Researcher pushes back on AI biorisk alarmism: threat vectors haven't changed since o1

anshulkundaje · x · 2026-09-16

Responding to the reasoning-scaling discussion around OpenAI's o1, researcher David Bellamy argues that AI biorisk alarmists owe an explanation of 'why now': it was evident over two years ago that RL training with test-time 'thinking' would rapidly improve cheap-to-iterate, easy-to-verify tasks — which is why he pivoted to that paradigm himself. But biorisk barely relates to these capability trends, and threat vectors haven't changed: COVID-19 predated ChatGPT. The right move, he argues, is strengthening biodefense regardless of AI.

Original post →

More from AGI Musings

AGI Musings channel →