OpenAI, Anthropic face new oversight as New York’s AI safety law takes effect

New York’s RAISE Act, which takes effect on January 1, 2027, will make frontier AI labs—those earning more than $500 million a year—mandatory to register through a state portal, publish safety frameworks on their websites, and report critical incidents within 72 hours. The law’s mechanism is a three‑step compliance chain: registration, transparency, and rapid incident reporting.

Compared to California’s SB 53, which requires reporting within 15 days, New York’s 72‑hour window tightens the safety net by a factor of 4.5 days. DIGIT, the Office of Digital Innovation, Governance, Integrity, and Trust, will enforce the Act and can levy fines up to $1 million for a first violation and $3 million for subsequent ones.

The consequence is a financial risk that could eclipse the revenue of smaller frontier labs, nudging them toward earlier compliance or exit. The limit is that the law does not yet specify how “critical safety incident” is defined, leaving room for interpretation and potential legal challenges. Governor Kathy Hochul announced the law alongside a $55 million AI Responsibility Center at Binghamton University.

The center will independently evaluate safety frameworks and collaborate with DIGIT.

This partnership signals a shift from reactive to proactive oversight, but the source does not detail the center’s audit frequency or scope, a gap that could affect the law’s effectiveness.

The RAISE Act is part of a broader state push: a companion law now requires AI companion operators to disclose bot status and screen for suicidal ideation, while a separate minor‑chatbot law will ban deceptive human‑like interactions by June 2027.

Together, these statutes create a layered safety architecture that could set a national benchmark.

Our read: New York’s approach, by tightening reporting timelines and coupling them with a dedicated research center, may force other states to adopt similar frameworks.

Watch how the first registration cycle unfolds—any delays could reveal cracks in the Act’s enforcement machinery.