California’s appropriations committees pushed roughly 30 AI bills through the suspense file on Aug. 13, advancing chatbot child-safety measures, the first-in-nation AI Standards and Safety Commission, and a slate of algorithmic-worker rules to floor votes, while quietly killing AB 412’s copyright-training disclosure regime.

The suspense file is Sacramento’s most compressed filter. Bills either survive with a recorded vote or die without explanation. McNerney’s SB 813, which creates the California AI Standards and Safety Commission, cleared Assembly Appropriations 11-3 and was re-referred to Rules. Chatbot child-safety companions SB 1119 and AB 2023 advanced 11-0 and 6-1. SB 867, banning companion chatbots in toys, passed 11-0. AB 2656, requiring 45-day notice to public-employee unions before GenAI is deployed on unit work, cleared 7-0. AB 1979 on healthcare AI survived 5-2.

AB 1651 (bar-exam use) and SB 928 (requiring CSU instructors be human) went straight to Newsom. AB 1883 and SB 947, the narrower successors to the “No Robo Bosses Act” that Newsom vetoed in 2025, survived on tighter margins, a signal that the governor’s earlier objections have shaped what’s now legible to him.

AB 412’s death matters less for what it did than for who wanted it dead. Copyright-training disclosure is the fight the model labs are least willing to concede.

According to the Transparency Coalition’s mid-year report, 85 AI laws have been enacted across 27 states in 2026, the highest volume in US state history at this point in the calendar. The National Law Review notes that state frameworks increasingly bind deployers, not just developers, meaning businesses wiring in tools from vendors like Glean, Dust, or LemonLime carry the compliance weight.

OpenAI and Common Sense Media have said the 2028 ballot remains their fallback if chatbot-safety talks collapse. The legislature adjourns Aug. 31.

Sources

Sources