TechAnthropic insider warns AI could ‘kill all humans’. Should we be scared?

Anthropic safety researcher Evan Hubinger says AI is advancing so rapidly that he puts the chance of it killing all humans within the next decade above 10%. He considers the risk from existing models low, but worries the technology could soon improve itself enough to threaten humanity’s existence. Former Anthropic and OpenAI researcher Jacob Coxon has voiced a similar concern.