Summary
Regulates frontier (highly advanced, large-scale, general-purpose) models and their large developers. Requires a published safety framework, transparency reports with catastrophic-risk summaries, 15-day critical-incident reporting to California's Governor's Office of Emergency Services, whistleblower protections, and penalties up to $1M. Directs the Government Operations Agency to design a “CalCompute” cloud (if funded).
Healthcare Implications
Indirect but meaningful for health AI. Safety frameworks and transparency reports strengthen due-diligence when procuring or integrating frontier models in clinical tools. Incident reporting to California's Governor's Office of Emergency Services and whistleblower protections help surface risks with public-health impact. CalCompute may expand academic/health-system access to compute for R&D and evaluation.
Operational Implications
- Large developers must maintain published safety framework for frontier models.
- Transparency reports with catastrophic-risk summaries; 15-day critical-incident reporting to Governor's Office of Emergency Services; whistleblower protections; penalties up to $1M.
- Government Operations Agency directed to design CalCompute cloud. Healthcare relevance indirect: due-diligence when procuring frontier models for clinical tools.