NYC Council Puts Labs Under Oath — No Catastrophic-Risk Number, No Release-Gate Promise; xAI Skips Subpoena
On Oct. 5, OpenAI, Anthropic, Google and Meta testified under oath to New York’s City Council on a package of proposed AI bills. None put a number on catastrophic risk or promised that failing an internal or third-party safety test would block a release. xAI did not appear despite a Council subpoena; Speaker Julie Menin said the Council is pursuing it in court. The bills remain proposals — nothing has passed.
New York’s City Council put the frontier labs on the record — and got process answers, not the commitments its proposed bills would make mandatory.
On October 5, 2026, the Council met as a Committee of the Whole (all 51 members; about 40 took part, per amNewYork) to hear a package of AI bills unveiled Sept. 25. Witnesses from OpenAI (Morgan Dwyer), Anthropic (Logan Graham), Meta (Shane Cahill) and Google (Alice Friend) testified under oath. None put a percentage on catastrophic risk. None gave a blanket commitment that failing an internal safety test or an independent third-party validation would automatically block a release. xAI — named SpaceXAI in Council materials and local coverage — did not appear despite a subpoena; Speaker Julie Menin said the Council is pursuing that subpoena in court.
The hearing landed days after the voluntary White House frontier responsibilities accord. On the record, lab answers stopped short of the kill-switch and third-party validation rules the city is weighing.
What the proposals say
These are legislative proposals, not enacted law. The Council’s Sept. 25 release and NY1’s hearing coverage describe a package that includes (among others):
- Intro 2602 — third-party validation plus a human kill switch for AI systems marketed, offered or deployed in the city; $25,000 per violation (business and validator)
- Intro 2605 — whistleblowers receive a portion of recovered fines (Council wording); NY1 reports that share as 25%
- Intro 2600 — private right of action for foreseeable harms from misuse or jailbreaking
- Intro 2601 — 24-hour reporting of AI safety incidents by city contractors to Cyber Command
- An AI emergency-response plan and related measures
Mayor Zohran Mamdani, at a separate press conference, said AI needs regulation at every level but the most urgent response must come from Washington, and did not endorse specific bills pending review.
What the labs would and would not say
Menin pressed for a catastrophic-risk probability. Dwyer declined to put a number on it and said no level is acceptable; Menin called that answer “flippant at best.” Graham discussed Anthropic’s risk work without a percentage. Cahill said he did not want to be imprecise. Friend said there is not yet a rigorous scientific method for assigning such a probability.
On release gates, Dwyer said OpenAI would not release models it did not believe were safe and has delayed releases before — but when asked whether failing an internal or third-party test would stop a release, she pointed back to process rather than a yes/no. Anthropic, Meta and Google likewise described review procedures without the blanket commitment Menin sought.
On liability for a rogue system causing serious harm, Friend gave the clearest line: existing legal frameworks apply, and “if it’s illegal without AI, it’s still illegal with AI.” Others largely declined to speak for legal departments.
Former OpenAI/Anthropic researcher Jacob Coxon testified it is “more likely than not” that humanity loses control of AI on the current path — his own assessment. Former researchers Daniel Kokotajlo and Alex Turner also testified; Turner put his personal estimate of an eventual AI takeover at roughly one in three.
Dwyer said OpenAI is reviewing possible agent misalignment incidents back to November 2025 and was not aware of any affecting city systems so far — consistent with the desk’s earlier 100+ organization notification story, not a new disclosure.
The RAISE Act fight on the side
After the hearing, Assemblymember Alex Bores posted on X that he believed OpenAI had “committed perjury” over its RAISE Act stance — an accusation reported by amNewYork; the X post itself was not inspected here. When Council Member Christopher Marte raised the dispute, Dwyer said OpenAI supports the New York RAISE Act as passed, without saying the company had backed the original, stronger draft that sponsors say lobbyists fought.
Limits
- Hearing facts and quotes are from amNewYork and NY1 secondary coverage plus the Council’s Sept. 25 proposals release — not a full hearing transcript.
- Bills are proposals. Nothing has passed; the mayor has not endorsed specific measures.
- Whistleblower share: Council says “a portion”; 25% is NY1’s figure.
- Bores’ “perjury” line is an X accusation via amNY — attributed, with OpenAI’s on-record answer included.
- Council and local outlets name the absent company SpaceXAI; house style is xAI.
- WSJ and Bloomberg Law also covered the hearing (paywalled; not inspected).
Sources
- amNewYork: AI giants give few clear answers to key safety questions at NYC Council hearing amid whistleblower warnings (October 5, 2026)
- NY1: City Council takes on debate over catastrophic risks of AI (October 5, 2026)
- New York City Council: Legislative Proposals to Safeguard New Yorkers from Potential Risks of Artificial Intelligence (September 25, 2026)