EU AI Act Labelling Rules for AI Ad Imagery: What Brands Must Disclose
Since 2 August 2026, the EU AI Act sets two labelling duties for AI-generated campaign imagery. The company providing the generation system must mark its outputs in a machine-readable format so the images are detectable as artificial. Separately, whoever deploys the system to create a deep fake must disclose that the content has been generated or manipulated. The Act defines a deep fake as AI-generated or manipulated image, audio or video content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic. A photorealistic campaign visual with a lifelike model can meet that definition, so brands and agencies publishing such imagery in the EU should plan for a disclosure. The information must reach viewers in a clear and distinguishable manner at the latest at first exposure. Fines for non-compliance reach EUR 15 million or 3 percent of worldwide annual turnover, whichever is higher.
That is the short version. Below: where the duties come from, who carries which one in a brand-agency-studio setup, and how the US position differs. An overview for campaign decision-makers, not legal advice.
Where the rules come from
The duties sit in Article 50 of Regulation (EU) 2024/1689, the AI Act, under the heading "Transparency obligations for providers and deployers of certain AI systems." They have applied since 2 August 2026, the Act's general application date, as the European Commission's AI Act page confirms.
You may have read that the AI Act was delayed. That is true only for a different part of the law: the amending regulation adopted in summer 2026, often called the AI Omnibus, moved the compliance dates for high-risk AI systems to December 2027 and August 2028. It did not touch the transparency rules, and a campaign image generator is not a high-risk system. Article 50 applies now.
When a campaign image counts as a deep fake
Article 3(60) of the Act defines a deep fake as "AI-generated or manipulated image, audio or video content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful."
Two elements matter for advertising: the content must resemble something that exists, and a viewer must be able to mistake it for a real recording. Read against campaign work:
A generated image of a real, recognisable model resembles an existing person. If it looks like a photograph, it sits inside the definition. The consent side of that scenario is covered in Model Likeness Rights in AI Campaigns.
A photorealistic image of a real product in a generated setting resembles existing objects and places. The closer the output is to camera footage, the harder it is to argue the definition does not apply.
Stylised or fantastical imagery is a weaker case. An illustration or a scene no viewer would take for a photograph does not "falsely appear authentic." The line is the viewer's impression, not the production method.
The Commission published draft guidelines on Article 50 in May 2026. Until final guidance or case law draws the line, the safe working assumption for photorealistic campaign output is that the definition applies.
The provider duty: machine-readable marking
Article 50(2) obliges providers of AI systems that generate synthetic audio, image, video or text to ensure the outputs "are marked in a machine-readable format and detectable as artificially generated or manipulated."
This duty belongs to whoever places the generation system on the market, not to the brand running the campaign. For a marketing team it still matters: images from a compliant pipeline carry embedded provenance data, and the marking question belongs in the vetting conversation with any production partner. Ask which systems the studio uses, how outputs are marked, and whether the marking survives retouching and export.
The Commission has also facilitated a voluntary Code of Practice on marking and labelling of AI-generated content, published in 2026. It helps providers and deployers implement the obligations; the duties themselves are mandatory either way.
The deployer duty: telling the audience
Article 50(4) is the paragraph that reaches brands and agencies directly. Deployers of an AI system that generates or manipulates content constituting a deep fake "shall disclose that the content has been artificially generated or manipulated." A deployer is whoever uses an AI system under its own authority in a professional context.
In a campaign built by a production studio for a brand, often through an agency, the first question is who the deployer is. The studio operates the generation system; the brand publishes the result. The contract should name the party responsible for the disclosure and fix its wording and placement, as it does for usage rights and approvals. How AI Production Studios Work With Ad Agencies describes where those decisions sit in the collaboration.
Article 50(5) sets the form: the information must reach viewers "in a clear and distinguishable manner at the latest at the time of the first interaction or exposure" and meet accessibility requirements. A note buried on a landing page three clicks from the ad fails that standard.
There is a softer regime for artistic work. For content that forms part of an "evidently artistic, creative, satirical, fictional or analogous work," disclosure is limited to a form that does not hamper the display or enjoyment of the work. Two cautions before leaning on that clause for advertising: the exception changes the form of the disclosure, not its existence, and no court has yet decided whether a commercial campaign counts as an evidently artistic work. Building a compliance position on it is a gamble a brand does not need to take.
What non-compliance costs
Article 99(4)(g) of the Act puts breaches of the Article 50 obligations, by providers and deployers alike, in the fine bracket of up to EUR 15 million or, for an undertaking, up to 3 percent of total worldwide annual turnover, whichever is higher; for SMEs the lower amount applies. Beyond the fine, an unlabelled campaign that draws a regulator's attention creates the kind of coverage a premium brand runs campaigns to avoid.
The US comparison: no label, but no safe harbour either
US federal law has no general requirement to label AI-generated advertising imagery. The Federal Trade Commission polices advertising through its deception authority: an ad is a problem when it misleads consumers about something material, whatever technology produced it. The FTC's Trade Regulation Rule on the Use of Consumer Reviews and Testimonials, in effect since October 2024, bans fake reviews and testimonials, including AI-generated ones that misrepresent the reviewer or their experience.
The difference is the trigger: the EU duty attaches to the nature of the content, the US rules to whether someone is misled about a material fact. A transatlantic campaign needs the EU disclosure anyway, and one global labelling standard beats maintaining two versions of the same asset.
What campaign teams should do now
Inventory running assets. Identify which published visuals are AI-generated or AI-manipulated and whether they meet the deep fake definition.
Allocate the duty in contracts. Every production agreement should say who discloses, where and in what wording. Silence leaves both parties exposed.
Ask providers about marking. Confirm that the systems in your pipeline mark outputs machine-readably and that post-production preserves the marking.
Set a disclosure format per channel. A label that is clear at first exposure looks different in a social feed than on out-of-home. Decide formats once, centrally.
Watch the Commission's guidance. The draft Article 50 guidelines and the marking Code of Practice will shape what regulators accept. The wider duties around rights and disclosure: Legal Framework for AI Campaigns.
Frequently Asked Questions
Do AI-generated ad images have to be labelled in the EU?
Yes, in many cases. Since 2 August 2026, Article 50(4) of the AI Act requires deployers to disclose content that constitutes a deep fake: AI imagery that resembles real persons, objects, places or events and could be mistaken for authentic. Photorealistic campaign imagery can meet that definition; stylised imagery is a weaker case.
Who is responsible for the label, the brand, the agency or the studio?
The Act puts the disclosure duty on the deployer of the AI system, the party using it under its own authority. In a multi-party production the roles overlap, so the contract should assign who discloses, in what wording and where.
Does the AI Act delay of 2026 postpone the labelling rules?
No. The amending regulation moved the dates for high-risk AI systems to December 2027 and August 2028. The Article 50 transparency obligations, including deep fake disclosure and machine-readable marking, were not postponed and apply since 2 August 2026.
Is a fashion campaign an "artistic work" that is exempt from disclosure?
No exemption removes the disclosure. For evidently artistic or creative works, Article 50(4) only softens the form: the generated content must still be disclosed, in a way that does not hamper enjoyment of the work. Whether a commercial campaign qualifies as evidently artistic is unsettled.
What fines apply if AI ad imagery is not labelled?
Up to EUR 15 million or 3 percent of total worldwide annual turnover, whichever is higher, under Article 99(4)(g) of the AI Act. For SMEs the lower amount applies.
Do US rules require labelling AI-generated ad images?
No general federal labelling duty exists. The FTC applies its deception standard: an ad is unlawful when it materially misleads, however it was produced. Its 2024 rule on consumer reviews bans fake testimonials, including AI-generated ones. Brands with EU campaigns tend to apply the EU disclosure globally rather than maintain two asset versions.
About The Bridges AI
The Bridges AI is an independent AI and hybrid campaign production company working in strategic partnership with Most Wanted Models®, the German model agency founded in 1997. Combining nearly 30 years of fashion, talent and campaign expertise with generative AI, The Bridges AI produces high-end imagery and video for fashion, beauty and premium brands worldwide.
Written by the editorial team of The Bridges AI. Last updated October 2026.
Comments