An EU rule now makes some AI outputs say they are not human

Start here

Someone in the European Union talking to a chatbot now has a specific legal answer to how they would know this was not written or said by a person. The EU's AI Act sets that answer out in one article, and it took effect only weeks before this guide was prepared.

What the documents say

Regulation (EU) 2024/1689, the AI Act, was adopted 13 June 2024 and published in the Official Journal on 12 July 2024. Its Article 50 requires providers to ensure a person knows they are interacting with an AI system, unless that is ‘obvious from the point of view of a natural person who is reasonably well-informed’ — an exception that does not apply to law-enforcement systems handling public crime reports. Providers must mark synthetic audio, image, video or text output in a machine-readable format that is ‘effective, interoperable, robust and reliable as far as this is technically feasible,’ and deployers of systems that generate deepfakes must disclose that content ‘has been artificially generated or manipulated,’ with a lighter, non-disruptive disclosure allowed for clearly artistic, creative or satirical work. AI-generated text published to inform the public on matters of public interest needs the same disclosure, unless it went through human review with a named person or organisation holding editorial responsibility. Article 50 applies from 2 August 2026. The European Commission's own regulatory framework page, retrieved 16 September 2026, confirms this phased timeline and states the Commission ‘published supporting guidance documents in July 2026’ to help organisations comply.

Check this

If an AI system in the EU responds without disclosing it is AI, and that is not obvious from context, that is the gap Article 50 targets — check the interface for a stated disclosure, not your own guess about what feels artificial. For an article or image you suspect is AI-generated, look for the editorial-responsibility exception: is a named outlet or journalist taking responsibility for reviewing it, the one condition that removes the disclosure duty.

What holds and what fails

What holds is the principle across every clause: disclosure is required unless a specific, named exception applies, not the reverse. Not yet established, because the rule is only weeks old at the time of writing, is how consistently it is enforced across the chatbots and content tools EU users encounter, or how regulators will treat borderline cases such as a lightly edited AI draft. Editorially, a reader outside the EU should not assume the same disclosure exists elsewhere; this obligation is specific to this regulation's jurisdiction.

• Look for an explicit AI-disclosure notice the next time you use a chatbot based in or serving the EU.
• Check whether an image or article you suspect is synthetic carries a machine-readable or visible label.
• Do not assume this disclosure duty applies outside the EU without checking your own jurisdiction's rules.

A duty this new is still being tested in practice. What the text establishes is the standard a system is now supposed to meet; whether every provider meets it on day one is a separate question the regulation's own enforcement record will answer over time.