SBS NEWS

News > World

"Don't Know How to Control AI Yet... Humanity Could Go Extinct," Warns Ex-Anthropic Researcher

Kim Young-a

Published : Oct 6, 2026 9:24 AM


▲ Jacob Coxon speaks at a New York City Council hearing.

A researcher who left Anthropic to warn against the dangers of the artificial intelligence (AI) development race stated at a New York City Council hearing that "it is more likely that humanity will lose control to AI."

Jacob Coxon, a former researcher at Anthropic, appeared at a New York City Council AI hearing on the 5th local time and made the remarks, saying, "We still do not know how to control any AI system."

Coxon stated, "At our current trajectory, it is more likely than not that humanity will lose control to AI, which could end in human extinction."

Criticizing the industry's approach as "extremely reckless," Coxon said, "The mindset of 'move fast and break things' (Silicon Valley's long-standing motto) may work for a photo-sharing app, but it does not work when building the most powerful technology in history."

Coxon also pointed out that "some form of speed limit is needed in the development of frontier models," noting that an AI "kill switch" (emergency shut-off mechanism) being reviewed by the New York City Council could help in the short term, but is not enough to guarantee safety in the long term.

When stepping down from Anthropic last month, Coxon criticized Anthropic and OpenAI, saying they are "gambling with our lives."

Coxon took the witness stand at the request of City Council Member Julie Menin, who is pushing for a package of AI regulation bills.

In a press conference prior to the hearing, Council Member Menin said, "The notion by AI companies that it will just stop at self-regulation will not work."

Alex Turner, a former Google DeepMind researcher, also testified via video link, saying, "China is not the only potential enemy. We are racing with a fairly high probability of creating our own enemy domestically, an unaligned AI," adding that "this could someday become more powerful than China."

Representatives from OpenAI, Anthropic, Google, and Meta Platforms also attended the hearing via video, taking the oath and testifying.

Among them, OpenAI, Anthropic, and Google decided to attend after the City Council warned it would issue subpoenas.

Council Member Menin stated that she will ask the court to enforce a subpoena against SpaceX's AI division, which did not respond to the attendance request.

While the Trump administration largely maintains a hands-off stance on AI regulation, state and local governments such as New York are speeding up independent regulations, widening the misalignment between federal and local authorities.

(Photo: Getty Images)