California moved two dozen AI bills in a single day
Appropriations committees in both chambers cleared more than twenty AI bills on 13 August, covering chatbots, workplace surveillance, therapy transcription, toys and real-estate ads. The one requiring developers to document copyrighted training data was held.
On 13 August the appropriations committees of both California chambers worked through the AI backlog, and the sheer count is the story: more than twenty bills advanced in a day, two more went to the Governor, and a handful were held.
Read the list of survivors and a strategy appears that nobody announced. SB 867 would keep companion chatbots out of toys. SB 903 regulates AI transcription in mental-health therapy. AB 2025 requires disclosure when AI is used in real-estate promotional material. AB 1883 restricts workplace surveillance. SB 951 requires ninety days' notice of digital displacement. None of these is an AI safety law. Each is an existing body of law — consumer protection, employment, health privacy, advertising — being extended to cover a new instrument.
Now read what did not advance. AB 412, which would have required AI developers to document the copyrighted material in their training sets, was held in committee. So were AB 2545 (an AI worker-impact assessment) and SB 1146 (AI health-product advertising claims). The pattern is not subtle: rules about how a model is used are moving; rules about how a model is built are not.
That asymmetry matters more than any single bill. Deployment rules are enforceable against a company with California customers, and there are many. Training-data rules would reach into the construction of models built elsewhere, and they keep dying at the committee stage. The disclosure law that did take effect this month follows the same line — it governs what a deployed system must tell a user, not what went into it.
Two bills cleared entirely and went to the Governor: AB 1651, on AI in the State Bar exam, and SB 928, requiring that California State University instructors be human. Both passed their chambers without a dissenting vote. When legislators agree unanimously it is usually because the question was narrow.
Transparency Coalition — AI Legislative Update: August 14, 2026 → · Latham & Watkins — California Assumes Role as Lead US Regulator of AI → · O'Melveny — California Continues its Push to Regulate AI →