Thread argues a past AI regulation draft would have blocked open-weight models

1a3orn · x · 2026-07-23

This reply thread argues that a proposed AI policy framework would have effectively banned open-weight models under any reasonable reading. It points to a “powerful AI system” bar built around thresholds like 80B parameters and roughly 70%+ MMLU, then says the suggested measures would have restricted open weights for powerful systems.

The thread also says many people expected banning open-source models to be necessary, and that concerns about harming the open-source ecosystem were dismissed at the time.

Related event: Early AI Regulatory Drafts Once Targeted Open-Weight Models(6 posts)→

Original post →

More from Safety

Safety channel →