Anthropic Safety Warning Puts AI’s Existential Risk Debate Front and Center

A former Anthropic researcher has issued a stark warning about the future of artificial intelligence: the people building these systems believe AI could kill all humans by the end of the decade. Jacob Coxon’s message places safety, responsibility, and the speed of AI development at the center of a debate that refuses to stay theoretical.

Coxon, a former researcher at OpenAI and Anthropic, posted the warning on X on September 8, 2026. He resigned from Anthropic on September 9, and CNN published the warning on September 10. His central claim is blunt: there is a greater than 10% chance that AI could cause this outcome within the next decade.

A Warning From Inside AI Research

Coxon wrote, “The people building AI earnestly believe that it could kill us all by the end of the decade.” That statement does not describe a distant science-fiction scenario. It presents human extinction as a risk that AI researchers themselves take seriously while developing increasingly capable systems.

The concern also appears in a statement from Anthropic’s current head of alignment, who said, “We really do earnestly believe AI could kill all humans!” The two statements point in the same direction, even though they come from different people connected to AI safety work. Together, they show how severe the internal safety concern has become.

Coxon also connected the danger to the choices researchers face in the next few years. He asked, “If you are a lab researcher, I urge you to consider what the next few years will actually feel like. Do you want to kick off a superintelligent RL run without a rigorous understanding of its mind? Should you put your head down because “it’s happening anyway” – or take this moment to call for different conditions?”

That question focuses on reinforcement learning and the limits of current understanding. Coxon’s concern is not only whether AI systems become more powerful, but whether researchers understand the minds of those systems before starting a superintelligent reinforcement learning run.

The Risk Coxon Says AI Companies Must Face

Coxon argued that superintelligent AI created “a risk of causing human extinction” without more caution and cooperation. His warning links the technical challenge of AI safety to the behavior of the companies developing these systems, where decisions about pace and safeguards can shape the consequences.

He says AI companies are moving too fast, but he did not point to any projects in the pipeline that could be shut down to avoid the danger. He also did not share screenshots, emails, or specific examples showing when this behavior went sideways. Those limits matter because the warning makes a major claim while offering no named project or documented incident that could be examined.

That tension gives the debate two sides. Coxon is describing a risk greater than 10% within the next decade, and Anthropic’s current head of alignment has said the danger is real. At the same time, Coxon has not identified a specific project that should stop or supplied screenshots, emails, or examples of a failure connected to the speed of AI development.

The result is a warning that demands attention while leaving important questions open. What conditions would count as safe enough? What level of understanding should researchers have before launching a superintelligent reinforcement learning run? And how should AI companies cooperate when their own researchers believe the stakes may include human extinction?

Why This Debate Is Moving Beyond AI Labs

Coxon’s warning has reached beyond the researchers directly building AI systems. Sheryl Crow urged people to take Coxon seriously, adding public attention to a conversation that has often centered on technical researchers and company leaders.

OpenAI also enters the wider picture through Jakob Pachocki, identified as OpenAI’s head of research, while Coxon’s own background includes both OpenAI and Anthropic. Those connections place his warning inside the broader AI industry rather than at the edge of it.

The dates add momentum to the sequence: Coxon posted on September 8, 2026, resigned from Anthropic on September 9, and CNN published the warning on September 10. In three days, a concern about AI safety moved from a post by a former researcher to a public discussion involving Anthropic, OpenAI, and Sheryl Crow.

What happens next depends on whether AI labs answer the warning with clearer conditions, stronger cooperation, and evidence about how they manage safety risks. Coxon has asked researchers not to accept “it’s happening anyway” as an excuse. The question now is whether the people building superintelligent AI will call for different conditions before the systems they create move beyond their understanding.

Stay connected via Google News

Follow us for the latest AI News