Proposal: a legal ontology separating open-weight hosts from frontier labs for AI liability

jd_pressman · x · 2026-09-13

jdpressman proposes a legal ontology as the first step toward satisfactory AI liability rules: distinguish a hosted service running an open-weight model it didn't train from a frontier lab that trains and serves its own model. He suggests a Section 230-style law for raw inference providers — but, as with social media algorithms straining 230's spirit, passing liability to users would require strict conditions.

Related event: Who's Liable When AI Misbehaves? X Users Debate Lab, User, and Model Responsibility(8 posts)→

Original post →

More from Safety

Safety channel →