Two AI Ad-Disclosure Laws Took Effect August 2. Here's What Actually Applies to You.

August 2, 2026 was billed by a wave of compliance blogs this week as the day "AI ad disclosure becomes mandatory." That's not quite what happened. Two real laws did take effect on that date — California's AI Transparency Act (SB 942, as amended by AB 853) and Article 50 of the EU AI Act — and a third, New York's synthetic-performer law, has already been in force since June 9. But as Total Licensing reported citing Billo CEO Donatas Smailys, all three laws target narrow, specific actors and scenarios — "most ad campaigns fall under neither." If you're running AI UGC ads for a DTC brand, the honest answer to "do I have to label this now?" is almost certainly no law says so — but the direction of travel says disclose anyway.

What actually changed on August 2

California's AI Transparency Act (SB 942) doesn't regulate advertisers at all. It binds providers of publicly accessible generative AI systems with more than 1 million monthly users or visitors — think the model vendors, not the marketers using them. Covered providers must offer a free public detection tool, an optional visible disclosure, and a mandatory hidden watermark embedded in AI-generated image, video, and audio output. Violations carry a $5,000 civil penalty per violation, with each day of continued non-compliance counted separately — confirmed across multiple legal trackers including TrustArc and CalMatters' bill tracker. If you're a brand using Prizmad, Creatify, HeyGen, or any AI video tool, this law's obligations sit with the tool vendor, not with you.

Article 50 of the EU AI Act is the one that does reach brands directly — but only in a specific case. It requires that AI-generated or AI-manipulated content be made clear to users when that content depicts a real or plausibly-real person, object, place, entity, or event in a way that could pass as authentic — the deepfake scenario. It applies to any advertiser whose ads reach EU consumers regardless of where the company is headquartered, and penalties can reach €15 million or 3% of global annual turnover, per the Act's Article 99 penalty schedule. A standard AI UGC avatar ad — a consistent, disclosed-as-synthetic avatar persona delivering a product testimonial — doesn't obviously trigger this; the rule is aimed at content designed to be mistaken for something real, not stylized or clearly synthetic creative. The EU is still finalizing exactly what the consumer-facing label should look like (a proposed "AI" icon is circulating in draft Code of Practice guidance), so treat that visual as unsettled.

New York's synthetic-performer law, live since June 9, covers the opposite case from Article 50: an invented performer who resembles no real person, appearing in an ad reaching New York audiences. First violation is $1,000, each subsequent one $5,000. This is the law most likely to actually touch AI UGC avatar ads, since an AI avatar is by definition a synthetic performer — but it only applies "where the advertiser has actual knowledge," per the law's language, and carves out several categories.

Why the confusion is understandable

Three laws, three different triggers — provider-side watermarking, real-person deepfake depiction, and invented-performer disclosure — landing in the same six-week window reads like one big mandate if you only skim headlines. It isn't. As the Total Licensing piece puts it: "Neither rule creates a general requirement for every U.S. brand to tell viewers that an advertisement was made with AI." A companion piece from Adriel aimed at agencies makes the same point from the compliance-audit side: the real risk for most ad accounts isn't a fine, it's a client finding out AI touched their creative from a platform rejection notice or a Reddit thread instead of from the agency itself.

What's actually mandatory right now vs. what's platform policy

Separate from all three laws, Meta and Google already label AI-assisted ad creative on their own platform policy, not legal requirement. Meta auto-detects its own generative tools (Background Generation, Image Generation, Add Animation) and applies an "AI Info" label with no advertiser action needed; it also reads C2PA provenance metadata from third-party tools where present. Google's "How this ad was made" panel, rolled out through My Ad Center, works similarly — automatic for Google's own tools, advertiser-declared for everything else, and Google has said it won't verify whether that self-declaration is accurate. TikTok remains the strictest, requiring advertisers to manually enable an AI Disclosure tag before submission. We covered the platform-by-platform mechanics in more depth in our AI ad disclosure labels guide.

The practical takeaway for DTC advertisers running AI UGC ads

1. You're very likely not directly covered by SB 942 — that's a provider obligation, and Prizmad and comparable tools carry it, not you.
2. Article 50 matters only if your creative depicts a real, identifiable person or event in a way meant to pass as authentic. A disclosed AI avatar delivering a scripted testimonial is a different category from a deepfake of an actual person.
3. If your spend touches New York, the synthetic-performer disclosure is the one to actually check — it's the narrowest-scoped of the three but the most likely to technically apply to an AI avatar ad.
4. Platform labels (Meta, Google, TikTok) are the more immediate operational reality for most advertisers than any of these three laws, since they apply regardless of jurisdiction the moment you run the ad.
5. Consumer sentiment is running ahead of the law anyway — a Q2 2026 Fractl/Search Engine Land survey cited in the Total Licensing piece found 91% of consumers want AI-generated video labeled, even though only 20% of organizations say they always disclose. That gap, not the legal minimum, is the actual reputational exposure.

None of this changes whether AI-generated ads are allowed to run, or whether they perform. It's a transparency question, not a content restriction — and for most AI UGC advertisers, the honest move remains what it was before August 2: disclose by default rather than parse which of three narrow laws technically applies to your specific campaign.

Building AI UGC ads and want the compliance question to be a five-minute checklist, not a legal review? Prizmad generates the ad — script, avatar, voiceover, captions, music — from a product URL in about 5 minutes; see the pricing or paste a URL into the generator to see the workflow, and check our disclosure labels guide for the platform-by-platform checklist.

Sources: Total Licensing, July 30, 2026; Adriel, updated July 23, 2026; TrustArc SB 942 tracker; EU AI Act Article 99.