New York’s Gounardes Proposes RAISE Act 2.0 With Lower Compute Threshold and Licensed Frontier AI Examiners

New York State Sen. Andrew Gounardes has proposed a major expansion of the state’s frontier AI safety law, according to a New York State Senate press release on October 29. The bill, S.10731, has been described as RAISE Act 2.0 and is co-sponsored by Sen. Patricia Fahy and Assemblymember Jordan Wright. Gov. Kathy Hochul signed the original RAISE Act in December, with compliance due from January 2027.

The new bill would change who is covered. Instead of applying to developers with more than $500 million in revenue, it would cover companies that spend more than $1 billion on frontier AI research and development over 36 months. It would also lower the compute threshold for covered models from 10^26 to 10^25 operations, matching the European Union’s level, and broaden the definitions of critical safety incidents and catastrophic risk.

Why it matters

The bill goes beyond disclosure. Developers would file quarterly reports on automated AI research and development and report significant increases in risk within 72 hours. From April 1, 2028, state-licensed frontier examiners would produce risk reports every 90 days. A state office within the Department of Financial Services could limit training compute, the attorney general could seek court orders restricting deployment, and penalties could reach $1 million a day, rising to $3 million a day for subsequent violations. Binding standards would apply by January 1, 2029.

Gounardes said it is time to rein in the AI arms race, and Assemblymember Alex Bores, who sponsored the original law in the Assembly, described it as a floor, not a ceiling. The advocacy groups Encode and the Secure AI Project support the bill. The move comes as frontier labs shape their own proposals, after Anthropic pitched its own frontier AI safety bill to lawmakers, and as federal pressure builds, with House Democrats asking frontier labs to report safety incidents within 48 hours and the White House mandating AI incident reporting.

What to watch

The bill has been referred to the Senate Rules Committee and is expected to be considered in the 2027 legislative session. Watch how frontier developers and industry groups respond to the licensed examiner model and the power to cap training compute, whether Hochul signals support, and whether federal preemption efforts in Washington try to override state frontier AI rules before the new standards take effect.