Across the Atlantic, U.S. states have been making moves of their own. New York has introduced rules around synthetic performers and digital replicas. California has its own AI transparency requirements. And Hawaii has been considering legislation around everything from AI-generated likenesses to synthetic performers in advertising.
We should also say upfront that we're looking at all of this from an e-commerce perspective, not a legal one, so nothing here should be taken as legal advice. For that side of the conversation, we have a session with AI and media lawyer Kelsey Farish for a webinar on the EU AI Act and what e-commerce teams should know.
Put the two regions side by side, though, and one of the biggest differences becomes pretty obvious.
The EU is taking an across-the-board approach. One AI Act creates a broader regulatory framework across its Member States. The U.S., meanwhile, is building its AI rulebook much more piece by piece. There's still no federal equivalent to the EU AI Act, which means much of the movement we're seeing is happening at the state level.
For global brands, that distinction matters.
If you're selling across both the U.S. and Europe, there isn't just one set of AI rules you need to keep an eye on. The same AI-generated fashion model, campaign image, or piece of content could potentially fall under different requirements depending on where it appears and who sees it.
And even if your brand operates primarily in one market, looking at what's happening elsewhere gives us an interesting preview of where regulation could be heading. Different governments are tackling many of the same questions, such as transparency, synthetic people, digital replicas, and consumer deception, but they aren't always arriving at the same answers.
So, what happens when we actually put them side by side?
EU vs. U.S. AI Regulation, Side by Side
| AI requirement | EU | New York | California | Hawaii |
|---|---|---|---|---|
| Label/disclose realistic AI-generated or manipulated content | YES, for covered deepfakes — Article 50 of the EU AI Act requires disclosure when AI-generated/manipulated image, audio or video content constitutes a deepfake. | PARTIAL — New York S.8420/A.8887 specifically requires disclosure of synthetic performers in advertisements. | PARTIAL — California's AI Transparency Act focuses primarily on disclosure/provenance requirements for covered GenAI providers rather than a universal visible label on brand imagery. | NO comparable general labeling requirement in the final Act 247. Earlier HB 2137 versions proposed broader disclosure requirements, but the final enacted Act 247 was narrowed. |
| Disclose AI-generated / synthetic models in advertising | YES, potentially — A realistic synthetic person can fall within the AI Act's definition of a deepfake and Article 50 disclosure requirement. | YES — S.8420/A.8887 requires advertisements using a synthetic performer to disclose that use. | NO equivalent general advertising requirement identified. | NO in the final law. Interestingly, earlier HB 2137 versions did propose conspicuous disclosure of synthetic performers in advertising, but this did not survive into Act 247. |
| Machine-readable / embedded AI disclosure | YES — Article 50 requires providers of covered AI systems to make outputs machine-readable and detectable as artificially generated/manipulated. | NO equivalent general requirement identified. | YES — Covered GenAI providers must embed machine-readable information in AI-generated content identifying its AI origin. This is a provider requirement, not a general retailer/advertiser requirement. | NO equivalent requirement identified in Act 247. |
| Rules around digital replicas of real people | PARTIAL / DEPENDS — No harmonized EU-wide digital-replica right. GDPR may govern the personal data used to create and operate a Digital Twin, while consent and likeness rights vary by Member State. The AI Act separately imposes transparency requirements where the resulting content qualifies as a deepfake. | YES — Under the New York Fashion Workers Act, models must give separate, explicit written consent for model management companies and clients to use their digital replicas. | YES, in certain contexts. | YES — Hawaii Act 247 addresses unauthorized realistic digital imitations of identifiable individuals. |
| AI manipulation of products, objects or locations can fall within transparency rules | YES, potentially — The EU AI Act's definition of "deep fake" expressly includes existing persons, objects, places, entities and events. | NO equivalent broad AI-specific transparency rule identified. | NO equivalent broad consumer-facing rule identified. | NO equivalent broad rule. Act 247 focuses on realistic digital imitations of identifiable individuals, rather than objects/places generally (Hawaii Data Capitol). |
| Specific AI disclosure rules aimed at advertising | YES, but Article 50 isn't limited to advertising. Article 50 applies based on the type/use of AI content rather than simply whether something is an advertisement. | YES — New York's synthetic-performer law specifically applies to advertisements. | NOT IN THE SAME WAY — California's AI Transparency Act regulates covered GenAI providers and technical transparency rather than creating a general synthetic-performer advertising disclosure rule. | NOT IN FINAL ACT 247. Earlier versions proposed this exact type of requirement. |
| One overarching AI framework? | YES — The EU AI Act creates an EU-wide regulatory framework. | NO — state-specific laws. | NO — state-specific laws. | NO — state-specific laws. |
Where This Leaves Brands
We hope this gave you some interesting insight into where AI legislation and regulation are heading and maybe introduced you to a few laws and approaches you hadn't come across yet.
There's still a lot being figured out, and the rules are far from consistent from one place to the next. But what's interesting is how quickly the conversation has moved from whether AI should be regulated to how. Labels, embedded disclosures, digital replicas, consent — these are already becoming real considerations for the teams creating and using AI content.
And if the last year is anything to go by, this is probably only the beginning.