California's AI Transparency Act Is Now Live: What It Means for Brands Publishing AI-Assisted Content
California's new AI law requires disclosure labels on generated images, video, and audio.
CAITA defines a "covered provider" as any entity building a generative AI system with more than one million monthly users, publicly accessible within California. OpenAI, Anthropic, Google, and Microsoft sit squarely in scope. Most brands do not. But that gap closes through licensing contracts, which we will get to shortly.
The three obligations covered providers must meet as of August 2, 2026
All three requirements went live August 2, 2026: a free public AI detection tool covering image, video, audio, and combined media; a manifest disclosure giving users a visible AI-generated label; and a latent disclosure embedded in AI-generated content where technically feasible. One boundary that matters for content strategy: text-only output, blog posts, ad copy, email, currently falls outside CAITA's scope entirely.
How the C2PA standard is the technical infrastructure behind latent disclosures
C2PA is the open standard most covered providers use to satisfy latent disclosure. Content Credentials work like a nutrition label for digital media: cryptographically signed metadata recording a file's origin and any AI enhancements applied. TikTok has labeled over 1.3 billion videos using this system; Google has watermarked over 20 billion images via SynthID.
That raises an important question for anyone managing a content pipeline. C2PA data lives in the file wrapper, and most image compressors, CMS platforms, and social schedulers strip it by default. A compliant image at creation turns non-compliant the moment it hits a standard upload workflow.
How CAITA reaches brand content teams through licensing contracts
Section 22757.3(c) requires covered providers to contractually obligate licensees to preserve disclosure capability. If your team uses OpenAI, Anthropic, or Google tools, your license almost certainly prohibits stripping watermarks, and violating that clause triggers a 96-hour revocation clock.
The phased compliance timeline through 2028 and what each stage adds
August 2, 2026 is live. January 1, 2027, large online platforms must surface provenance data and cannot knowingly strip compliant credentials. January 1, 2028, camera manufacturers must embed latent disclosures by default for devices first produced for sale in California.
Penalties, enforcement authority, and what a $5,000-per-day structure means in practice
The civil penalty runs $5,000 per violation, each day counted separately. Thirty days of a missing detection tool equals $150,000 from that single deficiency alone. The California Attorney General, city attorneys, and county counsels all hold enforcement authority, and prevailing plaintiffs collect attorney's fees on top.
Where FTC requirements stack on top of CAITA for brand content teams
The FTC has issued requirements around AI disclosure obligations for content teams. The FTC penalty runs $53,088 per violation in 2026, and unlike CAITA, FTC obligations cover text-based content directly.
How CAITA relates to the EU AI Act and what dual-market brands need to know
Both laws share architecture, but a compliance program built for Article 50 will not automatically satisfy CAITA. CAITA's free public detection tool requirement has no Article 50 equivalent, and EU fines run up to 15 million euros or 3% of global annual turnover. If your workflow already includes C2PA metadata for EU compliance, the technical infrastructure overlaps; CAITA's contractual and detection-tool obligations still need separate attention.
Why consumer trust trends make proactive disclosure a brand strategy argument, not just a legal one
Consumer research has consistently found that a meaningful share of U.S. consumers say heavy AI use would decrease their trust in a favorite brand, and that concern has grown over time across age groups.
Brands that disclose AI use honestly, before regulators force them to, are building a position that voluntary disclosure eventually makes irreplaceable. Content automation platforms such as Letterstory, which runs the full lifecycle from drafting to publishing, are where those disclosure practices need to be baked in, not bolted on.
