California's AI Transparency Act is live, and a rewrite is with the governor
Watermark and disclosure duties began in August, and the operative requirements may change at short notice.
Background
California's AI Transparency Act, as amended by AB 853, became operative on 2 August 2026 for covered generative AI providers. A significant amendment bill was then sent to the governor, who had until 30 September to sign or veto it.
Three key points
1. What the operative law requires. The Act as amended obliges covered generative AI providers with more than one million monthly users to offer detection tools for AI-generated content and to make disclosures available. Duties for large online platforms to apply detection tools follow in 2027, and for capture device manufacturers in 2028.
2. A rewrite is pending. SB 1000, an urgency bill, would remove the user threshold, replace the detection tool requirement with a disclosure verification tool and drop the optional visible disclosure provision. It would take effect on signing, so the operational requirements could change at short notice.
3. The practical impact on advertisers is indirect. The law applies primarily to AI system providers rather than advertisers directly. Its effect on advertising workflows comes through the tools creative and media teams use, and what detection and disclosure capabilities those tools offer.
Why this matters
California sets the pace for US AI regulation, and the requirements are still in motion. Advertisers using AI content tools at scale should ask vendors how they are responding to both the current operative rules and the pending amendment before assuming the picture is settled.
Sources: Cooley; Vorp Labs, September update.
