California’s AI transparency law went live today. Every AI image, video, and audio generator with more than a million monthly users in the state now has to embed provenance data, run a free detection tool, and offer visible AI labels.
Miss it, and it’s $5,000 a day. The biggest name sitting out on day one? Midjourney.
What the AI transparency law requires
The law is SB 942, signed in 2024.
California pushed the operative date to August 2 specifically to line up with the EU AI Act’s Article 50. That means the same global AI providers are now under two provenance regimes on the same calendar day.
The AI transparency mandate isn’t subtle. Covered providers have to offer a free, public detection tool that lets anyone check whether an image, video, or audio clip came from their system.
They also have to embed a latent disclosure, which means machine-readable provenance data in the file itself. And they have to give users the option of a visible label that identifies content as AI-generated.
The practical standard for that latent disclosure is C2PA, the cryptographic provenance format backed by the Content Authenticity Initiative. OpenAI, Google, and Adobe already ship C2PA manifests or SynthID watermarks on outputs, which is why their compliance posture looks ready on paper.
The threshold is per system, not per company. OpenAI, Google, Meta, Adobe, ElevenLabs, Stability AI, and Anthropic all clear the one-million-user bar comfortably. So do consumer products like Midjourney and Character.AI.
The law does not carve out open-source developers, nonprofits, or free products. If a system clears the user threshold, it’s covered, even if nobody is paying for it.
The $5,000-a-day enforcement bite
The penalty is where this gets real. The bill text sets a civil penalty of $5,000 per violation, with each day of non-compliance treated as a separate violation.
The California Attorney General, any city attorney, and any county counsel can bring enforcement actions. There’s no notice-and-cure window, and no private right of action, but the fee-shifting means prevailing plaintiffs recover attorney’s costs.
Do the math on a provider that sits out for a month. Thirty days without a compliant detection tool is $150,000 in exposure before legal fees, per Startup Fortune.
Midjourney is the gap
The most prominent compliance gap is Midjourney. As of today it still ships no C2PA content credentials and no documented pixel watermark, according to TechTimes.
What this means for people who make AI images
If you generate an image in ChatGPT or a Google product, that image now carries a C2PA manifest and a SynthID pixel watermark.
Anyone can check it through the provider’s detection interface, whether you chose to disclose it or not.
That’s the practical shift. The AI transparency push moves disclosure from a user choice to a default baked into the file.
For creators, that means generated content has a paper trail. For platforms and apps that redistribute AI media, the provenance question is no longer optional.
This is the same trust problem that keeps surfacing in AI media.
I wrote about the AI music track that hit the Hot 100 with nobody able to prove it was fake.
And about Google’s Earth AI image tool dying in a day over misinformation fears. Provenance data is the mechanism those stories kept pointing at.
The platform problem, and what comes next
Here’s the catch the law hasn’t solved yet. C2PA metadata gets stripped when images are uploaded to Instagram, reposted on X, or sent through WhatsApp.
A provider can be fully compliant at the point of generation and still have its provenance destroyed before the image reaches most viewers.
That’s exactly what Startup Fortune flags. The January 1, 2027 phase of the law adds obligations for large online platforms to detect provenance data, surface it, and stop knowingly stripping it. Until then, enforcement covers the provider side only.
There’s also an urgency bill, SB 1000, that passed the California Senate 33-1 in May and could rewrite parts of SB 942 if it clears the Assembly and gets signed.
Anyone building compliance plans on today’s text should check the current status through the California Legislative Information portal before relying on it.
Bottom line
The AI transparency era started with real teeth today. California is the first US state to enforce a comprehensive provenance mandate, and it picked a date that syncs with the EU’s.
Midjourney is the test case. Everyone in the AI media space is watching whether enforcement actually lands. I’d bet on the fines arriving before the platform fix does.
