Analysis
New York City’s Council convened a Committee of the Whole hearing at City Hall on October 5 to examine AI risks. CNBC reported that representatives from four major AI companies testified under oath. The official agenda included proposals covering model testing, human shutdown controls, chatbot data and responsibility for harm. The hearing did not itself put those proposals into law. [1][2]
Companies face questions at City Hall
CNBC identified the witnesses as Anthropic’s Logan Graham, OpenAI’s Morgan Dwyer, Google’s Alice Friend and Meta’s Shane Cahill. Council Speaker Julie Menin pressed the companies on who should bear legal responsibility when an AI system causes serious harm. Google’s representative argued that existing law applies to AI and supported a comprehensive federal framework. [2]
Former AI researchers also testified, warning about the pace of development and calling for stronger oversight. Those warnings are the witnesses’ assessments of future risk, rather than findings that a catastrophic outcome is inevitable. [2]
What the proposed rules would do
One measure would require an AI model marketed, sold or deployed in the city to undergo independent validation and include a way for a human operator to stop it. Testing would cover matters including performance, unequal impacts, privacy and safety. The city’s Cyber Command would write implementing rules and set qualifications for validators. [3]
A separate proposal would allow people to sue an AI provider for harm occurring in the city when a third party misuses its model. The proposed conditions include foreseeable harm and the provider’s failure to implement reasonable safeguards. That is a proposed basis for a lawsuit; it is not a finding that any company at the hearing is liable. [4]
The chatbot proposal focuses on users’ information and how systems describe themselves. It would impose privacy and security requirements, give users access to their own data, and require disclosure that they are interacting with a machine. It would also prohibit representations that a chatbot can give advice equivalent to a licensed professional’s. [5]
City purchasing is also in view
The hearing agenda includes measures on AI safety incidents involving city contracts, whistleblower protection and emergency planning. Under the incident-reporting proposal, contractors and agencies would notify Cyber Command within 24 hours of a reportable incident, and the office would publicly disclose reports within 24 hours. Other measures address AI advertising and the effects of algorithmic tools on city employees. [1]
The package illustrates how local government could influence both consumer-facing systems and the tools it buys for public services. Its eventual reach depends on the language that advances, any implementing rules and the outcome of the legislative process.
What remains uncertain
As reviewed October 6, the council’s hearing minutes were still draft and the legislative pages showed no enactment date or law number. Several proposal records display October 8 introduction dates, which were still in the future; this article treats them as preconsidered proposals on the October 5 agenda, not completed later actions. Witness details are attributed to CNBC.
Sources
- New York City Council • October 5 Committee of the Whole agenda and hearing record ↗Official source
- CNBC • October 5 reporting on AI-company and researcher testimony ↗Source
- New York City Council • T2026-2602: proposed validation and shutdown requirements ↗Official source
- New York City Council • T2026-2600: proposed civil claims for third-party AI misuse ↗Official source
- New York City Council • T2026-2599: proposed chatbot privacy, security and transparency rules ↗Official source