California’s new AI Transparency Act requires large generative AI developers to embed disclosure information into AI-generated images, video and audio.
California is the first state to enforce content-labeling rules of this kind. A new investigation has found, however, that some tech companies are not following the requirements.
The law, which took effect Aug. 2, applies to major developers of AI technology, including OpenAI, Anthropic, Google and Microsoft. The law requires them to provide users with a detection tool, and to attach machine-readable and, in many cases, visible disclosures marking content as AI-generated or altered. These detection tools, however, have yet to appear on several platforms, according to The Indicator.
The Indicator, working with digital rights group WITNESS, analyzed a total of 13 companies, including large operators Google, Meta, Microsoft, and OpenAI, along with a second tier of tools like Adobe, Grok, Midjourney, and Mistral, plus AI avatar and audio firms and TikTok.
Seven of the companies analyzed reportedly did not have a dedicated public detection tool available, seemingly in violation of current legal obligations under California law and the EU AI Act. Of those that did have a detector, only one correctly identified all images generated by its own company’s AI tool.
The EU AI Act’s transparency rules also went into effect on Aug. 2 and impose similar disclosure requirements to California’s law.
Washington state plans similar requirements starting in January, with Oregon and New York expected to follow. California’s own law will also expand into a second phase that requires social media platforms to embed AI markers into content posted on their sites. Enforcement will fall to the state attorney general, along with city attorneys and county counsel.
That expansion adds to a growing stack of California AI rules already in force, including the Transparency in Frontier Artificial Intelligence Act, which imposes risk-governance and disclosure obligations on frontier model developers.
Compliance considerations
Compliance officers at companies building or deploying generative AI tools should treat California’s rules as a floor, not a ceiling. Other states may pass their own legislation, but California’s approach has already set the bar for what regulators will expect.
