The new frontier: Lessons from California's SB 53 and New York's RAISE Act

Throughout 2025, artificial intelligence governance remained a central topic of discussion among state policymakers with numerous AI-focused bills introduced across the country. These bills typically addressed one or two discrete types of AI harm rather than adopting a comprehensive approach akin to the EU AI Act. Policymakers continue to intensify their focus on how to govern foundation or frontier models, the advanced, large-scale, general-purpose AI systems such as large language models or multimodal models. California and New York are the first states to pass legislation that imposes transparency requirements and accountability on developers of these models, particularly with respect to their most catastrophic possible risks.

California moved first, adopting SB 53, the Transparency in Frontier Artificial Intelligence Act, 29 Sept. 2025. New York followed several months later with the enactment of the Responsible AI Safety and Education Act, or RAISE Act, 19 Dec. 2025, subject to an amendment negotiated between the legislature and Gov. Kathy Hochul, D-N.Y., as a condition of her approval of the bill.

Initially, the approaches to foundation models differed markedly between the two states, with disagreements over definitions, obligations and penalties. However, in a remarkably rapid example of the Sacramento Effect, the amendment to the RAISE Act was introduced in January 2026 and signed into law in March, bringing New York neatly into alignment with California.

Key differences remain, including provisions in SB 53 aimed at protecting whistleblowers who work with foundational models and a shorter incident reporting deadline in the RAISE Act, as highlighted in Hochul's original signing statement.

Scoping the final frontier