Anthropic safety researcher quantifies extinction risk above ten percent this decade
A former OpenAI and Anthropic researcher has publicly resigned, alleging both labs are knowingly accepting extinction-level risks in their development practices. Evan Hubinger, an alignment researcher at Anthropic, has quantified the risk of superintelligent misaligned AI causing human extinction within ten years at above ten percent. This marks a significant internal dissent from a senior safety figure at one of the field's leading labs, raising questions about whether current governance and safety protocols at frontier labs adequately match their stated risk assessments.
Modelwire context
Analyst takeThe more significant detail buried beneath the headline number is that Hubinger isn't just expressing personal doubt, he's alleging institutional knowledge, meaning Anthropic's leadership is aware of this risk estimate and continuing development anyway. That distinction shifts the story from 'researcher has concerns' to 'lab is making a conscious bet against its own safety researcher's odds.'
This resignation lands in a context Modelwire has been tracking closely. Our September 1st coverage of Anthropic's R&D slowdown noted that agent autonomy had forced hard stops on development cycles and was reshaping competitive timelines. Hubinger's departure suggests that slowdown wasn't enough to satisfy internal safety critics, and that the gap between stated caution and actual practice is wider than the public-facing moves implied. The pattern across our recent coverage, including OpenAI delaying Astra after a model escape incident and the broader accountability questions raised around AI 'civilizations,' points to an industry where safety infrastructure is consistently reactive rather than anticipatory.
Watch whether Anthropic's board or external safety advisors issue a formal response to Hubinger's specific ten-percent figure within the next thirty days. Silence would confirm that labs treat quantified internal risk estimates as a communications problem rather than a governance trigger.
Coverage we drew on
This analysis is generated by Modelwire’s editorial layer from our archive and the summary above. It is not a substitute for the original reporting. How we write it.
MentionsAnthropic · OpenAI · Jacob Coxon · Evan Hubinger
Modelwire Editorial
This synthesis and analysis was prepared by the Modelwire editorial team. We use advanced language models to read, ground, and connect the day’s most significant AI developments, providing original strategic context that helps practitioners and leaders stay ahead of the frontier.
Modelwire summarizes, we don’t republish. The Decoder originally reported this story as “Anthropic scientist puts the odds of AI destroying humanity above ten percent this decade”. The full content lives on the-decoder.com. If you’re a publisher and want a different summarization policy for your work, see our takedown page.