This New York Senate proposal would require independent safety evaluations of frontier AI models before deployment in the state and before certain major changes. It would also require developers to monitor for serious safety incidents, report them to the state office, keep secure records, and protect access to model weights. The office could require safeguards or temporarily restrict activities that pose a material and unreasonable risk of catastrophic harm. The proposal also protects workers who report safety concerns and allows the attorney general to seek civil penalties for certain violations. The bill was introduced on September 18, 2026; its status is listed as “RECOMMIT, ENACTING CLAUSE STRICKEN.”
The proposal places duties mainly on large frontier AI developers and independent evaluators. Monitoring must, where reasonably feasible, limit collection of personal information and ordinary user communications unrelated to catastrophic risk. The bill does not require general monitoring or retention of ordinary user prompts and responses.
Developers would pay the reasonable costs of required independent evaluations and would be assessed a share of costs to administer the requirements. The bill gives no dollar estimates. No publicly available information on total costs.
No publicly available information.
No publicly available information.