Generative AI companies with more than one million monthly users must now offer free tools that let the public check whether images, video, or audio were made or altered by AI, under the first phase of the California AI Transparency Act that took effect August 1, 2026. Senate Bill 942, signed in 2024, carries fines of up to $5,000 per violation per day, with a second phase requiring visible AI labels starting January 1, 2027.
The law targets deepfakes, digital scams, and disinformation campaigns. State Senator Josh Becker championed the measure, and Governor Gavin Newsom signed it in 2024. It applies to covered providers: companies with generative AI systems that draw more than one million monthly users or visitors and are publicly available in California, which includes several major platforms used daily.
Under the first phase, covered providers must offer a free verification tool. Users can upload an image, video, or audio file, or paste a link to a post, to check whether the material came from one of the company's models and learn about its digital provenance. The law bars these tools from retaining submitted files longer than necessary and prohibits them from collecting personal information from users.
The second phase begins January 1, 2027. Platforms will then have to add visible labels to AI-generated content, making it easier for users to identify synthetic material without running a check.
What the law means by "digital provenance"
The law centers on digital provenance data: information embedded in content or its metadata that shows who created it, which AI system generated it, when it was made, and whether it was later modified. The goal is to give content a verifiable history, particularly for sensitive situations such as elections, public health campaigns, or material that affects user safety.
Where technically feasible, AI-generated content must include the provider's name, the system version used, the creation date, and a unique identifier for tracing its origin.
Licensing obligations and fines
SB 942 also applies to companies that license their AI systems. Licensors must ensure the licensed technology keeps the required identification mechanisms, and if a licensee modifies the system so it can no longer include mandatory tags or provenance data, the licensor has 96 hours to revoke the license.
Penalties can climb quickly. Civil fines reach $5,000 per violation, and each day a violation persists counts as a separate offense. The California Attorney General, city attorneys, and county counsels across the state can all bring enforcement actions.
Why this matters for legal professionals
Legal teams should first determine whether their organization meets the covered-provider threshold of one million monthly users. If it does, counsel needs to confirm that free detection tools are live, provenance information is embedded where technically feasible, and licensing agreements include the 96-hour revocation requirement. Because fines accrue per day and multiple state and local prosecutors can enforce the law, a single compliance gap carries real liability.
For outside counsel, SB 942 is now part of the due diligence checklist for AI procurement and vendor agreements. California is again positioning itself as a pioneer in AI regulation, and the enforcement machinery under this law gives that position practical weight.
Your membership also unlocks:
![[Wafer: response was truncated before the model finished its internal reasoning. Increase max_tokens, or disable thinking on this model (e.g. chat_template_kwargs.enable_thinking=false), then retry.]](https://cdn.completeaitraining.com/news_images/wafer-response-was-truncated-before-the-model-finished-its-internal-reasoning-in_2026-08-09.jpg)