Requires an office within the department of financial services to establish minimum standards for large frontier developers' frontier AI frameworks to prevent unreasonable levels of catastrophic risk.
S10456 would require the relevant office within the Department of Financial Services to adopt regulations by July 1, 2028 establishing minimum standards for large frontier developers’ frontier AI frameworks. The stated purpose is to prevent “unreasonable levels of catastrophic risk” from advanced AI systems. The bill does not itself set detailed technical requirements; instead, it directs the office to create those standards through rulemaking under the State Administrative Procedure Act.
In developing the regulations, the office must consider industry best practices, national and international standards, and laws and regulations from other states and jurisdictions. It must also solicit input from academia, the Empire AI research institute, the attorney general, civil society, frontier model developers, and foundation model developers. The office must review the rules annually and update them as needed based on new scientific understanding, technical advances, safety incidents, or gaps in existing regulation.
The bill would add a new section 1430 to the General Business Law and expand the regulatory authority of the designated office within DFS over frontier AI developers operating in New York. Its practical effect is to create a state-level framework for AI safety regulation focused on large frontier models, with standards to be set later by regulation rather than directly in statute. It would affect frontier AI and foundation model developers by subjecting them to minimum safety standards intended to reduce catastrophic-risk scenarios.
No committee transcript or vote record is available, so there is no documented floor or committee debate to gauge support or opposition. Based on the bill text and caption, the measure appears to be framed as a precautionary AI safety bill, with an emphasis on expert consultation and adaptive regulation. The overall tone of the proposal is preventive and technical rather than punitive.
The main potential point of contention is the scope of state regulation over frontier AI development, especially whether New York should impose standards on advanced AI systems before federal or broader international rules are established. Developers may be concerned about compliance burdens, uncertainty from evolving annual updates, and the breadth of the term “catastrophic risk.” Supporters are likely to emphasize the need for proactive safeguards, while critics may argue the bill could be overbroad, difficult to implement, or duplicative of emerging standards elsewhere.