New York City's council wants AI models to undergo independent validation and include a human shutdown capability. Council Speaker Julie Menin announced the proposal on September 25 as part of a legislative package.
The validation proposal also requires proof that a human operator can stop the model. Validators would disclose conflicts of interest, while the city's Cyber Command would define validation requirements.
The official description sets out $25,000 penalties per instance for certain violations. Other proposals address incident reporting and protections for whistleblowers.
The council announced a hearing for October 5. The legislative record lists the shutdown measure as a proposal with no law number or enactment date. What participants say at the hearing must be checked against their own original sources.