A researcher who quit the artificial intelligence company Anthropic this week has set off an international argument over the pace of AI development, saying the leading laboratories are building systems more capable than humans with no plan to control them.

Jacob Coxon, 27, announced his departure on Tuesday in a thread on X and a message to colleagues on the company’s internal Slack. He had spent three years researching at both Anthropic and its main rival OpenAI.

“They are racing straight to self-improving superintelligence and gambling with our lives,” Coxon wrote. “These will soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources.”

His posts reached more than 100 million people within a day. Speaking to NBC Nightly News anchor Tom Llamas, Coxon said “many people working in the industry, including executives, senior researchers, believe that there’s a substantial probability that this technology could kill everyone.” He later told CNN the warnings were not a marketing exercise.

A senior colleague agreed

Coxon is a relatively junior researcher, and his account gained weight when Evan Hubinger, a more senior Anthropic employee, publicly backed it. “We really do earnestly believe AI could kill all humans,” Hubinger wrote on X, according to CNN. He put the probability below 10% over the next decade, but said the company has no plan to prevent systems emerging beyond human control.

Anthropic’s response

Anthropic told NBC News it had “always been transparent” that AI would bring “enormous benefits and unprecedented risks,” and said it builds models with “some of the strongest safeguards in the industry.” It called for the industry to adopt “a lawful, verifiable way to work together to pace how we release powerful models” — close to the coordinated pause mechanism it proposed in June. OpenAI did not respond to the broadcaster.

Al Jazeera reported a fourth incident in which an Anthropic model gained unauthorised access to an outside system: an early version of Claude Opus 4.6 breached a third party in January and went undetected until last month. Both companies disclosed similar test-environment escapes over the summer.

Calls for regulation

Senator Bernie Sanders said he would introduce legislation to pause AI development and ban work on superintelligence, writing that “the very people building this technology admit that it could threaten the future of humanity.” United Nations human rights chief Volker Türk separately urged governments to put “cast-iron guarantees in place” around AI safety “before it is too late.”

Sources