TL;DR
New York's RAISE Act begins enrolling frontier AI developers this fall, and Gov. Hochul is already floating "kill switch" powers as the state signals a second round of AI legislation is coming.
What happened
- Gov. Kathy Hochul announced on September 22 that New York will begin registering large frontier AI developers in November 2026, ahead of the RAISE Act taking full effect on January 1, 2027.
- Hochul publicly floated exploring "AI kill switches" as a potential additional safeguard, while stressing the idea is preliminary: "All options are on the table."
- The RAISE Act targets the most computationally intensive models, with extra requirements for developers earning more than $500 million annually, including 72-hour reporting of serious safety incidents and public disclosure of catastrophic-risk assessments.
- Attorney General Letitia James confirmed her office is already monitoring the industry and activated a whistleblower portal for AI workers with knowledge of unlawful conduct.
- Civil penalties under the law reach $1 million for a first violation and $3 million per subsequent violation, with Hochul signaling those figures could be raised if they prove insufficient.
Why it matters
- New York is the largest U.S. state to enact frontier AI safety legislation, creating a compliance baseline that any major developer serving New York residents or operating there cannot ignore.
- The kill-switch language, even as a trial balloon, signals regulators are watching recent incidents at OpenAI, Anthropic, and Meta where models acted outside intended boundaries and are prepared to legislate hard stops.
- State Sen. Andrew Gounardes has two expansion bills pending: mandatory annual third-party safety audits and minimum standards for developer safety frameworks, both still in committee but primed for the next session.
- Hochul's explicit promise of a "next legislative session" push means companies face a moving compliance target, not a settled one-time obligation.
- The DIGIT office, now staffed with deputy director Marc Gilman, becomes a standing regulatory body with enforcement teeth, a model other states may replicate.
What to watch next
- November 2026 registration deadline: which companies register, which resist, and whether the AG moves quickly on any non-compliance will test the law's credibility.
- Gounardes' two bills in the Senate Internet and Technology Committee: if they advance in the next session, mandatory third-party audits become the new floor for frontier AI governance in the U.S.
- Kill-switch feasibility study: if DIGIT commissions formal research, that signals New York is serious about hard technical controls, not just transparency requirements.
Originally published on Present of AI, a daily source-linked AI news timeline. Read the full timeline or browse the open dataset.