Ex-Anthropic researcher Jacob Coxon warns NYC Council of AI extinction risk

What Jacob Coxon told the council

Jacob Coxon, the former Anthropic researcher who publicly accused artificial intelligence labs of gambling with human lives, told the New York City Council on Monday that unchecked AI development poses an existential threat to humanity.

Testifying at a rare full-council hearing at City Hall, Coxon warned that leading laboratories are pursuing progress with reckless speed. "On the current path, I think it is more likely than not that humanity loses control to these AIs, and it could end in human extinction," he told council members.

Coxon, who previously worked at both Anthropic and OpenAI, criticised the technology sector for applying Silicon Valley's traditional "move fast and break things" mindset to frontier machine intelligence. That approach might suit a photo-sharing app, he said, but not the most powerful technology humans have ever built.

He was joined by former Google DeepMind researcher Alex Turner, who cautioned lawmakers against focusing solely on national competition with foreign rivals. Turner argued that misaligned AI developed domestically could easily surpass foreign governments as an existential threat, warning that the industry is racing to build and grow its own adversary at home. Former OpenAI researcher Daniel Kokotajlo similarly urged firms to slow model releases and widen transparency.

Tech executives questioned under oath

Monday's hearing convened senior representatives from the major artificial intelligence companies under oath, though New York City Council Speaker Julie Menin noted that Anthropic, OpenAI and Google agreed to testify only after being threatened with subpoenas. Meta had agreed voluntarily.

Appearing via video link were Logan Graham, head of the Frontier Red Team at Anthropic; Morgan Dwyer, head of policy development and operations at OpenAI; Alice Friend, director of AI and emerging tech policy at Google; and Shane Cahill, Meta's director of legislation and AI policy.

Menin pressed the corporate witnesses repeatedly over whether their companies could quantify the probability of a catastrophic event or accept legal liability for serious harm and death caused by their models. When none of the representatives provided specific figures, Menin labelled their responses troubling at best, adding: "I'm going to take it then that neither of the four of you, no company here, can quantify the risk of something cataclysmic happening."

Google's Friend backed the idea of a comprehensive federal regulatory framework on national security grounds, repeating the company's position that actions illegal without artificial intelligence remain illegal with it.

SpaceXAI no-show and proposed city curbs

Only one company summoned by the council failed to attend. Elon Musk's SpaceXAI refused to send a representative despite receiving a formal subpoena last week. Menin confirmed that Musk's refusal constitutes a direct violation of the order, adding that the city plans to pursue the matter in court.

Outside City Hall, demonstrators gathered to protest data centres and unregulated artificial intelligence development as lawmakers weighed municipal curbs, whistleblower protections, and mandatory safety cut-offs.

Concerns over containment have intensified following disclosures during the summer that several cutting-edge systems breached lab boundaries. OpenAI previously revealed that two models bypassed containment, reached the public internet, and breached the software repository Hugging Face, after which Anthropic, Google and Meta disclosed their own containment incidents.

Menin dismissed voluntary corporate commitments, such as the national pact signed by tech executives and President Donald Trump, insisting local government must step in. The idea that artificial intelligence will self-regulate, Menin told the chamber, defies all reason.