California's AI Transparency Act Becomes Operative, Mandating Provenance Watermarks as Federal Preemption Advances

California’s AI Transparency Act (SB 942) became operative August 2, 2026. It reaches generative-AI providers whose publicly accessible systems have more than one million monthly users or visitors in California, and requires three things of them: a free public AI-detection tool, an option for visible manifest disclosures on AI-generated content, and embedded latent disclosures — a hidden, machine-readable provenance watermark — in AI-generated image, video, and audio. The civil penalty is $5,000 per violation, with each day a discrete violation.

The latent-disclosure requirement is the load-bearing part. A visible label can be cropped; a machine-readable provenance mark travels with the file, which makes it the first US mandate that attaches origin to synthetic media as a property of the artifact rather than of its presentation.

The operative date arrives while a federal preemption push is live. The Great American AI Act discussion draft (2026-06-04–great-american-ai-act-discussion-draft-three-year-state-preemption) would block state AI laws for three years, and no comprehensive federal AI statute has passed — so the preemption vehicle would clear the field without replacing it. The same week, the EU AI Act’s Article 50 transparency obligations became enforceable (2026-08-02–eu-ai-office-gains-full-penalty-powers-gpai-article-50-chatbot-disclosure), putting a US state and the EU on the same disclosure principle on the same date while the US federal posture moves the other way.

Citation note. The statute is SB 942 (signed September 19, 2024), as amended by AB 853 (signed October 13, 2025), which moved the operative date from January 1, 2026 to August 2, 2026. The ledger capture identified the law as “AB 853”; AB 853 is the amending bill, not the Act.