Skip to the content
AI and Cybersecurity Learning and Practice Center Test automation Open the Learning Center

NewsPolicy

New York City Council questions OpenAI, Anthropic, Google and Meta under oath, and gets few firm answers on control

On October 5 the New York City Council held a rare Committee of the Whole hearing on the risks of AI agents, with OpenAI's Morgan Dwyer, Anthropic's Logan Graham, Meta's Shane Cahill and Google's Alice Friend testifying virtually. Asked how likely a loss of control is and whether a failed safety test would stop a release, the companies described their risk assessments and review processes but gave no percentages or blanket commitments; Meta promised written answers, and Google said no rigorous scientific method exists to put a probability on a catastrophic event. SpaceXAI, which the Council had subpoenaed, did not appear. Former researchers from OpenAI, Anthropic and Google DeepMind warned that the companies do not fully control their systems. The Council is weighing ten bills, among them third-party validation before AI systems are deployed in the city, a human kill switch, incident reporting within 24 hours and a share of fines for whistleblowers. Speaker Julie Menin said the Council will send the companies written questions.

Why it matters

A city is now writing rules for AI agents the way it would for elevators or buildings: test before use, a way to stop it, and a deadline to report when it goes wrong. Teams deploying agents in New York should expect to show all three.

Sources

Read it with the rest of the news in the Learning Center