New York RAISE Act sets tough transparency rules for frontier AI

The RAISE Act is a clear marker that state governments are no longer content to let Washington set the pace on AI safety. By legally mandating safety frameworks, incident reporting and an oversight office for “frontier” models, New York is turning what were previously voluntary commitments by labs like OpenAI, Google and Anthropic into hard compliance obligations. In practice, that means safety teams and policy staff move from “nice to have” to core infrastructure for any company training the largest systems.([governor.ny.gov](https://www.governor.ny.gov/news/governor-hochul-signs-nation-leading-legislation-require-ai-frameworks-ai-frontier-models?utm_source=openai))

Strategically, this law tightens the regulatory noose around the highest‑end models while leaving smaller startups mostly untouched, mirroring California’s SB 53 and creating a bi‑coastal de facto standard for US AI governance. For the biggest labs, operating in New York now means building auditable safety pipelines and being prepared for legal scrutiny if a model is stolen, misused for large‑scale crime, or exhibits dangerous capabilities. That raises the fixed cost of chasing the frontier and may push more consolidation toward players that can absorb compliance overhead.

For the broader race to AGI, the message is that political legitimacy is becoming as important as benchmark scores. Labs that want to keep scaling will need robust governance stories as much as bigger clusters of GPUs. New York’s move also sets up a looming constitutional fight with Trump’s executive order attempting to pre‑empt state AI rules, which could ultimately decide how fragmented or centralized US AI regulation becomes.([techcrunch.com](https://techcrunch.com/2025/12/20/new-york-governor-kathy-hochul-signs-raise-act-to-regulate-ai-safety/?utm_source=openai))