Sam Altman says bio would be a reasonable bet if AI goes visibly wrong in 2026, and doubts blocking misuse will hold

OpenAI · Video interview

Read the source (opens in a new tab)
What happened
At an OpenAI town hall streamed on YouTube on January 26, 2026, CEO Sam Altman answered a student who runs a biosecurity startup asking how OpenAI approaches biosecurity.
Key points
  • Altman said one risk "we are quite nervous about is bio" and that "the models are quite good at bio."
  • He said he does not think the world's strategy of restricting access and adding classifiers "to not help people make novel pathogens" is "going to work for much longer."
  • He urged moving from blocking to resilience, as with fire codes, saying AI will be a problem for bioterrorism and cybersecurity but "also a solution to those things."
  • He said that if something goes "visibly really wrong for AI" this year, "I think bio would be a reasonable bet for what that could be."
Evidence
Town hall answer, one executive's view with no data. Quotes come from YouTube auto-captions, read rather than watched; attribution rests on the video title and the questioner addressing Sam.

See it in the latest record