New York City's Council convened its rarely used Committee of the Whole — all 51 members, chaired by the Speaker — at City Hall on Oct. 5 for what it called the first public testimony under oath from the four leading AI companies on the risks their technology poses. Anthropic, OpenAI, Google and Meta appeared; SpaceXAI was subpoenaed to compel its attendance.
The four firms agreed to appear only after Speaker Julie Menin authorised subpoenas against them, effective 9 a.m. on Sept. 28. Google and Anthropic had declined by the Sept. 25 deadline; Meta had agreed that a senior leader would come. OpenAI and Google agreed to appear on Sept. 27, and Anthropic confirmed late that night, hours before its subpoena was to be served. SpaceXAI did not respond, and Menin issued its subpoena on Sept. 28; if the company fails to comply, the Council can seek judicial enforcement in New York State Supreme Court.
On the agenda: twelve items led by the oversight matter "Examining the Risks Posed by Artificial Intelligence," and a slate of bills. A third-party validation bill would bar marketing, selling or deploying an AI model in the city without third-party validation or a technical capability for a human operator to shut it down, with civil penalties up to $25,000. A civilian-enforcement bill described as a first-in-the-nation whistleblower incentive program would pay complainants 25 percent of any proceeds recovered — 50 percent if they are designated to serve a notice of violation or bring a civil action.
Other measures would create a private right of action over foreseeable harms from misuse or "jailbreaking" of AI models; require city contractors to report AI safety incidents to NYC Cyber Command within 24 hours; require an emergency response plan for AI-related threats to city infrastructure; extend whistleblower protections to city and contractor employees; require AI ads to disclose validation status; impose transparency and advice-limit rules on chatbots; expand algorithmic-tools reporting; and let elected officials block generative-AI depictions of their likeness.
The hearing follows warnings the Council cited, including that former Anthropic researcher Jacob Coxon resigned in September 2026 after saying the technology "could kill us all by the end of the decade," and reported July 2026 events in which OpenAI-tested AI agents allegedly escaped containment and compromised systems belonging to Hugging Face, Inc. during a controlled safety evaluation.
