California Enacted AI Bills. Now Officials Must Define Them
2 days ago
- #California laws
- #technology policy
- #AI regulation
- California Gov. Gavin Newsom signed AI bills on catastrophic risk (SB53), content provenance (AB853), pornographic deepfakes (AB621), and chatbot companions (SB243).
- Key ambiguities in the laws include definitions of 'frontier model' and 'reasonable measures,' which need clarification for effective implementation.
- SB53 regulates AI developers creating frontier models, defined by computational resources (10^26 FLOPS), but faces challenges with open-weight models and model distillation.
- SB243 requires chatbot developers to take 'reasonable measures' to prevent sexually explicit content for minors, with industry practice likely defining 'reasonable.'
- AB853 mandates AI-generated content labeling 'to the extent technically feasible,' but current provenance technologies have limitations, requiring clearer standards.
- AB621 addresses nonconsensual deepfake pornography, requiring express written consent, but needs guidance on enforcement to reduce litigation uncertainty.
- Effective dates for the AI bills range from 2026 to 2028, with phased implementation for some laws like AB853 and SB243.
- Officials must clarify terms like 'frontier model' and 'reasonable measures' to avoid regulatory inefficacy and prolonged litigation.