Developers of the most powerful AI models would need to publish their safety and security protocols to prevent catastrophic risks like weapons development under a measure cleared by California lawmakers on Saturday.
The state Senate approved revisions to the measure (SB 53) by state Sen.