California AI Transparency Act SB 942: Disclosure 2026

Quick Answer

The California AI Transparency Act, SB 942, requires large generative-AI providers to make AI-generated content detectable and disclosable, and after amendment by AB 853 it becomes operative on August 2, 2026, with additional obligations for large hosting platforms following on January 1, 2027. The Act applies to a 'covered provider' — a person that creates a publicly accessible generative-AI system with over one million monthly visitors or users in California — and imposes three main duties. First, a free, publicly available AI-detection tool that lets anyone check whether image, video or audio content was created or altered by the provider's system. Second, a manifest disclosure: covered providers must give users the option to include a clear, visible label identifying content as AI-generated. Third, a latent disclosure: covered providers must embed provenance information in AI-generated image, video or audio content in a way that persists and can be read by the detection tool. The watermarking and disclosure obligations apply to images, video and audio, not to text-only outputs. For advertisers, the Act is primarily an obligation on the AI tools they use rather than a direct duty, but its effect is concrete: AI-generated ad creative made with covered tools will increasingly carry embedded provenance that detection tools and platforms can read, which interacts with platform AI-label rules and other synthetic-media laws. The practical response is to know which of your AI tools are covered, expect provenance metadata in AI creative, keep manifest labels where required, and align with platform disclosure. Check creative language with the Keyword Risk Checker, define terms in the compliance glossary, and track the timeline on the Policy Change Tracker.

Predict ad disapprovals before you submit — try the free Meta Rejection Predictor

California's AI Transparency Act (SB 942) in 2026: Manifest and Latent Disclosure, Detection Tools and the August Start Date

What SB 942 Requires and When

California's AI Transparency Act, SB 942, is a law aimed at making AI-generated content identifiable, so that people can determine whether an image, video or audio clip was created or altered by a generative-AI system. It was adopted in 2024, and after an amendment by AB 853 its operative date became August 2, 2026, with a further set of obligations for large hosting platforms taking effect on January 1, 2027. The core of the Act places duties on the providers of large generative-AI systems rather than on the people who use them to make content.

The Act matters to advertising because so much ad creative is now produced or modified with generative AI, and SB 942 shapes how that creative is marked and verified at the source. Rather than requiring each advertiser to label its work, the Act requires the AI systems themselves to embed provenance information and to offer detection and labelling tools — which means the AI creative flowing into advertising increasingly carries machine-readable signals about its synthetic origin, whether or not the advertiser adds anything.

"The California AI Transparency Act requires covered providers to give consumers the tools to know whether content was created or altered by generative artificial intelligence.
— Summary of California SB 942"

This guide explains who counts as a covered provider, the three disclosure mechanisms — manifest, latent and the detection tool — and what the Act means for AI ad creative, before placing it alongside the other synthetic-media rules advertisers face. For the EU equivalent see the EU AI Act Article 50 guide, and for the detection technology the AI content-detection guide.

Who Is a Covered Provider

The Act's obligations attach to a defined actor called a 'covered provider', and getting this definition right is essential to knowing who must comply. A covered provider is a person that creates, codes or otherwise produces a generative-AI system that is publicly accessible within California and has over one million monthly visitors or users. This is a threshold aimed at the large, widely-used AI systems rather than every small tool.

The Covered-Provider Test

Element	Requirement	Effect
Creates the system	Creates, codes or produces a generative-AI system	Targets the maker of the model/tool, not the user
Public accessibility	Publicly accessible within California	Reaches widely-available consumer AI systems
Scale threshold	Over one million monthly visitors or users	Focuses on large, mainstream systems

The important consequence for advertisers is that the Act regulates the AI tool provider, not the advertiser who uses the tool to make an ad. If you generate ad creative with a large, widely-used AI image, video or audio system, that system's provider is likely the covered provider carrying the SB 942 obligations, and the provenance and disclosure features exist at the tool level. Advertisers are, in effect, downstream beneficiaries and users of those features rather than the regulated party — though, as discussed below, AB 853 adds hosting-platform duties that broaden the picture. Define the underlying concepts in the compliance glossary.

Manifest, Latent and Detection Disclosures

SB 942 works through three complementary mechanisms, each addressing a different way of surfacing that content is AI-generated. Understanding the distinction between them is key, because they operate very differently — one is visible, one is hidden, and one is a lookup tool.

The Three Mechanisms

Manifest disclosure: covered providers must offer users the option to include a clear, visible disclosure in AI-generated image, video or audio content, identifying it as AI-generated in a way understandable to a reasonable person and appropriate to the medium.
Latent disclosure: covered providers must include a hidden, persistent disclosure in AI-generated image, video or audio content, conveying provenance information — such as that it was AI-generated and details identifying the system — that can be detected by the provider's tool.
Detection tool: covered providers must make available a free, publicly accessible tool that lets anyone check whether a piece of image, video or audio content was created or altered by the provider's system, reading the latent disclosure.

A crucial scope limit is that these watermarking and disclosure obligations apply to image, video and audio content — text-only outputs are not covered by the requirements. The three mechanisms fit together: the latent disclosure embeds durable provenance in the content, the detection tool reads it, and the manifest disclosure gives a visible signal a person can see directly. For advertising, the latent disclosure and detection tool are the most consequential, because they mean AI creative carries verifiable provenance that persists even when no visible label is present. For how such provenance is technically implemented, see the AI content-detection guide.

What It Means for AI Ad Creative

For advertisers, SB 942 is best understood as reshaping the raw material of AI creative rather than imposing a new filing duty. Because the covered providers of large AI systems must embed latent provenance and offer manifest labels, the AI images, video and audio that advertisers generate with those systems increasingly arrive with machine-readable signals of their synthetic origin built in.

The Practical Effects

Provenance travels with the creative: AI ad assets made with covered tools can carry latent disclosure that detection tools and, increasingly, platforms can read — so synthetic origin is discoverable.
Manifest labels are available: where a visible AI label is appropriate or required, the covered provider's manifest-disclosure option supports adding one.
Alignment with platform rules: platform AI-labelling policies and this provenance layer reinforce each other, so keeping AI creative properly labelled satisfies multiple regimes at once.
Hosting-platform duties from 2027: AB 853 adds obligations for large hosting platforms starting January 1, 2027, extending the ecosystem's transparency responsibilities.

The prudent posture is to treat AI provenance as a feature of the creative rather than something to strip out: knowing which of your AI tools are covered providers, expecting latent provenance in the assets they produce, preserving manifest labels where appropriate, and aligning with platform AI-disclosure rules keeps advertising consistent across the overlapping regimes. Attempting to remove provenance to hide AI origin runs directly against the direction of the law and platform policy. Track the timeline on the Policy Change Tracker, and screen AI creative captions with the Keyword Risk Checker.

How It Fits the Wider Disclosure Landscape

SB 942 does not operate in isolation. It is one of several converging regimes requiring that AI-generated and synthetic content be identifiable, and advertisers benefit from seeing how they line up rather than treating each as a separate surprise. The common thread is that synthetic media should be detectable and, where it could mislead, disclosed.

The Converging Regimes

Regime	Focus	Who it primarily binds
California SB 942	Provenance, detection and labelling of AI media	Large generative-AI providers (and 2027 hosting platforms)
New York synthetic-performer law	Disclosure of AI synthetic performers in ads	Advertisers producing the ad
EU AI Act transparency	Labelling AI-generated and deepfake content	AI providers and deployers, including advertisers
Platform AI-label policies	On-platform labelling of AI content	Advertisers and creators

Read together, these regimes point advertisers toward a single disciplined practice: assume AI-generated creative should be identifiable as such, preserve the provenance and labels the tools provide, and add clear disclosure where a synthetic element could mislead — which satisfies the spirit of all of them. The differences are in who is bound and exactly what is required, but the direction is uniform. For the New York advertiser-facing duty see the New York synthetic-performer guide, and for the platform-label mechanics on a visual network the Pinterest AI-label guide.

AI Transparency Readiness Checklist

[ ] Identified which AI image, video and audio tools you use, and whether they are covered providers
[ ] Expected latent provenance disclosure in AI creative from covered tools
[ ] Used the manifest-disclosure option to add a visible AI label where appropriate
[ ] Preserved rather than stripped provenance metadata in AI ad assets
[ ] Aligned AI creative with platform AI-labelling policies
[ ] Noted that text-only AI outputs are outside SB 942's watermarking duties
[ ] Distinguished the covered-provider duty from the advertiser's role
[ ] Noted the August 2, 2026 operative date and the January 1, 2027 hosting-platform duties
[ ] Cross-checked obligations under other synthetic-media regimes you are subject to
[ ] Confirmed current SB 942 and AB 853 requirements against official California sources

Frequently Asked Questions

What does SB 942 require, and when does it take effect?

California's AI Transparency Act, SB 942, requires large generative-AI providers to make AI-generated content identifiable through three mechanisms — a free public detection tool, an optional visible ('manifest') disclosure, and an embedded, persistent ('latent') provenance disclosure in AI-generated images, video and audio — and after amendment by AB 853 it becomes operative on August 2, 2026, with additional obligations for large hosting platforms following on January 1, 2027. The law was originally adopted in 2024 with an earlier effective date, and AB 853, signed in 2025, moved the operative date and layered on the hosting-platform duties, which is why the current timeline centres on August 2026 and January 2027. Breaking down the substance: the detection tool must be free and publicly available, letting anyone submit a piece of image, video or audio content and learn whether it was created or altered by the provider's AI system. The manifest disclosure is a visible label: covered providers must give users the option to include a clear, conspicuous disclosure in AI-generated content that identifies it as AI-generated in terms understandable to a reasonable person and appropriate to the medium. The latent disclosure is hidden but durable: covered providers must embed provenance information in AI-generated image, video or audio content — conveying that it is AI-generated and identifying details about the system — in a way that persists in the content and can be read by the detection tool. Together these let both machines (via the latent disclosure and detection tool) and people (via the manifest disclosure) determine that content is synthetic. An important scope limitation is that the watermarking and disclosure obligations apply to image, video and audio content, not to text-only outputs, so AI-generated text is outside these particular requirements. For advertisers, the timing matters because so much creative is AI-assisted: from August 2, 2026, the large AI systems commonly used to generate ad imagery, video and audio will be operating under these obligations, so the assets they produce will increasingly carry latent provenance and support manifest labelling. The 2027 hosting-platform layer then extends transparency duties further into the distribution side. Because the dates and details were changed by amendment and could be refined further, advertisers should confirm the current requirements against official California sources rather than relying on the original 2024 text. For the EU counterpart see the EU AI Act Article 50 guide, and track the timeline on the Policy Change Tracker. The organizing principle is that SB 942 requires large AI providers to offer detection, manifest and latent disclosures for AI media, operative August 2, 2026, with hosting-platform duties from January 1, 2027.

Who is a 'covered provider' and does that include advertisers?

A 'covered provider' under SB 942 is a person that creates, codes or otherwise produces a generative-AI system that is publicly accessible within California and has over one million monthly visitors or users — in other words, the maker of a large, widely-used AI system — and this generally does not include advertisers, who are users of such systems rather than their creators, so the Act's core obligations fall on the AI tool providers, not on the businesses that use those tools to make ads. Understanding this allocation is essential to knowing where the compliance burden actually sits. The definition has three elements. The provider must create the system — build, code or produce the generative-AI model or tool — which targets the entity responsible for the AI itself. The system must be publicly accessible within California, which reaches consumer-facing AI systems available to Californians. And it must exceed one million monthly visitors or users, a scale threshold that focuses the law on large, mainstream systems rather than every niche tool. An entity meeting all three is a covered provider bearing the detection-tool, manifest-disclosure and latent-disclosure duties. Advertisers typically do not meet this definition, because they use AI systems to generate creative rather than creating the systems. If a marketing team uses a large, widely available AI image or video tool to produce an ad, the provider of that tool is the covered provider, and the SB 942 features — the embedded provenance, the manifest-label option, the detection tool — exist at the tool level. The advertiser is a downstream user who benefits from and interacts with those features, not the regulated party carrying the statutory duties. There are two important caveats. First, an organisation that itself creates a qualifying generative-AI system — for instance a large company that builds and publicly offers its own AI model at scale — could be a covered provider in its own right, so the analysis depends on what the business actually does, not on its industry label. Second, AB 853 adds obligations for large hosting platforms from January 1, 2027, which broadens the set of regulated actors beyond the original covered-provider category and could touch platforms that host and distribute AI content. For advertisers, the practical takeaway is to identify which of the AI tools they rely on are covered providers, understand that those tools carry the disclosure machinery, and use it appropriately — rather than assuming they themselves must build detection tools or embed watermarks. Define the underlying terms in the compliance glossary. The organizing principle is that covered providers are the makers of large public generative-AI systems, so the Act binds AI tool providers rather than the advertisers who use them, with hosting platforms added from 2027.

What is the difference between manifest and latent disclosure?

The difference between manifest and latent disclosure under SB 942 is visibility and audience: a manifest disclosure is a clear, visible label that a person can see directly and understand as identifying content as AI-generated, while a latent disclosure is a hidden but persistent piece of provenance information embedded in the content itself, designed to be read by a detection tool rather than seen by a casual viewer — and covered providers must support both, because they serve different transparency purposes. Manifest disclosure is the human-facing mechanism. Covered providers must give users the option to include, in AI-generated image, video or audio content, a disclosure that identifies the content as AI-generated in a manner that is clear, conspicuous, appropriate for the medium, and understandable to a reasonable person. The point is that an ordinary person encountering the content can tell, from something they can perceive, that it was made with AI. Because it is a visible label, it communicates directly to the audience without any tool or technical step. Latent disclosure is the machine-facing mechanism. Covered providers must embed, in AI-generated image, video or audio content, a hidden disclosure conveying provenance information — such as the fact that the content is AI-generated and identifying details about the system that produced it — in a way that persists in the content and can be detected by the provider's tool. Because it is embedded rather than displayed, a viewer does not normally see it, but it travels with the content and can be surfaced by the detection tool even if no visible label is present. This persistence is what makes provenance verifiable after the fact: a piece of AI media can be checked for its latent disclosure to confirm its origin. The detection tool is the third piece that connects to the latent disclosure: covered providers must offer a free, public tool that reads the latent disclosure and tells a user whether a piece of content was created or altered by the provider's system. So the latent disclosure and detection tool form a pair — one embeds provenance, the other reads it — while the manifest disclosure stands alone as the visible signal. For advertising, the latent disclosure and detection tool are particularly significant, because they mean AI creative carries verifiable, durable provenance regardless of whether a visible label is attached, which supports downstream verification by platforms and others. All of these obligations apply to image, video and audio, not to text-only outputs. For the technical underpinnings of embedded provenance see the AI content-detection guide. The organizing principle is that manifest disclosure is a visible label for people while latent disclosure is embedded, persistent provenance for machines, and covered providers must support both plus a detection tool.

How does SB 942 affect advertisers who create AI ad creative?

SB 942 affects advertisers who create AI ad creative mainly by changing the nature of the raw material rather than by imposing a direct filing or labelling duty on the advertiser: because the covered providers of the large AI systems advertisers use must embed latent provenance and offer manifest labels, the AI images, video and audio those systems produce increasingly arrive with machine-readable signals of their synthetic origin built in, and advertisers should work with that provenance rather than against it. The most concrete effect is that provenance travels with the creative. When an advertiser generates an ad asset using a covered AI tool, that asset can carry a latent disclosure identifying it as AI-generated, and detection tools — and, increasingly, platforms integrating similar provenance reading — can surface that origin. In practice this means the synthetic nature of AI ad creative becomes discoverable by third parties, which is precisely the transparency outcome the law intends. A second effect is the availability of manifest labels: where a visible AI label is appropriate or required by another regime or a platform policy, the covered provider's manifest-disclosure option gives the advertiser a supported way to add one, rather than improvising. A third effect is alignment across regimes: platform AI-labelling policies, other jurisdictions' synthetic-media rules, and SB 942's provenance layer all push in the same direction, so an advertiser that keeps AI creative properly labelled and preserves its provenance tends to satisfy several regimes at once, reducing rather than multiplying the compliance burden. A fourth, forward-looking effect comes from AB 853's hosting-platform obligations starting January 1, 2027, which extend transparency responsibilities to large platforms that host content — part of a broader trend toward provenance being maintained across the whole content lifecycle. The practical guidance that follows is to treat AI provenance as a feature to preserve, not an inconvenience to remove. Advertisers should know which of their AI tools are covered providers, expect latent provenance in the assets those tools generate, keep manifest labels where appropriate, and align with platform disclosure rules. Critically, they should not attempt to strip provenance metadata to conceal AI origin, because doing so runs directly counter to the direction of the law and platform policy and could create exposure under the disclosure regimes that require synthetic content to be identifiable. The overarching point is that SB 942 makes AI creative carry its own transparency, and the advertiser's job is to preserve and build on that rather than to defeat it. Screen AI creative captions and claims with the Keyword Risk Checker, and track the rollout on the Policy Change Tracker. The organizing principle is that SB 942 gives AI ad creative built-in provenance and labelling from the tool side, so advertisers should preserve and align with it rather than treat labelling as a burden to avoid.

How does SB 942 fit with the EU AI Act, the New York law and platform AI-label rules?

SB 942 is one of several converging regimes that all require AI-generated and synthetic content to be identifiable, and while they differ in who they bind and exactly what they demand, they point advertisers toward a single consistent practice: assume AI creative should be identifiable as such, preserve the provenance and labels the tools provide, and add clear disclosure wherever a synthetic element could mislead. Seeing them as a set rather than as isolated rules makes compliance far more manageable. Consider how they line up. California's SB 942 primarily binds large generative-AI providers, requiring them to embed provenance, offer manifest labels and provide detection tools for AI images, video and audio, with hosting-platform duties added from 2027. The EU AI Act's transparency provisions bind AI providers and deployers — including advertisers who deploy AI — requiring that AI-generated or manipulated content, including deepfakes, be marked and disclosed, with obligations applying from August 2026. New York's synthetic-performer law binds advertisers directly, requiring conspicuous disclosure when an advertisement includes an AI-generated synthetic performer. And platform AI-labelling policies, on networks such as the major social platforms, require advertisers and creators to label AI content on-platform, sometimes reinforced by automatic detection of provenance metadata. The actors differ — AI providers, deployers, advertisers, platforms — and the precise triggers differ — provenance embedding, deepfake labelling, synthetic-performer disclosure, on-platform labels — but the underlying expectation is uniform: synthetic media should be detectable and, where it could mislead, disclosed. That convergence is what allows a single disciplined approach to satisfy the spirit of all of them. An advertiser who preserves the provenance that covered AI tools embed (SB 942), labels AI and deepfake content it deploys (EU AI Act), discloses AI synthetic performers in its ads (New York), and applies platform AI labels (platform policies) is meeting the common standard across regimes, even though each has its own mechanics. Conversely, an advertiser who strips provenance or hides AI origin risks running afoul of several regimes at once. The practical method is therefore to build a standing habit — identify AI elements in creative, keep their provenance and labels, and disclose clearly where a viewer could be misled — and then check any regime-specific extras, such as New York's conspicuous-disclosure standard or a particular platform's labelling mechanics. For the New York advertiser duty see the New York synthetic-performer guide, and for platform-label mechanics the Pinterest AI-label guide. The organizing principle is that SB 942, the EU AI Act, the New York law and platform policies converge on making synthetic media identifiable, so a single practice of preserving provenance and disclosing AI elements satisfies them together.

Don't miss the next policy change.

Create a free account — track every policy change across 8 platforms, get instant alerts, and access every free compliance tool. Or try our Meta Rejection Predictor first.

Create Free Account

Report Keywords — Run AI Compliance Audit

#AI Disclosure#California#SB 942#Synthetic Content#Ad Compliance#Regulation#Watermarking#Advertisers#2026 Policy#Generative AI#Content Provenance#Compliance Guide 2026

Share This Report

Tweet Share

Platforms

Industries

Knowledge

Blog

California's AI Transparency Act (SB 942) in 2026: Manifest and Latent Disclosure, Detection Tools and the August Start Date

What SB 942 Requires and When

Who Is a Covered Provider

The Covered-Provider Test

Manifest, Latent and Detection Disclosures

The Three Mechanisms

What It Means for AI Ad Creative

The Practical Effects

How It Fits the Wider Disclosure Landscape

The Converging Regimes

AI Transparency Readiness Checklist

Frequently Asked Questions

Don't miss the next policy change.

Report Keywords — Run AI Compliance Audit

Share This Report

Related Posts

The FTC Consumer Review Rule in 2026: Fake Reviews, AI Testimonials and Fake Social Proof

FTC 'Made in USA' Enforcement in 2026: The July Warning Letters and Origin Claims on Social Media

UK DMCC Act in 2026: The Fake Reviews Ban, Drip Pricing Rules and CMA Direct Enforcement

Related Resources

All Tools

Platform Guides

Knowledge Hub

Policy Tracker