EU AI Act Article 50: What fashion brands must disclose from 2 August 2026

From 2 August 2026, fashion brands operating in the European Union must disclose when they use AI-generated or manipulated photorealistic images of people in their marketing materials. The requirement, set out in Article 50 of the EU Artificial Intelligence Act, obliges deployers of AI systems that generate or manipulate image, audio or video content to label such material in a way that reveals its synthetic nature.

What does Article 50 of the EU AI Act require?

Article 50 mandates that deployers of general-purpose AI models which generate synthetic content must disclose when that content resembles real people, objects, places or events and could be mistaken for authentic material. The provision targets deepfakes and other photorealistic manipulations that could mislead the public. For fashion brands, this means any use of AI-generated models, digitally altered body shapes or synthetic faces in product imagery, lookbooks or advertising campaigns must be clearly marked as artificial.

The regulation includes specific exceptions. Lawfully created content for artistic, satirical or journalistic purposes is exempt, as are AI systems used for fraud detection or cybersecurity. The disclosure requirement applies only to content that could reasonably appear authentic to a person, not to obviously stylised or abstract imagery.

When did Article 50 come into force?

The EU AI Act entered into force on 1 August 2024. Article 50, which addresses transparency obligations for general-purpose AI models, became applicable on 2 August 2026, two years after the Act's entry into force. This staggered implementation gave brands and technology providers time to adapt their workflows and labelling systems.

How should fashion brands label AI-generated model imagery?

The AI Act does not prescribe a single disclosure format, but it requires that labels be clear, conspicuous and machine-readable. Fashion brands have adopted various approaches to meet the standard:

• A visible text overlay on the image itself, such as 'AI-generated model' or 'This image has been digitally created'.
• A watermark or icon in the corner of the image, paired with a legend or caption explaining the use of synthetic content.
• Metadata embedded in the image file using standards such as the Coalition for Content Provenance and Authenticity (C2PA) framework, which records whether an image was generated or manipulated by AI.
• A disclaimer in the caption or adjacent text, for example: 'The model in this image is not a real person and was created using artificial intelligence'.

Best practice combines human-readable labels with technical metadata to ensure transparency for both consumers and automated systems that may scan content for compliance.

What are the penalties for non-compliance?

Breaches of Article 50 fall under the AI Act's tiered penalty structure. Non-compliance with transparency obligations can result in fines of up to 1.5 per cent of the infringing company's total worldwide annual turnover for the preceding financial year, or 7.5 million euros, whichever is higher. The European Commission and national authorities have discretion to impose lower fines based on the severity and duration of the violation, the size of the company and whether the breach was intentional or negligent.

Which fashion use cases require disclosure?

Article 50 applies to any photorealistic depiction of a person that has been generated or manipulated by AI. Common scenarios in fashion include:

• Fully synthetic models created by generative AI to display clothing, accessories or beauty products.
• Real models whose body proportions, skin tone or facial features have been digitally altered beyond standard retouching.
• AI-generated faces or avatars used in virtual try-on tools or personalised marketing.
• Composite images that blend AI-generated elements with real photography, such as replacing a model's head or torso.

Standard retouching practices, such as colour correction, blemish removal or background adjustments, do not trigger the disclosure requirement unless they substantively alter the appearance of a person in a way that could mislead viewers about their true likeness.

How does Article 50 affect virtual influencers and avatars?

Virtual influencers and branded avatars present a grey area. If the avatar is clearly non-human or stylised,such as a cartoon character or fantastical figure,it typically falls outside the scope of Article 50 because it does not purport to depict a real person. However, if the avatar is photorealistic and could plausibly be mistaken for a real human being, disclosure is required. Brands working with hyper-realistic virtual influencers are advised to label such content proactively to avoid regulatory risk.

The obligation to disclose rests with the deployer of the AI system, which in most cases is the fashion brand or retailer commissioning or publishing the content, rather than the technology provider or creative agency that produced it. Brands should ensure that contracts with agencies and vendors clarify responsibility for compliance and that labelling protocols are agreed in advance.