Guides · Article 50 · EU AI Act
What newsrooms and publishers should know about Article 50
Newsrooms and publishers now use AI to draft copy, generate images and audio, and assist editing at scale. Two parts of Article 50 of the EU AI Act are especially relevant: the content-marking duty in 50(2), and the limb of 50(4) aimed squarely at AI-generated text published to inform the public — a limb that carries its own editorial carve-out. This guide explains both from the regulation text. It is informational only and not legal advice.
AI-generated text published to inform the public — Article 50(4)
Alongside its better-known deepfake rule, Article 50(4) provides that deployers of an AI system that generates or manipulates text which is published with the purpose of informing the public on matters of public interest must disclose that the text has been artificially generated or manipulated (Article 50(4), Regulation (EU) 2024/1689). Note the conditions packed into that sentence: it addresses deployers, the text must be published, and the publication must have the purpose of informing the public on matters of public interest. Whether a given article meets those conditions is fact-specific.
The human editorial-review carve-out
The same limb carves the obligation back. As written, it does not apply where the use is authorised by law for certain criminal-offence purposes, or where the AI-generated content has undergone a process of human review or editorial control and a natural or legal person holds editorial responsibility for the publication of the content.
That second carve-out is the one newsrooms tend to lean on, and it turns on both elements together: a genuine process of human review or editorial control, and a person or entity holding editorial responsibility for what is published. Whether a particular workflow satisfies it is fact-specific — a lightly skimmed AI draft is not obviously the same as a piece run through your standard editorial process. If you intend to rely on it, document how review works and who holds editorial responsibility, and confirm the analysis with qualified counsel.
Generated images, audio, and video — Article 50(2)
Publishing is not only text. Where the AI system generates synthetic audio, image, or video, Article 50(2) can ask the provider to mark those outputs in a machine-readable format so they are detectable as artificially generated, subject to its carve-outs. Many publishers pair machine-readable marking with a visible, human-readable note for readers. Whether your outputs fall inside 50(2), and whether a carve-out applies, are fact-specific questions for counsel.
Common questions
Do AI-drafted articles reviewed by a human editor need a disclosure?
The regulation text says the text-publication obligation does not apply where the AI-generated content has undergone a process of human review or editorial control and a natural or legal person holds editorial responsibility for the publication. Whether your workflow meets that description — both elements — is fact-specific. Document it and confirm with qualified counsel.
Is a visible reader-facing note enough for AI-generated images?
Article 50(2) asks providers to mark synthetic outputs in a machine-readable format detectable as artificially generated; a visible note often accompanies that but is a different thing. Whether your images fall under 50(2), and what marking fits, is fact-specific — confirm with qualified counsel.
See what may apply to your business
Answer seven quick questions for an automated, informational indication of which Article 50 obligations appear likely to apply — free, and not legal advice.