New York Passes America's First High-Risk AI Safety Bill 'RAISE Act'

On June 12, 2025, the RAISE Act (Responsible AI Safety and Education Act), introduced by New York State Senator Andrew Gounardes, passed the state senate — America's first high-risk AI safety legislation, mandating that cutting-edge AI models maintain safety plans preventing 100+ casualties or $1B+ property damage. Requirements: covered entities must implement safety plans; mandatory incident reporting and system transparency obligations; New York Attorney General can impose civil penalties up to $30 million for violations. "High-risk" AI definition focuses on frontier models with sufficient capability to cause large-scale harm — not all AI systems. The act shifts regulatory philosophy from "innovation-first, regulate-later" to "accountability-first" for the highest-capability systems. Comparison with other AI governance approaches: EU AI Act (risk classification with pre-licensing for high-risk systems); California SB 1047 (vetoed by Governor Newsom in 2024, similar scope); federal RAISE Act analog pending in US Congress. RAISE Act significance: (1) First state law specifically targeting "frontier AI" safety rather than specific application domains (healthcare, employment, education); (2) Establishes precedent that private AI development carries public safety obligations even before deployment; (3) Creates enforcement mechanism through civil penalties rather than criminal liability; (4) New York's enactment creates regulatory pressure on other states and federal level. Industry reaction: concerns about compliance uncertainty for rapidly evolving capabilities; questions about who constitutes a "covered developer"; the 100+ casualty threshold is seen as both setting a high bar and creating a perverse incentive to ensure model capabilities stay just below triggers.