California Enacted AI Bills. Now Officials Must Define Them (www.lawfaremedia.org)

California has recently enacted several significant AI-related bills, including regulations on catastrophic risk disclosures (SB53), content provenance (AB853), pornographic deepfakes (AB621), and chatbot companions (SB243) signed by Gov. Gavin Newsom. While these laws aim to enhance accountability and transparency in AI technologies, their practical implications hinge on how key terms—such as "frontier model" and "reasonable measures"—are defined by state officials. For instance, SB53 outlines a "frontier model" based on compute power thresholds, posing challenges for enforcement due to ambiguities in how developers account for compute use, particularly when fine-tuning open-weight models. The significance of these regulations lies in their potential to shape industry standards and compliance practices in the evolving AI landscape. Particularly, the definition of "reasonable measures" in SB243 mandates chatbot developers to prevent explicit content for minors, which could establish minimal safety standards based on current industry practices. Similarly, the feasibility assessments outlined in AB853 for AI-generated content labeling could influence how effectively platforms adopt and implement new technologies. Clear guidance from California’s attorney general will be crucial to navigate these complexities and ensure the laws achieve their intended protective goals without overburdening innovation.