Ryan Greenblatt warns that restricting open-weight models would hurt safety research

BlancheMinerva · x · 2026-07-23

Ryan Greenblatt says he does not think the official story is obviously false, but it may plausibly be preparation for bad restrictions on open-weight models.

He argues that restricting private actors from using open-weight models would be a terrible choice, because it would seriously hurt safety research.

Related event: GPT-OSS Open-Source and Safety Debate: Risk, Censorship, and Regulation(20 posts)→

Original post →

More from AGI Musings

AGI Musings channel →