EU AI Act Article 50: what brands must label when AI content features real people

Article 50 of the EU AI Act has applied since 2 August 2026. For the first time there is an EU-wide rule that AI-generated content must be recognisable. For brands working with AI images of real people, such as the digital twin of a model or athlete, this is no longer theory. It shapes how campaigns look and what your contracts need to say.

Two roles, two duties

Article 50 splits responsibility between two parties.

• Providers of generative AI systems, the companies building the tools, must mark their output in a machine-readable way, such as metadata or an invisible watermark that lets software detect AI-generated content.
• Deployers, the parties using AI content, must visibly disclose when content is a deepfake.

The AI Act defines a deepfake as AI-generated image, audio or video that resembles existing people, places or events and could falsely appear authentic. If your brand runs a campaign with the digital twin of an existing model, you are the deployer. The disclosure duty sits with you, not with the tool or the agency that made the image.

Creative work: disclose, don't ruin it

The law makes room for marketing and creative work. For content that is evidently artistic, creative, satirical or fictional, disclosure can be limited to a notice that does not hamper the display of the work. A campaign visual doesn't need to be covered in warnings, but the label has to be there.

What the Digital Omnibus changed

This summer the EU amended the AI Act through the so-called Digital Omnibus, which entered into force on 27 July 2026. For Article 50 the main line held:

• The transparency obligations apply from 2 August 2026, including deepfake disclosure.
• Only for the technical marking by providers did generative systems already on the market before 2 August get extra time, until 2 December 2026.
• Content already online before 2 August does not need to be marked retroactively, although the European Commission recommends it where technically feasible.

The Commission also published a voluntary Code of Practice on marking and labelling AI-generated content on 10 June 2026 and guidelines on Article 50 on 20 July 2026. Signing the Code is not mandatory, but it is a recognised way to show you are doing it right.

What's at stake

Breaching Article 50 can lead to fines of up to €15 million or 3% of worldwide annual turnover, whichever is higher. Audience trust matters at least as much. A campaign later exposed as an unlabelled deepfake costs more than a fine.

Good practice already exists

H&M shows how it can work. Images featuring digital twins of its models carry a visible label in the bottom-right corner. Transparent about AI, without ruining the image.

Checklist for brands

1. Take stock. For each campaign, check whether recognisable real people appear in AI content. If so, label it.
2. Put it in writing. Agree in the contract who is responsible for disclosure: brand, agency or talent.
3. Ask your vendor. How does the tool mark its output: metadata, watermark or both?
4. Document. Keep a record per asset of which tool was used and what consent the person gave.
5. Label everywhere. On social, in ads and on product pages too, not only in the hero campaign.
6. Check older material. Reusing content from before 2 August in a new campaign? Treat it as new and label it.

How ADAPT handles this

AI disclosure is a fixed clause in our campaign contracts. The brand is responsible for the disclosure and every use is tied to the talent's recorded consent. Labelling isn't an afterthought, it is part of the deal. Read also why talent is usually better off licensing their likeness than selling it.

Working with AI content of real people as a brand? Here is how to do it with a digital twin of real talent.

This article is general information, not legal advice.