California's SB 1000 Overhauls AI Transparency Rules, Effective Immediately

Enacted on September 30, 2026 as an urgency measure, SB 1000 immediately rewrites the California AI Transparency Act (CATA), which took effect weeks earlier and was expanded by AB 853. Three changes matter most. First, CATA's one-million-monthly-user threshold is gone: any person that creates, codes, or produces a GenAI system publicly accessible in California is now a "covered provider," and the exemption for non-user-generated television, streaming and movie content narrows to video games only. Second, the free "AI detection tool" becomes a "disclosure verification tool," with carve-outs for minor modifications, express permission to rely on third-party tools, a bar on outputting personal information absent express consent, and tighter retention limits. Third, CATA's optional manifest disclosure is eliminated; latent disclosures must interoperate with widely adopted standards, must state whether the system created or altered the content, and — from January 1, 2029 — whether the system is primarily assistive technology. Licensee governance is rebuilt around notice, a 72-hour provider response window, a 96-hour licensee cure period, and reporting to the California Attorney General, with a temporary $50,000-per-violation-per-day penalty for false assistive-technology claims. AB 853 dates (January 1, 2027 and 2028) are unchanged. Clients should re-scope covered-provider status, re-test detection and provenance tooling, and re-paper GenAI license agreements now, and watch for guidance or litigation over pre-effective-date claims.