Warnings meet industry testimony
Former employees of Anthropic, OpenAI and Google DeepMind warned New York City lawmakers Monday that artificial-intelligence development is moving faster than the systems designed to control it. Their testimony came during a rare Committee of the Whole hearing that brought the full City Council together to examine safety risks and a package of proposed local laws.
Current representatives of OpenAI, Anthropic, Google and Meta also appeared. They emphasized potential benefits, described internal safety work and expressed varying degrees of openness to regulation. The hearing therefore produced two records at once: unusually direct criticism from people who had worked inside leading laboratories and public commitments from companies that could be measured against future policy.
The rules under consideration
Council proposals include third-party validation of certain AI systems and a requirement that a human operator be able to shut them down. Other measures would require disclosure of safety incidents involving city contractors, clarify whistleblower protections, regulate chatbot privacy and make some misleading safety claims in advertisements subject to penalties.
One proposal would establish a private right of action for certain harms caused by foreseeable misuse when a company failed to adopt reasonable safeguards. Another would direct city agencies to develop an emergency-response plan for AI-related threats to infrastructure, operations, public health or welfare.
These are introductions, not enacted laws. Drafts can change after testimony, fiscal review and negotiation. The Council must also decide which systems merit heightened scrutiny, what evidence an independent validator should examine and how enforcement would protect security-sensitive information and legitimate trade secrets.
Where the debate sharpened
The most contentious exchange came when Council Speaker Julie Menin asked company representatives to quantify worst-case risks. The answers underscored how difficult it is to translate broad warnings into a number that can guide legislation. A risk estimate may be uncertain, but uncertainty does not answer what controls are appropriate, who verifies them or what happens after a failure.
Industry representatives maintained that advanced systems are already helping people and that their companies devote substantial resources to safety. Former employees argued that competition creates pressure to release more capable models before outside institutions can independently assess them. Those claims now sit in a public record alongside the companies’ descriptions of their safeguards.
New York’s limited but significant role
City government cannot create a complete national framework for developing frontier models. It can, however, regulate municipal procurement, city contractors, local consumer practices and some products offered within the five boroughs. New York’s market size can also give local standards influence beyond the city, even when formal jurisdiction ends at its borders.
That influence brings a drafting challenge. Rules aimed at a handful of powerful developers could unintentionally reach small businesses that merely use outside software. Definitions must distinguish between building a model, distributing it and applying it in a local service. They must also remain useful as technical architecture and product names change.
What happens next
The Council’s official agenda still lists the minutes as draft and shows no final action on the measures. Lawmakers can revise the bills, request more information and hold additional hearings before committee votes. Any proposal that advances would then need approval by the Council and action by the mayor.
The immediate outcome is transparency rather than regulation. The hearing forced competing claims into the same forum and gave city officials specific questions to pursue: what companies disclose after incidents, whether audits can be genuinely independent, how workers report safety concerns and which remedies are practical when harm occurs.
Sources: Associated Press report on the October 5 testimony; New York City Council hearing agenda and bill summaries. Reporting reviewed October 6, 2026.