California became the first state in the nation to force the largest AI companies to label their machine-made images, video and audio on Saturday, as the core provisions of the California AI Transparency Act took effect after a seven-month delay. Beginning August 2, 2026, generative-AI providers that reach more than one million monthly users in California must embed provenance data in the content their systems produce and give the public a free tool to detect it.
The law, better known by its bill number SB 942, was signed by Governor Gavin Newsom in September 2024 but was originally scheduled to become operative on January 1, 2026. A companion measure, AB 853, signed October 13, 2025, pushed the compliance date to August 2 and layered on a phased set of new obligations that will reach hosting platforms and device makers in 2027 and beyond. The result is that the first binding, consumer-facing AI-labeling regime in the United States is now live in the country's largest economy and the home of most of the companies it governs.
What the law requires
SB 942 draws a line between two kinds of disclosure. A "manifest" disclosure is the visible label a user can see and choose to apply, telling a viewer that a piece of content was created or altered by generative AI. A "latent" disclosure is the hidden, machine-readable signal baked into the file itself, carrying provenance information such as the name of the provider, the system used and the date the content was produced. That embedded data is designed to survive downloads and re-uploads and, in practice, is expected to follow the specification developed by the Coalition for Content Provenance and Authenticity, the industry group known as C2PA whose standard has become the de facto template for content credentials.
Covered providers must also publish a free, publicly accessible AI-detection tool that lets anyone check whether an image, video or audio clip carries the latent disclosure and, if so, surface the provenance details. The requirements apply only to "covered providers," defined by the one-million-California-user threshold, a bar meant to capture the largest model developers while sparing startups and smaller shops.
Enforcement is not toothless. The statute authorizes civil penalties of up to $5,000 per violation per day, and it hands enforcement power to the California Attorney General as well as to city attorneys and county counsels, who may sue on behalf of state residents. Prevailing plaintiffs can recover attorney's fees and costs on top of the per-day penalties.
"Artificial intelligence has become an integral part of our daily lives, influencing the products we use. It is crucial that consumers have the right to know if a product has been generated by AI," Senator Josh Becker, the Bay Area Democrat who authored SB 942, said in describing the bill's purpose. Becker has framed the measure as a response to deepfakes and AI-driven political misinformation, arguing that it will "advance provenance, transparency, accountability, and empower individuals to make choices aligned with their values."
A collision with Washington
The law arrives directly in the path of an escalating fight between Sacramento and Washington over who gets to regulate AI. On December 11, 2025, President Trump signed an executive order, "Ensuring a National Policy Framework for Artificial Intelligence," that directs federal agencies to challenge state AI laws deemed inconsistent with a national approach, establishes an AI Litigation Task Force to bring those court challenges, and threatens to condition some federal funding on states standing down from enforcement of conflicting rules.
California's recent AI statutes, including SB 53 and the transparency act, are widely viewed by legal analysts as likely targets. Whether the executive order can actually preempt state law is an open question; preemption ordinarily flows from an act of Congress, not from the White House alone, and a coalition of 36 state attorneys general wrote to Congress in November 2025 opposing efforts to bar states from regulating AI. Firms advising clients have cautioned that the order's practical reach is uncertain and that companies should not assume it relieves them of California obligations that are now in force.
That leaves covered providers in an awkward spot: comply with a live California mandate carrying daily penalties, or bet on a federal challenge that has not yet been filed, let alone won. Most compliance guidance circulating among law firms this summer has landed on the former.
What to watch
The immediate test is enforcement posture. The Attorney General's office has not signaled whether it will move quickly or grant an informal grace period as providers stand up detection tools and audit their provenance pipelines. Watch, too, for the first litigation under the December executive order and whether it names California by name. And keep an eye on the 2027 tranche of AB 853, which extends transparency duties to generative-AI hosting platforms and large online platforms, broadening the law well beyond the model developers now on the hook. For anyone tracking whether the U.S. gets AI content rules from the states or from Washington, California's August 2 start date is the opening data point.
"It is crucial that consumers have the right to know if a product has been generated by AI."— Josh Becker, California State Senator and author of SB 942