Article 50

Transparency

In force since 2 August 2026

Article 50 became applicable on 2 August 2026. These are disclosure duties rather than engineering ones, which makes them both the cheapest obligations to meet and the most visible to fail: tell people when they are dealing with AI, and mark what AI produced.

What it requires

Systems interacting directly with people, such as a support chatbot, must tell them they are dealing with an AI system, unless that is obvious to a reasonably observant person.
Providers of systems generating synthetic audio, image, video or text must mark the output in a machine-readable format, detectable as artificially generated or manipulated.
Deployers of emotion recognition or biometric categorisation must inform the people exposed to it.
Deployers publishing deepfakes must disclose that the content is artificially generated or manipulated. Where the work is evidently artistic or satirical, the disclosure need not spoil it.
Deployers publishing AI-generated text to inform the public on matters of public interest must disclose it, unless a human reviewed the text and someone holds editorial responsibility for it.

Who it applies to

Providers and deployers both, with different duties falling on each. Most small companies are deployers.
It bites hardest on ordinary marketing and customer support: a chatbot on the website, blog posts written by a model, generated images of people who do not exist.
Unlike the high-risk regime, there is no classification exercise to do first. If you run a chatbot, the duty is already yours.

What changed in July 2026

The duties themselves were not rewritten, and the 2 August 2026 commencement stands.
A transition was added. Providers of systems generating synthetic content that were already on the market before 2 August 2026 have until 2 December 2026 to comply with the machine-readable marking duty in Article 50(2).
That transition sits in Article 111(4), which the Digital Omnibus inserted, and it covers the marking duty only. The disclosure duties were live from August.
Article 50(7), on codes of practice, was amended, which affects how compliance will be evidenced rather than what is owed.

What most guidance still gets wrong

Assuming a chatbot notice covers everything. Marking generated content is a separate obligation, falls on providers rather than deployers, and has its own date.
Assuming the deferral of the high-risk rules moved these too. It did not. The high-risk obligations went to December 2027; transparency stayed where it was.
Treating disclosure as a legal page. The duty is to inform the person interacting with the system, at the point they interact with it, not to bury a sentence in the terms.

Which of these apply to you?

Ten questions, about three minutes, no account and nothing stored on our side. It tells you which of these obligations reach your organisation and which dates you are already past.

Start the free assessment

Written against Regulation (EU) 2024/1689 as amended by Regulation (EU) 2026/1744. Not legal advice. Every change to this drafting is dated and published on the change history page.