At a New York City Council hearing on October 5, 2026, representatives of OpenAI, Anthropic, Meta and Google reportedly stopped short of the absolute assurance sought by Speaker Julie Menin: that their AI agents would always follow safeguards intended to prevent catastrophic harm. The official meeting record lists the hearing at 11 a.m. in Council Chambers at City Hall.
How the representatives reportedly answered
The reported answers differed. OpenAI representative Morgan Dwyer said technology could not be guaranteed risk-free. Anthropic representative Logan Graham described control science as difficult and unsettled, while Google representative Alice Friend said promising perfection was not possible. Meta representative Shane Cahill said he could guarantee Meta’s commitment to safe AI development and deployment. That was a commitment to the company’s safety efforts, not the absolute assurance Menin requested about how its agents would behave.
AI-safety proposals on the Council agenda
The agenda included proposals concerning third-party validation and human-operated shutdown capability for AI models, reporting certain AI-safety incidents involving city contracts, and private or civilian enforcement. The relevant items were marked as heard or laid over.
The official meeting record
The New York City Council’s meeting page lists the agenda and minutes as final and links a recording of the October 5 hearing.
