Article 50 of the AI Act: what you need to know about labelling AI content

Who Article 50 applies to, which content must be labelled, how to disclose it, when the exception applies and what the penalties are. With references to the Regulation, the Commission's guidelines and the EU Code.

Zespół AIMARKER· 6 October 2026· 5 min read

Text Art. 50 and the official EU AI icon

Article 50 of Regulation (EU) 2024/1689 on artificial intelligence, the AI Act, has applied since 2 August 2026 [1]. It concerns not only technology companies but anyone who professionally publishes content generated or altered by AI: online shops, agencies, newsrooms and creators. Below are the key points, each with its source.

Who does Article 50 apply to?

The AI Act distinguishes two roles [1, Art. 3(3) and (4)]:

provider: develops an AI system (e.g. an image generator or a language model) and places it on the market,
deployer: uses an AI system, except in the course of a personal non-professional activity.

If you run a website or a shop, or prepare content for clients as an agency, you are usually a deployer.

For providers, Article 50(2) requires AI outputs to be marked in a machine-readable format and detectable as artificially generated or manipulated [1]. Regulation (EU) 2026/1744 (the "Digital Omnibus on AI") gives providers of systems placed on the market before 2 August 2026 until 2 December 2026 to comply [2, new Art. 111(4) AI Act]. This transition period does not apply to deployers' obligations.

Which content must deployers disclose?

Article 50(4) covers two kinds of content [1]:

deep fakes, defined as "AI-generated or manipulated image, audio or video content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful" [1, Art. 3(60)],
text: deployers of an AI system that generates or manipulates text published with the purpose of informing the public on matters of public interest must disclose that the text has been artificially generated or manipulated [1, Art. 50(4), second subparagraph].

The Commission's guidelines on Article 50 give examples [4, para. 131]. Matters of public interest include politics and democratic processes, public administration, justice, public health, the environment, consumer safety and economic, financial, scientific and cultural developments. In-scope examples include a summary of a town council meeting, an article on diet and diseases, and weather warnings. Out of scope are, for example, fantasy novels, advertisements and product descriptions that make no claims about e.g. health, consumer safety or sustainability, and a summary generated by a chatbot that is only available to the user who prompted it.

Where a deep fake forms part of an evidently artistic, creative, satirical or fictional work, the obligation is limited to disclosing the existence of such content in a way that does not hamper the display or enjoyment of the work [1, Art. 50(4)]. This is a limitation, not an exemption.

When and how to disclose?

Under Article 50(5), the information must be provided "in a clear and distinguishable manner at the latest at the time of the first interaction or exposure" and conform to the applicable accessibility requirements [1].

The Commission's guidelines add that:

information that can easily be overlooked, e.g. only in a manual, hidden in menus or in terms of use, is not clear and distinguishable [4, para. 142],
for text, disclosure should appear at the start of the publication [4, para. 143],
people should be able to perceive the label without specific technical tools [4, para. 132],
for deep fakes, deployers cannot rely only on the provider's machine-readable marking [4, para. 117],
Article 50 imposes no additional accessibility requirements; existing law such as Directives 2016/2102 and 2019/882 applies [4, para. 144].

The Code of Practice and the EU icons

The final Code of Practice on Transparency of AI-Generated Content was published on 10 June 2026 [3]. Following it is voluntary. In its opinion of 8 July 2026, the Commission concluded that the Code adequately covers the obligations in Article 50(2), (4) and (5) [5]. The Code introduces a common "AI" icon [3, Annex 1]. Our separate guide explains how to use it.

Is there an exception for text?

Yes. The 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 [1, Art. 50(4)]. The guidelines set a high bar, including fact-checking as a minimum [4, paras. 134–138]. Our article on the editorial exception covers the details. The exception does not apply to deep fakes.

What about content from before 2 August 2026?

According to the guidelines, texts both generated and published before 2 August 2026 do not need to be labelled retroactively. However, texts generated earlier but published on or after 2 August 2026 must be labelled [4, para. 154].

Penalties

Non-compliance with the Article 50 transparency obligations is subject to administrative fines of up to EUR 15 000 000 or, if the offender is an undertaking, up to 3% of its total worldwide annual turnover for the preceding financial year, whichever is higher [1, Art. 99(4)(g)]. For SMEs, including start-ups, and small mid-caps, the lower amount applies [1, Art. 99(6); 2, Art. 99(6a)].

And in Poland?

The Polish Act of 3 July 2026 on artificial intelligence systems (Journal of Laws 2026, item 1003) designates the Commission for the Development and Security of Artificial Intelligence (KRiBSI) as the market surveillance authority under the AI Act [6, Art. 5(1)]. Its provisions on inspections, proceedings and penalties enter into force three months after publication [6, Art. 127]. The Article 50 obligations themselves apply directly under the Regulation from 2 August 2026.

Where to start?

Identify which content you publish is created with AI: text, graphics, photos, recordings.
Check whether it is a deep fake or text on matters of public interest.
Decide whether you rely on the editorial exception for text and who is responsible for the review.
Label content clearly and distinguishably at first exposure, with accessibility in mind.
Keep records: what was labelled or approved, when and by whom.

AIMARKER helps with these steps: it labels content with the Code's icons on any website, keeps a register and an audit log, and produces a report. See how it works.

Sources

Legal status as of 6 October 2026.

This article is for information only and is not legal advice. For a specific case, consult a lawyer.