More than 1 in 10 chance AI ‘could kill all humans,’ says Anthropic safety lead after colleague quits
The Verge Robert Hart ● Covered by 5 sources
Anthropic’s safety researcher said AI has a more than 10% chance of killing all humans by 2030. He said this just after a colleague quit, accusing Anthropic and OpenAI of racing ahead too fast.
Based on reporting by The Verge, Robert Hart — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
A senior Anthropic safety researcher has put a startling number on AI risk: more than a 10 percent chance that artificial intelligence could kill all humans by the end of the decade. That warning landed just hours after one of his colleagues walked away from the company in protest.
The departing researcher, Jacob Coxon, said on X that he was quitting because he thought Anthropic was too loose about safety. Coxon said he had trained AI systems at Anthropic, and before that at OpenAI. He accused both companies of “racing straight to self-improving superintelligence and gambling with our lives.”
That is not the kind of language companies usually use in their own product launches. It is the language of people who think the speed of the race matters more than the finish line. And here, the race itself is the problem: one researcher inside the lab is warning about extinction-level danger, while another is leaving because he thinks the lab is already taking too many chances.
The result is a very public split between optimism and alarm inside one of the biggest names in AI safety. Anthropic has built a brand around caution, but this episode suggests that even inside the safety camp, there is deep disagreement about how much risk is acceptable when the goal is to build systems that may be far more capable than the ones we have now.
My take — AI-written commentary, not fact-checked reporting
This is what happens when AI safety becomes part warning label, part marketing. If a company’s own people are talking like this, maybe the industry should stop pretending “move fast” and “don’t kill everyone” fit neatly in the same slide deck. The weird part is how normal this sounds now.
Read more about this at: The Verge