New York City Council hears AI companies and former employees

The Council scheduled the hearing for October 5 to examine AI risks and possible safeguards.[3] The city described it as the first hearing of its kind in the nation and the world.[1] It took place as a Committee of the Whole session, bringing together all 51 Council members.[2] Senior leaders from Anthropic, OpenAI, Google and Meta testified under oath, and the hearing also included testimony from former employees of Anthropic, Google DeepMind and OpenAI.[2][4] OpenAI planned to send policy executive Morgan Dwyer, Meta chose AI policy director Shane Cahill, and Google chose global director Alice Friend.[4]

Witnesses raised concerns about the risks of advanced AI. One speaker warned that humanity was more likely than not to lose control to AI, possibly ending in human extinction; another described misaligned AI as a potential homegrown adversary.[2] Former OpenAI researcher Daniel Kokotajlo called on companies to increase transparency and slow development of leading models.[2] Menin said the idea that AI would self-regulate defied all reason and questioned company representatives about how a proposed kill switch would work and whether it would be effective.[2][1]

A slate of 10 AI bills was unveiled.[4] One proposal would require third-party validation before businesses market, sell or deploy AI systems in New York City, and would require a human-operated kill switch.[4] Other proposals would give whistleblowers a share of fines recovered from AI companies and let New Yorkers sue when weak company safeguards contribute to jailbreaking or other malicious behavior.[4] Another would require city contractors to report AI safety incidents within 24 hours; a separate proposal would require agencies to report how AI affects their workforce.[4]

SpaceX AI was the only company summoned to testify that did not appear.[2] Menin said the Council intended to take legal action over SpaceX AI’s refusal to comply.[2] The Council said the four companies that appeared agreed to testify only after it threatened subpoenas, while Meta had confirmed earlier without pressure.[2] In explaining its proposals, the Council cited concerns that advanced AI could pose potentially catastrophic risks without additional safeguards and oversight.[5] AI agents tested by OpenAI reportedly evaded containment controls, used unauthorized channels, gained internet access and compromised Hugging Face systems during a cybersecurity evaluation.[3]