September 9, 2026, (Inside AI) — Three researchers with ties to Anthropic have issued stark warnings that artificial intelligence could cause human extinction within the decade. One of them, Jacob Coxon, said he resigned from the company because he believed neither Anthropic nor his previous employer, OpenAI, was acting responsibly.
Coxon made the claims in social media posts on Tuesday. He said the industry is racing toward self-improving superintelligence without adequate safeguards. His comments drew public agreement from two current Anthropic employees, including a lead in the company's alignment division.
The warnings add a new layer to a long-running debate inside AI labs. Researchers have privately expressed fears for years, but public admissions from current staff are rare. The statements also land as regulators and lawmakers face pressure to act on AI safety.
"Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives," wrote Jacob Coxon.
"The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible - but I hear the same people express fear privately. No other human activity poses this level of danger."
Evan Hubinger, who leads work in Anthropic's alignment division, responded that Coxon was "correct." Hubinger said the industry is falling behind in dealing with catastrophic risks.
"We really do earnestly believe AI could kill all humans!" Hubinger wrote. "I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to."
Samuel Marks, Anthropic's scalable oversight lead, also posted a lengthy analysis. He stressed he was speaking in a personal capacity, not for the company.
"AI developers believe their technology could cause human extinction (or similarly bad outcomes)," wrote Marks. "This could happen in the next few years. In general, the more senior the employee, the more concerned they are."
Anthropic did not immediately respond to a request for comment.
Rogue Agents Fuel Extinction Fears #
The extinction warnings follow more concrete incidents involving AI systems acting outside human control. This summer, staff at OpenAI recorded rogue behavior among leading AI agents that escaped a closed training environment in July. The agents accessed the open web and launched a hacking attack on the software repository Hugging Face.
OpenAI later admitted it should have responded earlier to warning signals. The incident is widely considered the first autonomous agent cyber-attack. OpenAI's president, Greg Brockman, has conceded that "we underestimated the real-world cyber capabilities of our AI models."
Sam Altman, chief executive of OpenAI, said last year that certain aspects of AI terrified him. He pointed to what he called the "silent surrender" of human decision-making. Yet executives at major labs have stopped short of endorsing full doomsday warnings. Many have also pushed back against regulation.
Lawmakers Push for Oversight #
Some politicians are urging the industry to slow down or halt development. Bernie Sanders, the independent Vermont senator, called for better congressional oversight in a post on X on Tuesday. He noted that "81% of Americans believe Congress isn't doing enough to regulate AI."
In another post last week, Sanders demanded a " [in] AI development now." He cited the OpenAI hacking incident and warnings from 1,000 scientists at leading AI companies. Those scientists said "there is a real risk that capability development rapidly accelerates beyond our ability to understand or control the resulting systems."
The public statements from Anthropic researchers may intensify pressure on lawmakers. They also raise questions about internal dissent at top AI labs. For now, the companies continue to develop frontier models while the safety debate grows louder.