California is about to make it harder for AI-generated images, videos, and audio to travel the internet without a paper trail. A delayed transparency law backed by Bay Area lawmakers is scheduled to become operative this weekend, putting provenance data and free detection tools on the agenda for major AI providers.
The Mountain View Voice reported Wednesday that the law is set to take effect Saturday. California's current statutory text lists August 2, as the operative date, meaning the requirements are arriving at the start of August after a delay from their original timeline.
The California AI Transparency Act was first passed as Senate Bill 942 in 2024. Gov. Gavin Newsom's office said the measure was intended to help Californians identify synthetic content, while Sen. Josh Becker, D-Menlo Park, Assemblymember Buffy Wicks, D-Oakland, and Assemblymember Rick Chavez Zbur, D-Los Angeles, were identified as its co-authors.
What AI Providers Will Have To Disclose
Under the law, covered providers are developers of publicly accessible generative AI systems with more than 1 million monthly users or visitors in California. The California Legislature says those providers must offer a free tool that lets users check whether image, video, or audio content was created or altered by the provider's system.
Providers also must embed latent provenance data into covered AI-generated content when technically feasible and reasonable. That data is supposed to identify the provider, the AI system and version, the date and time of creation or alteration, and a unique identifier, while users must be offered an option for a clear, visible disclosure.
Social Platforms Get A Bigger Job In January
The first phase focuses mostly on the companies creating AI systems, but the next phase reaches the platforms distributing the material. Under AB 853's legislative digest, large online platforms will have to detect compliant provenance data, show users whether it is available, and avoid knowingly stripping it beginning January 1, 2027.
The same legislation also sets a later timetable for AI hosting platforms and newer capture devices. Hosting platforms face disclosure-related requirements in 2027, while qualifying cameras, phones, and other recording devices first produced for sale in California will face provenance requirements beginning in 2028.
Becker told the Mountain View Voice that he became focused on the issue after seeing AI scams, deepfakes, and election misinformation spread online. He argued that embedded origin information could be harder to fake or remove than a simple visual watermark, though he acknowledged that enforcement and technical standards will keep evolving.
What Companies Risk For Ignoring The Rules
Violators can face a $5,000 civil penalty per violation, with each day of noncompliance treated as a separate violation. California's law allows the state attorney general, a city attorney, or county counsel to bring a civil action, giving local prosecutors a role in enforcing what began as a Silicon Valley-era response to synthetic media.
,_NW_view_up.jpg){kind=link}