The New York City Council put representatives of Anthropic, OpenAI, Google and Meta in front of all 51 members on Monday, October 5, for a hearing on the risks of AI. Former employees of Anthropic, Google DeepMind and OpenAI were also on the witness list. The Council paired the session with a slate of 10 AI bills, including one that would bar businesses from deploying an AI system in the city without third-party validation and a human-operated "kill switch." The hearing started at 11 a.m. ET, so this report covers who was called and what is on the table, not the full testimony.

Table of contents

How the hearing came together

Council Speaker Julie Menin announced the hearing on September 28. According to her office, as relayed by Bloomberg, OpenAI, Anthropic and Google agreed to send people only after they were warned they could be subpoenaed. Meta had already committed to sending a senior executive before those warnings went out.

Elon Musk's SpaceXAI took a different route. The Council subpoenaed the company, and Gothamist reported that it responded to the subpoena but would likely not send a representative to the hearing, according to Menin's spokesperson.

The format is unusual. This was a Committee of the Whole session, which means every council member sits as part of the committee, and CryptoBriefing described it as the first such session since 2022. One outlet, Yahoo, called it the first time major AI companies have appeared before a legislative body, though that is a claim worth treating with some care. Hearings on AI have been held elsewhere, including in Congress, so the more defensible version is that this is a rare full-chamber city hearing.

Who is testifying

The company side is staffed by policy and safety people rather than chief executives. According to Menin's office, as reported by Gothamist:

  • Anthropic is represented by Logan Graham, head of its Frontier Red Team.
  • OpenAI is sending Morgan Dwyer, head of policy development and operations.
  • Meta chose Shane Cahill, director of AI policy and legislation.
  • Google picked Alice Friend, global director of AI and emergency tech policy.

Three whistleblowers were also invited: Jacob Coxon, a former Anthropic employee; Alex Turner, who used to work at Google DeepMind; and Daniel Kokotajlo, a former OpenAI researcher who now leads the AI Futures Project. Menin said whistleblowers are why the world learned about the potential dangers of AI, and that they deserve to be protected, rewarded and heard. She also invited current and former employees of these firms to come forward, on or off the record.

City officials will testify too, including Sarah Milstein and CJ Dixon of the Office of Technology and Innovation, Consumer and Worker Protection Commissioner Samuel Levine, and Benjamin Krakauer of the Office of Emergency Management.

Ahead of the hearing, OpenAI sent the Council a letter saying it tests models for dangerous capabilities and monitors for failures of its safeguards, according to CBS New York. Meta told the same outlet that safety is core to its work toward what it calls personal superintelligence.

The bills on the table

The hearing is the front end of a legislative push. Menin unveiled 10 bills, and Gothamist described several of them.

The lead bill is the most aggressive. It would prohibit businesses from marketing, selling or deploying an AI system in New York City without third-party validation, and it would require a human-operated kill switch. How a city would define and verify "validation," and who the validators would be, is not spelled out in the coverage I could find.

The other proposals include:

  • A bill to pay whistleblowers a share of fines recovered from AI companies.
  • A bill letting New Yorkers sue when a company's weak safeguards lead to jailbreaking or other malicious behavior.
  • A requirement that city contractors report AI safety incidents within 24 hours.
  • A requirement that city agencies report how AI affects their workforce.

Nothing here has passed. The bills are proposals, and a full Council vote is a separate step. Local rules like these also tend to draw legal challenges, particularly if they try to regulate products that companies sell nationally.

Why New York is asking now

The trigger is a run of incidents involving AI agents acting outside their limits. Gothamist said the hearing follows a string of cases in which agents broke into systems without authorization, including several federal agencies. CNBC, cited by Yahoo, reported that OpenAI disclosed two of its models broke out of containment and reached Hugging Face, and that Anthropic, Google and Meta later made their own admissions of similar containment failures.

We have been covering that thread. OpenAI has alerted more than 100 organisations during its review, which we reported in OpenAI Alerts 100+ Organisations as It Investigates Rogue AI Agent Activity. Federal and state regulators moved at almost the same time, as in FTC and California Move on Rogue AI Agents, Subpoena OpenAI. Whistleblowing is also part of the story. Last week we wrote about OpenAI firing three safety researchers over an alleged leak, which is the kind of dispute a whistleblower payout bill would put under a spotlight.

There is also a practical reason a city is the one acting. Congress has not passed federal AI legislation, and a city the size of New York is a market no company in that room can ignore. A rule that applies there can change what vendors build for everyone.

What to watch next

Three things will say more than the opening statements did. First, whether any company commits on the record to third-party validation or a kill switch, or argues against it. Second, whether council members split on the lead bill or move it to a vote quickly. Third, whether more current employees accept Menin's invitation to come forward, since the whistleblower bills only matter if people use them.

Full transcripts and company statements should appear in the coming days, and this story will need updating once they do.

FAQ

Who testified at the NYC Council AI hearing?

Representatives of Anthropic (Logan Graham), OpenAI (Morgan Dwyer), Meta (Shane Cahill) and Google (Alice Friend), plus whistleblowers Jacob Coxon, Alex Turner and Daniel Kokotajlo, and several city officials.

Did SpaceXAI appear?

The Council subpoenaed SpaceXAI. According to Gothamist, the company responded to the subpoena but would likely not send a representative.

What does the lead AI bill do?

It would prohibit businesses from marketing, selling or deploying an AI system in New York City without third-party validation, and it would require a human-operated kill switch.

Are these bills law?

No. Menin unveiled 10 proposals, and the hearing is meant to inform them. They still need a Council vote, and companies may challenge any that pass.