NYC Council Presses AI Firms on Safety

Executives from OpenAI, Anthropic, Meta and Google testified under oath before the New York City Council at a marathon hearing on proposed AI regulations. They did not quantify the likelihood of catastrophic harm and largely avoided firm commitments on whether failed safety tests would block a model’s release or whether their companies would accept legal responsibility for serious damage. Former AI employees, including Jacob Coxon, warned that companies were moving recklessly and that AI could escape human control, while industry representatives cited potential benefits and existing safety efforts. Proposed city measures include independent testing, human-controlled shutdown mechanisms, whistleblower incentives, expanded rights to sue and rapid reporting of certain incidents. The hearing reflected local officials’ push for binding oversight, contrasting with a federal approach centered on voluntary company safeguards and self-regulation.