Tomorrow, August 2, 2026, two major AI transparency rules take effect simultaneously: the EU AI Act’s Article 50 obligations (which we covered on July 20) and California’s AI Transparency Act. The EU angle got most of the attention. The California angle is worth understanding on its own terms.

California’s AI Transparency Act — originally passed as SB 942 and updated by AB 853, which pushed the start date from January 1 to align with the EU deadline — applies to any company operating a publicly accessible generative AI system with more than one million monthly users in California. If that’s you, three things become legally required starting tomorrow.

What the Law Actually Requires

A free AI detection tool. Covered providers must make available, at no cost, a tool that lets users check whether image, video, or audio content was created or altered by your generative AI system. The tool must also expose provenance data — metadata about the AI system that produced the content.

Manifest disclosure. When your system generates content, users must have the option to include a visible label identifying it as AI-generated. This is user-facing and optional for the user to turn on, but the capability must exist.

Latent disclosure. AI-generated images, video, and audio must carry embedded signals — watermarks, content credentials, or equivalent metadata — that allow downstream detection even when the visible label isn’t present.

There’s also a flow-down requirement: if you license your generative AI system to third parties and you know they’re not implementing these disclosures, you must revoke the license within 96 hours.

The Penalty Math

Enforcement is handled by the California Attorney General, city attorneys, and county counsel. The civil penalty is $5,000 per violation. The critical detail: each day of non-compliance is a separate violation.

A business running an AI system without the required detection tool that gets flagged in month one isn’t looking at a single $5,000 fine. They’re looking at 30 separate violations before the complaint is even filed.

What’s Still Uncertain: The SB 1000 Factor

Here’s the part most compliance guides aren’t mentioning: the California legislature is currently considering SB 1000, a pending urgency bill that would significantly revise what SB 942 requires. As of June 2026, SB 1000 would remove the one-million-user threshold entirely, delete the manifest disclosure requirement, and revise the rules around detection tools and latent disclosure.

If SB 1000 passes — and urgency bills move fast — the law that takes effect tomorrow could look meaningfully different within weeks. That doesn’t mean businesses should wait. Urgency legislation is uncertain by definition, and regulators aren’t going to pause enforcement while the legislature debates.

What It Means for Businesses Using Generative AI

The direct compliance burden lands on the platforms — the companies building and operating the AI systems themselves. If you’re using third-party AI tools (OpenAI, Anthropic, Google) to generate content in your business, the platform obligations sit with those providers, not with you.

But this doesn’t mean the law is irrelevant to every other business.

The latent disclosure requirements mean that AI-generated content will increasingly carry verifiable machine-readable signals. Detection tools will become more capable at identifying AI-origin content. The businesses caught without disclosure frameworks — the ones whose AI-generated marketing copy, voice agents, and automated communications don’t identify themselves as AI — are building reputational risk that goes beyond what any regulator can fine them for.

The EU AI Act made the same point for European operations. California’s law makes it for your entire North American presence if you have more than a million monthly users.

The Practical Checklist

If you operate a covered AI system, the August 2 requirements are specific. You need a live detection tool, manifest disclosure capability, and latent signals embedded in generated media. If you’re not there yet, document what you have, what you’re fixing, and your timeline — regulators in enforcement proceedings look for evidence of good faith, not just end-state compliance.

If you use AI-generated content in your business but don’t operate a covered AI system, this is still a good moment to review how your AI-generated communications are disclosed to customers. Not because the law requires it today, but because your customers already expect it, and the regulatory trend in every major market is pointing in one direction.

What This Means for Business

The dual enforcement date — EU and California, same day — was not a coincidence. The California legislature aligned AB 853’s effective date with the EU AI Act deliberately. The global market is starting to converge on disclosure as the baseline expectation for AI systems that interact with people.

Businesses that get ahead of disclosure — that make AI identification a feature rather than a compliance checkbox — tend to find it improves customer trust rather than degrading experience. Voice AI agents that disclose their nature upfront don’t lose meaningful conversion rates. AI-generated content that carries honest attribution builds brand credibility in markets where cynicism about AI is high.

The regulation is catching up to where customer expectations already are. That’s not a threat to good AI deployments. It’s a validation of doing it right from the start.

Enterprise DNA helps businesses design AI operations that meet regulatory requirements without compromising the customer experience. If you’re navigating AI compliance for your operations, book a discovery call with our team.