Dietterich hits back at LeCun: lock up bioterror models, keep the rest open

tdietterich · x · 2026-11-22

Yann LeCun argued that safetyist/doomer rhetoric is convincing governments to back Anthropic's push to keep AI locked up, making models less open and ultimately less safe.

AI safety researcher Thomas Dietterich pushed back: safety arguments come in multiple flavors, some compatible with openness. He believes models with specialized bioterrorism knowledge should be locked up, but all others should be open — and that rapidly shipping poorly tested software is unwise.

Original post →

More from AGI Musings

AGI Musings channel →