AI labels would cover more images, video and audio in California

In California, the real promise of an AI label is simple: help people tell when an image, video or audio clip came from a machine instead of a camera, microphone or human witness. The state’s AI Transparency Act already requires some generative AI content to carry provenance disclosures, and it already makes a free detection tool available to users. Those rules do not begin until Aug. 2, 2026, but SB 1000 would rewrite part of the system before that date arrives.

The bill, from Sen. Josh Becker, would drop the monthly-user threshold that currently decides who counts as a covered provider. That threshold is one of the law’s biggest gatekeepers, and removing it would move the focus away from how big a generative AI system is and toward whether its content is being labeled in a way people can actually use.

The gate gets wider

Under current law, a covered provider is a person who creates, codes or otherwise produces a generative AI system that has more than 1,000,000 monthly visitors or users and is publicly accessible in California. SB 1000 would delete that limit. That matters because the state’s disclosure rules, including manifest disclosures for some AI-generated image, video or audio content and latent disclosures for AI-generated image, video or audio content, only reach providers that fit that definition.

The bill would also rename the free checking feature from an AI detection tool to a disclosure verification tool. That is a small wording change on paper, but it points to the bill’s larger aim, which is to make the label itself easier to find and harder to sidestep when people are trying to verify what they are seeing or hearing.

A faster clock

SB 1000 is written as an urgency measure, so it would take effect immediately if enacted. That gives the rewrite more force than a typical cleanup bill, because it would land at the same moment California is preparing to turn on the existing transparency law.

Recorded votes show the bill cleared a floor vote. For users, the practical question is not whether AI will keep getting better at making convincing content. It is whether the rules around that content keep up quickly enough to give people a fighting chance to tell what is real.