Illustration by Megaton
Image: Illustration by Megaton
Regulation2-minute read

California's AI Transparency Act takes effect, mandating provenance in synthetic media

By Julius RobertTuesday, August 4th 2026

AI providers must now embed disclosures in images, video, and audio and offer free detection tools

Share
Subscribe to the Megaton newsletter

Get weekly AI video rankings, evaluation updates, and industry news delivered to your inbox.

AI providers must now embed disclosures in images, video, and audio and offer free detection tools

California's AI Transparency Act, known as CAITA, became active on August 1, 2026, moving AI content disclosure from voluntary industry practice to enforceable state law. The law targets large AI providers and covers three content types: images, video, and audio produced by AI systems.

The disclosure requirement runs in two directions. Providers must embed machine-readable provenance data, latent disclosures invisible to the naked eye, alongside visible disclosures that users can see directly. That dual-layer approach reflects a tension in synthetic media policy. Human-readable labels can be cropped or stripped, while embedded metadata persists through more distribution pathways.

CAITA also requires covered providers to make AI detection tools available at no cost. The free-tool requirement shifts the verification burden away from individuals who might otherwise lack access to commercial detection software.

The law applies to large AI providers, though the exact size thresholds that determine who qualifies as large remain unspecified. Some details of implementation scope will be tested in practice.

The provenance layer

Embedding provenance data in AI-produced content connects CAITA to a broader technical effort that predates the law. Industry coalitions and standards bodies have been developing content credentials and watermarking frameworks for several years. California's law now makes participation in some form of that infrastructure a legal baseline for qualifying providers in the state, rather than an optional industry commitment.

Editorial illustration for California's AI Transparency Act takes effect, mandating provenance in synthetic media
The law targets large AI providers and covers three content types: images, video, and audio produced by AI systems.

The first concrete test of CAITA's reach will come when California regulators or affected parties begin assessing whether large providers' existing disclosure and detection implementations satisfy the statute's dual-layer standard.

Subscribe to the Megaton newsletter

Get weekly AI video rankings, evaluation updates, and industry news delivered to your inbox.