours, and publicly release model transparency reports.

The newly established Office of Digital Innovation, Governance, Integrity, and Trust (DIGIT) will be responsible for implementing the RAISE Act and overseeing the operations of major frontier AI laboratories. Companies that violate the regulations may face fines of up to $1 million (approximately RMB 6.688 million at current exchange rates) per violation, with subsequent violations subject to fines of up to $3 million (approximately RMB 20.063 million at current exchange rates).

In addition, New York State also has the authority to take enforcement action against companies that fail to complete registration through the new registration platform. However, a state government spokesperson declined to disclose the specific penalties.

New York Governor Kathy Hochul said at a press conference on Tuesday local time: "Tech giants tell us that the AI development path they envision is inevitable. They tell us to either keep up or be left behind... They also believe that as long as we win this race toward the AI cliff, any collateral damage along the way is worth it."

On Tuesday, Hochul also announced the establishment of the New York State AI Responsibility and Research Center. Located at Binghamton University, part of the State University of New York, the center will independently evaluate the safety, reliability, and operational status of advanced AI systems, and work with DIGIT to review the safety frameworks submitted by companies. Hochul first announced plans for the center in January this year, when it received $55 million (approximately RMB 368 million at current exchange rates) in donations and construction funding support.

Hochul said: "We are building a national model for AI regulation that is both scientifically sound and practically feasible. This model will put community interests above corporate interests, and human interests above powerful technology."

The RAISE Act is not the only AI safety regulation passed by New York State. Last year, the state began implementing a law targeting AI companion services, requiring relevant operators to periodically remind users that they are communicating with chatbots, while also taking reasonable measures to identify users' potential suicidal tendencies or self-harm behavior.

In October this year, Hochul signed another chatbot safety bill aimed at minors. The bill restricts AI chatbots from engaging in deceptive anthropomorphic interactions, providing emotionally manipulative or purely sycophantic responses, and prohibits encouraging suicide, eating disorders, and other harmful behaviors. The law will take effect on June 1, 2027.

Other U.S. states have also successively passed AI safety and transparency regulations similar to the RAISE Act.

California's SB 53 took effect in January this year, requiring frontier AI laboratories to establish and disclose safety frameworks, provide protections for internal whistleblowers, and report major safety incidents to regulators within 15 days.

Illinois' SB 315 is scheduled to take effect in early 2027. Its requirements are similar to the above regulations, while also adding an independent safety compliance audit system.