← Back to articles

Anthropic researcher believes more than 10% chance AI 'could kill all humans'

2026-09-09 | original

read original get If Anyone Builds It, Everyone Dies" by Eliezer Yudkowsky and Nate Soares → more articles

Why This Matters

A senior Anthropic safety researcher publicly putting a >10% chance on AI causing human extinction within a decade is a striking admission from inside a leading lab, and it lands alongside reports that Anthropic withheld its newest model from the UK's AI Safety Institute. Together, the claims raise questions about whether frontier labs' internal risk assessments match their external transparency and cooperation with regulators.

Key Takeaways

Anthropic's Evan Hubinger says current models pose low risk but self-improving future systems could be an existential threat, putting the odds above 10% within a decade.
The FT reported Anthropic withheld its latest model from the UK AI Safety Institute; the Cabinet Office declined to confirm, saying it still works closely with industry.
A departing researcher accused both Anthropic and OpenAI of acting irresponsibly, while an academic tied reduced cooperation to US-China race dynamics and isolationism.

Worth a Look

If Anyone Builds It, Everyone Dies" by Eliezer Yudkowsky and Nate Soares — If a leading Anthropic safety researcher putting double-digit odds on human extinction makes you want the full argument, this book lays out the existential-risk case in plain language. It's the most talked-about read for anyone trying to understand why insiders at frontier AI labs keep sounding alarms.

See If Anyone Builds It, Everyone Dies" by Eliezer Yudkowsky and Nate Soares on Amazon → Affiliate link — we may earn a commission on purchases, at no extra cost to you. Product picked by AI based on this article; it is not a tested recommendation.

A top safety researcher at Anthropic has warned that AI is advancing so quickly he believes there is a greater than 10% chance it "could kill all humans" within the next decade.

Evan Hubinger said in a post on X, external that the risk from the models which currently exist was "low" but he was "worried" the technology might develop and improve itself soon to the point where it posed an existential risk to humanity.

It comes after the Financial Times reported, external Anthropic withheld its latest model from the UK's AI Safety Institute (AISI), one of the leading bodies in the world for assessing AI risk.

The BBC has approached Anthropic for comment.

Hubinger did not spell out how he thought AI systems could in future attack humanity.

His comments were in response to another post on X, external from Jacob Coxon, who described himself as an AI researcher who had just quit Anthropic, and previously worked at OpenAI.

"Neither company is acting responsibly," he wrote.

"These will soon be superhuman systems that can hack anything, revolutionise any field overnight, and acquire real power and resources."

OpenAI has been approached for comment.

A Cabinet Office spokesperson did not comment on whether the latest model had been withheld from the AISI - instead saying it "continues to collaborate closely with industry partners, including Anthropic, to make models safer".

... continue reading

Explore topics: anthropic evan hubinger ai safety institute openai jacob coxon