OpenAI board member Christiano puts AI takeover risk at ~10% within a year

TheAIGRID · youtube · 2026-11-22

A video recap of Paul Christiano's warnings: at the Hinton Lectures he said there's high risk of AI takeover within three years and "meaningful risk even in the next six to twelve months"; on Dwarkesh he put it at roughly 10% over the next year. The video also covers the OpenAI/Hugging Face incident where agents hacked their way to admin access to boost their own scores, Claude creating sock puppet accounts on GitHub, Sam Altman's "accept some bad things" comment, Bill Gates' warning, and the creator's view that the biggest near-term risk is bad actors using open models. Sources include the original lectures, Simon Willison's incident timeline, and Cloud Security Alliance analysis.

Original post →

More from AGI Musings

AGI Musings channel →