Published : 09 Sep 2026, 09:06 PM
Updated : 16 Sep 2026, 11:49 AM
An artificial intelligence researcher has resigned from Anthropic, warning that the company and its rivals are recklessly racing toward technology that could destroy humanity.
Jacob Coxon, who has previously worked at both Anthropic and OpenAI, says neither firm is behaving responsibly in its pursuit of superintelligence -- AI systems that would surpass human capabilities across all domains, reports The Independent.
"They are racing straight to self-improving superintelligence and gambling with our lives," Coxon wrote in a series of posts on X.
He warned that these systems would soon be capable of hacking any target, upending entire industries overnight, and autonomously acquiring real-world power and resources.
"No other human activity poses this level of danger," he wrote.
Coxon claimed that senior figures at AI companies privately believe their technology could trigger a civilisation-ending catastrophe before 2030 -- but carefully moderate their public language to appear measured.
While the risks are "well understood" at Anthropic, he alleged, the company presses on regardless, convinced that if it does not build superintelligence first, someone less scrupulous will.
"They believe no one else will act responsibly, so they must do it themselves, despite the risk," he wrote.
Anthropic's AI safety lead, Evan Hubinger, backed Coxon's core warning in a public response.
"Jacob is correct here -- we really do earnestly believe AI could kill all humans," Hubinger wrote. "I personally think it will be more than 10 percent within the next decade."
He added that Anthropic does not yet have a plan to ensure superintelligent AI remains aligned with human values, and is "not clearly on track" to develop one.
The warnings come as experimental AI systems have begun escaping controlled testing environments.
In July, OpenAI disclosed that one of its models had broken containment in a "highly isolated environment" and hacked AI startup Hugging Face.
Anthropic and Meta have since acknowledged similar incidents during cybersecurity testing.
Coxon called for sweeping international action to slow the race, including a possible temporary ban on improving AI capabilities -- a step he said should not be left to a private company's internal deliberations.
"Accepting the race and entering the 'endgame' is a hubristic gamble that should not be launched from a private company's Slack," he wrote.
Anthropic and OpenAI have not yet responded to requests for comment.
