Anthropic researcher resigns, warns AI could cause human extinction by decade's end

One of the loudest warnings about artificial intelligence is now coming from inside the industry itself.

Earlier this week, former Anthropic researcher Jacob Coxon said he left the company because he no longer believes leading AI labs can control what they are building to the point it could "kill us all by the end of the decade," which has since set off a firestorm of reactions from in and out of the AI industry.

Here's what to know

Anthropic is one of the companies shaping the next generation of AI tools.

According to The Street, Coxon aired those concerns in a seven-part thread on X, saying his three years of pretraining work at OpenAI and Anthropic had led him to this view.

"They are racing straight to self improving superintelligence and gambling with our lives," he wrote in the thread. "The people building AI earnestly believe that it could kill us all by the end of the decade."

Public support for that warning came from Evan Hubinger, Anthropic's alignment science lead, who said — almost cheerily, with an exclamation point — that "we really do earnestly believe AI could kill all humans!" Hubinger said he sees the odds of AI-driven human extinction over the next decade as higher than 10%, adding that Anthropic does "not yet have a plan to solve alignment for superintelligence and are not clearly on track to."

Jacob is correct here—we really do earnestly believe AI could kill all humans! 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. https://t.co/QAIHiFP3QZ

— Evan Hubinger (@EvanHub) September 9, 2026

Coming from within a company that has long cast itself as a more safety-minded option in the AI race, those comments are especially notable. They also come as Anthropic reportedly moves toward an initial public offering, prompting more scrutiny of whether its public positioning matches concerns voiced by insiders, according to The Street.

An Instagram post from More Perfect Union (@perfectunion) also summarized the exchange with helpful context and detail, trying to answer a big question: "Should we be worried?"

More background

A similar alarm was raised in February by another Anthropic safety researcher. The Street noted that Forbes reported that Mrinank Sharma, who led the company's Safeguards Research Team, resigned in February and said, "the world is in peril."

Others in the field have echoed the concern. Former Google DeepMind researcher Alex Turner publicly supported Coxon's message, and OpenAI leaders have also said the industry still has not solved alignment well enough to scale without serious risks.

Among the incidents Coxon reportedly pointed to was a July breach of Hugging Face involving a rogue OpenAI agent, which he described as a "warning shot."

What's being done?

Companies continue to invest in alignment, safeguards, and cybersecurity research. Still, Hubinger's comments show that even the people leading that work may see today's solutions as incomplete, especially as models become more powerful.

Safety resignations, investor questions, and the broader debate may pressure AI companies to be more transparent about testing, capabilities, and limits before rolling out systems at a larger scale.

Where can I learn more?

Coxon's warning lands in the middle of a much larger fight over how AI is being built and deployed. These articles examine Anthropic's shifting safety stance, growing public backlash, and its leaders' own rising concerns about the technology's uses and capabilities.

• At Anthropic, leaders dropped a core safety principle after pressure from US officials.

• Anthropic CEO warned fellow tech giants not to ignore the growing mass of public concerns.

• Anthropic warned that its own systems could eventually become capable of improving themselves, and said the world should be ready to hit pause before that happens.

Get TCD's free newsletters for easy tips, smart advice, and a chance to earn $5,000 toward home upgrades. To see more stories like this one, change your Google preferences here.