The New York City Council held a landmark AI hearing on October 5, 2026, where former Anthropic researcher and whistleblower Jacob Coxon testified in support of bills that would mandate "kill switches" on AI systems and fine non-compliant companies $25,000 per offense.
Jacob Coxon: the whistleblower
Coxon, 27, quit Anthropic in September 2026 after three years of AI pretraining research at OpenAI and Anthropic. On resigning, he warned that "the people building AI earnestly believe that it could kill us all by the end of the decade" and accused the leading labs of "gambling with our lives."
The mathematician — a UK International Mathematical Olympiad silver and bronze medalist and Cambridge graduate — has become the face of a growing safety revolt. Council Speaker Julie Menin convened Monday's hearing of the Committee of the Whole, all 51 council members, at Coxon's request, and he personally confirmed he would testify.
Fellow whistleblowers were expected alongside him: former Google DeepMind researcher Alex Turner and Daniel Kokotajlo, an ex-OpenAI researcher who now runs the AI Futures Project.
The bills: kill switches, $25,000 fines, cash rewards
The council is debating a package of AI regulation bills that would go further than any city in the country:
- Kill-switch mandate: any AI tool used in the city must include a "human override that can shut down the system."
- Third-party validation: companies could not "market, offer for sale, or deploy" AI systems in the five boroughs without independent experts vetting their internal safety claims.
- $25,000 fines for each offense.
- Whistleblower cash rewards: a first-in-the-nation approach paying informants from the fines collected from bad actors they help catch.
- Private right of action letting people harmed by an AI system sue the developer.
Tech giants send stand-ins, not CEOs
Menin invited the CEOs — Dario Amodei (Anthropic), Sam Altman (OpenAI), Sundar Pichai (Google), Elon Musk (SpaceX), and Mark Zuckerberg (Meta) — and only got commitments after threatening subpoenas. In the end, none of the CEOs showed up: Anthropic sent frontier red-team head Logan Graham, OpenAI sent Morgan Dwyer, Meta sent Shane Cahill, and Google sent Alice Friend. SpaceX's AI team had already been served with a subpoena around September 30 after non-compliance.
New York state moves in parallel
The push goes beyond the city. Governor Kathy Hochul announced in September that starting in November, AI developers with annual revenue over $500 million must register with the new DIGIT office, publish safety protocols, and report critical safety incidents within 72 hours — with enforcement beginning in January. She declined to rule out state-level kill switches.
Coxon's resignation had already "skyrocketed" concerns, and industry leaders — including Amodei, who published an essay urging rivals to jointly slow model development over recursive self-improvement risks — have unusually begun calling for stricter rules themselves.
Watch: "LIVE: Former Anthropic researcher testifies on AI at New York City Council" (Associated Press live coverage of the hearing).
Reel music: Kevin MacLeod - Electrodoodle (incompetech.com), licensed under CC BY 3.0
Sources: NY Post, NY Post (bill package), Bloomberg via Moneyweb, LiveMint