EU AI Act Article 50: the four transparency duties in plain words

EU AI Act Article 50 transparency duties apply from 2 August 2026. Here is who each of the four obligations binds and what you must do.

· 5 min read

What EU AI Act Article 50 says

Article 50 of the EU AI Act introduces transparency obligations for providers and deployers of certain AI systems. It applies from 2 August 2026 and sets four clear duties: disclose AI interactions, mark synthetic content, inform people exposed to emotion recognition or biometric categorisation, and label deepfakes and certain AI-generated public-interest text. Each duty is aimed at ensuring users know when they are engaging with AI or AI-generated content.

Who is a provider and who is a deployer under Article 50

In the EU AI Act, a provider is the person or organisation that develops an AI system or has it developed and places it on the market or puts it into service under their own name or trademark. A deployer is the person or organisation that uses an AI system under their own authority, except where the AI system is used in the course of a personal non-professional activity.

The four transparency duties in Article 50 and who must comply

The first duty in Article 50(1) requires providers to inform people when they are interacting with an AI system. This applies to chatbots, virtual assistants, or any system that engages in direct interaction with users. The disclosure must be clear and made at the latest at the first interaction.

The second duty in Article 50(2) requires providers of generative AI systems to mark synthetic audio, image, video and text output in a machine-readable format so it can be detected as artificially generated or manipulated. This marking must be effective, interoperable, robust and reliable as far as technically feasible. Generative AI systems placed on the market before 2 August 2026 have until 2 December 2026 to meet this duty under the Digital Omnibus on AI.

The third duty in Article 50(3) requires deployers of emotion recognition or biometric categorisation systems to inform the people exposed to such systems. Note that the business-to-business and industrial exemption applies only to the marking duty in Article 50(2), not to this one.

The fourth duty in Article 50(4) requires deployers to disclose deepfakes and to disclose AI-generated or manipulated text published to inform the public on matters of public interest. This duty ensures that users are aware when content has been artificially generated or altered, particularly in contexts where such information could influence public opinion or decision-making.

How to disclose: format, timing and exemptions

Article 50(5) sets out how the information must be given: it must be clear, distinguishable, provided at the latest at the first interaction or exposure, and accessible. The EU has also created a voluntary set of icons for labelling AI content, which can be used to indicate fully AI-generated content, partially AI-modified content, or general AI involvement.

There are narrow exemptions. The marking duty in Article 50(2) does not apply to outputs used in business-to-business or industrial contexts. Artistic, creative, satirical or fictional works only need disclosure in a way that does not hamper the display or enjoyment of the work. AI-generated text under human review with editorial responsibility is exempt from the public-interest text disclosure in Article 50(4).

Content generated and published before 2 August 2026 does not need to be labelled retroactively. This means that existing content does not have to be updated to comply with the new transparency obligations.

The Code of Practice and Commission Guidelines

The Code of Practice on Transparency of AI-Generated Content was finalised on 10 June 2026 and assessed as adequate by the Commission and the AI Board. Signing is voluntary, but signatories can rely on it to demonstrate compliance with Article 50(2), (4) and (5). The Code asks for a multi-layered approach to marking, combining metadata and invisible watermarks, and requires metadata to be digitally signed and time-stamped in a secure, tamper-evident way.

The Commission also published non-binding Guidelines on Article 50 on 20 July 2026. These Guidelines provide further clarification on how to interpret and apply the transparency obligations, but they do not introduce new legal requirements.

Practical steps to comply with Article 50

For providers of AI systems that interact with users, add a clear disclosure at the start of every interaction, such as a notice saying 'You are interacting with an AI system'. For generative AI, embed machine-readable metadata and use invisible watermarks to mark synthetic audio, image, video and text. Tools like C2PA Content Credentials can help bind provenance claims to files, and combining this with an invisible watermark addresses the risk of metadata stripping.

For deployers, ensure that any use of emotion recognition or biometric categorisation systems is disclosed to the people exposed. For deepfakes and AI-generated public-interest text, add a clear label or notice at the point of publication. Use the EU’s voluntary icons where appropriate to signal AI involvement.

For all duties, keep the disclosure clear, distinguishable, and accessible. Review the Code of Practice for best practices on marking and detection, and consider using a provenance certification service if you need to demonstrate compliance. For example, L'AItelier DNA certifies files and issues tamper-evident provenance certificates, binding C2PA manifests and invisible watermarks to the content.

Key dates for Article 50 and the Digital Omnibus on AI

The transparency obligations in Article 50 apply from 2 August 2026. The Digital Omnibus on AI, which came into force on 27 July 2026, extends the deadline for generative AI systems placed on the market before 2 August 2026 to meet the marking duty in Article 50(2) until 2 December 2026. It also introduces a prohibition on AI 'nudification' systems from 2 December 2026 and adjusts other high-risk deadlines.

Frequently asked questions

Does Article 50 apply to AI systems developed before 2 August 2026?

Yes, but with a transition period for generative AI. Systems placed on the market before 2 August 2026 have until 2 December 2026 to meet the marking duty in Article 50(2). Content generated and published before 2 August 2026 does not need to be labelled retroactively.

What is the penalty for not complying with Article 50?

Breaching the Article 50 obligations can lead to fines of up to EUR 15 million or 3% of worldwide annual turnover, whichever is higher. Lower ceilings apply to SMEs.

Are there exemptions to the marking duty in Article 50(2)?

Yes. There is a narrow exemption for outputs used in business-to-business or industrial contexts. Artistic, creative, satirical or fictional works only need disclosure in a way that does not hamper the display or enjoyment of the work.

How can I mark AI-generated content to comply with Article 50(2)?

Use a multi-layered approach combining machine-readable metadata (e.g., C2PA Content Credentials) and invisible watermarks. The Code of Practice recommends metadata that is digitally signed and time-stamped in a secure, tamper-evident way.

Do I need to label AI-generated content if it is under human review?

AI-generated text under human review with editorial responsibility is exempt from the public-interest text disclosure in Article 50(4). However, other duties may still apply depending on the context.

Put provenance on your next file

L'AItelier signs images and video with C2PA, adds an invisible watermark, and issues a certificate anyone can check. Your file is deleted as soon as the certificate is issued.

General information about the law as it stood when this was published, not legal advice for your situation.