New York enacts RAISE Act for frontier AI models

Governor Kathy Hochul signed the Responsible AI Safety and Education (RAISE) Act, making New York the second US state, after California’s Transparency in Frontier Artificial Intelligence Act, to impose statutory transparency obligations on developers of the largest AI models. Assemblymember Alex Bores, the bill’s sponsor, said it moved beyond California’s law in some respects, calling it the strongest state AI transparency law in the country.

As passed, the RAISE Act applies to “large developers” that have trained at least one “frontier model” — defined as a model trained with more than 10^26 operations and over $100 million in compute cost, or one produced by knowledge distillation costing more than $5 million — and that have spent over $100 million in aggregate training costs. Covered developers must publish a written safety and security protocol addressing risks of “critical harm,” test models for misuse potential, and report safety incidents to the state attorney general within 72 hours of learning of them — a materially tighter window than California’s 15-day requirement (24 hours for imminent risk of death or injury). The Attorney General can bring civil actions for false statements or failure to report, with penalties reported at up to $1 million for a first violation and $3 million for subsequent ones. A new oversight office within the Department of Financial Services will issue annual assessments of large frontier developers.

Hochul’s signing was accompanied by an agreement with legislators on chapter amendments to be passed in the following session: replacing the compute-and-cost threshold with one based on developer revenue exceeding $500 million annually, aligning New York’s definition with California’s, and setting the law’s effective date at 1 January 2027 — a year after California’s equivalent statute takes effect. The bill’s original provision restricting deployment of models posing “unreasonable risk” of critical harm was also flagged for removal, narrowing the law toward disclosure rather than substantive limits on deployment.

The RAISE Act was signed eight days after President Trump’s executive order directing federal agencies to challenge state AI laws seen as burdensome, setting up New York’s law as an early test case for that preemption effort.