New York City Council introduces bill requiring AI models to undergo third-party verification and have shutdown capability

On October 8, the New York City Council introduced Int. No. 1100, which would prohibit the marketing, sale, or deployment in New York City of AI models that have not passed third-party verification or do not have human shutdown capability. The verification is proposed to cover task performance, data sources, bias, data handling, and safety, among other items; the verifier must also certify to the developer and the city's Cyber Command whether the model is suitable for deployment, and disclose its own conflicts of interest. The text provides for a $25,000 civil penalty per occurrence for marketing, selling, or deploying without verification or shutdown capability, as well as for falsifying verification; other violations carry a maximum penalty of $25,000. The bill is still at the committee stage, and the text provides that if enacted into law it would take effect 180 days later.