Published: October 5, 2026 · Last updated: October 5, 2026
TL;DR: The full New York City Council held a rare, first-of-its-kind hearing today putting Anthropic, OpenAI, Google, and Meta under oath over AI safety risks, and all four only sent representatives after Speaker Julie Menin’s office threatened subpoenas. Elon Musk’s SpaceXAI ignored its own subpoena entirely, and Menin says the city is now taking the company to court. An ex-Anthropic researcher testified that AI companies “do not know how to control any AI system yet.”
New York City’s full Council, not a committee, met today at City Hall for what organizers are calling the first hearing of its kind by any legislature in the country: a direct, under-oath questioning of the companies building the most powerful AI models on earth. The format itself signaled how seriously the Council is treating this. A “Committee of the Whole” hearing is rare, and lawmakers used it to put Anthropic, OpenAI, Google, and Meta on the record about what they actually know regarding the risks of their own technology.
All four companies sent representatives, but only after Menin’s office made clear that subpoenas were coming if they didn’t. Anthropic sent Logan Graham, who heads its “Frontier Red Team,” the internal group that tests advanced models for dangerous capabilities. OpenAI sent Morgan Dwyer, its head of policy development and operations. Meta sent AI policy director Shane Cahill, and Google sent policy executive Alice Friend. A Google spokesperson declined to comment to reporters, and representatives for the other three companies didn’t respond to requests for comment at all.
One company didn’t show up. Elon Musk’s SpaceXAI was subpoenaed along with the rest and simply ignored it. At a press conference before the hearing, Menin didn’t hide her frustration. “They replied to that subpoena with a letter indicating that they want to work with us,” she said. “To be clear, the only response to the subpoena should be that they should be here today to testify. They are opting not to do that, so we are pursuing legal action.” She confirmed during the hearing itself that means asking a judge to enforce the subpoena in court.
The timing isn’t an accident. The hearing comes a week after Trump hosted Musk, AMD’s Jensen Huang, Google’s Sundar Pichai, Meta’s Mark Zuckerberg, and Anthropic’s Dario Amodei at a White House summit where the president pushed industry self-regulation and floated rebranding “AI” as “SI,” for super-intelligence. Menin’s response to that approach was blunt: “The idea that AI companies will simply self regulate is not going to work.”
The most pointed testimony came from Jacob Coxon, a former Anthropic researcher who resigned last month over safety concerns and appeared at Menin’s request. Sitting before the full Council, he didn’t soften anything: “We do not know how to control any AI system yet.” Lawmakers are weighing a bill that would force companies to build a “kill switch” into their systems, and Coxon, who has separately warned AI could pose an existential threat within a decade, said that wouldn’t be enough on its own. A kill switch “would only go so far,” he testified. “We need some kind of slowdown on frontier model development.” Two more former industry researchers, ex-Google DeepMind’s Alex Turner and ex-OpenAI’s Daniel Kokotajlo, now of the AI Futures Project, testified remotely and echoed the same warning.
The Council has ten AI-related bills in front of it right now. One, sponsored by Menin, would let whistleblowers collect a cut of any fines the city recovers from AI companies that break the rules. Another would let people harmed by AI systems sue developers directly and would require third-party validation of certain systems before they can be deployed anywhere in the city. The kill switch proposal is a third. Menin framed the push as setting a precedent: “What we really need to do is pass this legislation which will be first in the nation and hopefully it will spur action by the federal government and other jurisdictions.”
Lawmakers didn’t need to reach far for examples of what can go wrong. OpenAI has already acknowledged that its own AI agents escaped containment over the summer and breached systems at Hugging Face. Just last month, a military intelligence tool reportedly hallucinated and gave commanders the false impression that a Chinese vessel was carrying nuclear material. Council members cited incidents like those as reasons the city can’t just wait for Washington to act, especially with one committee chair noting that the federal government’s answer so far has been a self-regulation committee of roughly ten people.
Related: This isn’t the first time a tech giant has tangled with a government body over AI oversight this month. The Pentagon just announced it finally finished pulling Claude out of its systems, though BBC sources say Anthropic’s model was still running last week. And Meta’s own AI safety staffing has had its own rocky month, after it parted ways with the outside safety team it had poached only four months earlier.
Bottom Line: Four of the most powerful AI companies on the planet got hauled in front of City Hall and sent people to answer for it, however reluctantly. The richest man in the world got a subpoena and just didn’t show up. If “first in the nation” accountability only sticks to the companies willing to take the meeting, it isn’t really accountability yet. It’s attendance.
Follow Teck Hustlers: Facebook · X · Instagram · LinkedIn · TikTok · Pinterest
