Law · Updated October 2026

AI labelling: what Art. 50 AI Act means for your website (as of October 2026)

Draft. This page has not been editorially reviewed yet. The legal review is still pending.

You must make an AI chatbot on your website recognisable as AI, and you must label realistic-looking AI images or videos (deepfakes). An AI-written "About us" text that you've reviewed yourself usually doesn't need a label. The duties are set out in Art. 50 of the AI Act (Regulation (EU) 2024/1689) and have applied since 2 August 2026. According to specialist reports, the Digital Omnibus did not postpone this date. The text duty only covers texts on matters of public interest that haven't been reviewed by a person.

What the law requires

Art 50 Abs 1 VO (EU) 2024/1689 (KI-VO): Providers must design AI systems that interact directly with people so that those people are informed that they are interacting with an AI system – unless this is obvious to a reasonably well-informed, observant and circumspect person. According to the RTR, users must be informed at the latest at the first interaction; a notice only in the terms and conditions is not enough.
Art 50 Abs 4 UAbs 1 KI-VO (Deepfakes): Deployers of an AI system that generates or manipulates image, audio or video content constituting a deepfake must disclose that the content has been artificially generated or manipulated. For evidently artistic, creative, satirical or fictional works, the duty is limited to a disclosure that does not hamper the display of the work.
Art 50 Abs 4 UAbs 2 KI-VO (KI-Texte zu Angelegenheiten von öffentlichem Interesse): Deployers who publish AI-generated or AI-manipulated text to inform the public on matters of public interest must disclose this. This does not apply if the text has undergone human review or editorial control and a natural or legal person holds editorial responsibility. According to the RTR, matters of public interest include politics, public administration, public health or environmental protection; a mere spell check does not count as review.
Art 50 Abs 5 KI-VO: The information must be provided clearly and distinctly, at the latest at the time of the first interaction or exposure, and must meet the accessibility requirements.
Art 113 KI-VO; VO (EU) 2026/1744 (Digital Omnibus zur KI): Art 50 has applied since 2 August 2026. According to secondary sources, the Digital Omnibus on AI (Regulation (EU) 2026/1744 of 8 July 2026, in force since 27 July 2026) did not postpone the date from which Art 50 applies; only for the providers' marking duty under Art 50 Abs 2 is there a transition period until 2 December 2026 for systems placed on the market before 2 August 2026. What was postponed were mainly the high-risk obligations.
Art 99 Abs 4 und 6 KI-VO: Breaches of Art 50 can be punished with fines of up to EUR 15 million or 3 % of worldwide annual turnover, whichever is higher; for SMEs, the lower amount applies in each case.

Checklist

Item	Legal basis
The chatbot shows a notice that it is AI at the latest with the first message (Visible in the chat window, not only in the terms and conditions or privacy policy (RTR).)	Art 50 Abs 1 und 5 KI-VO
Realistic-looking AI-generated images/videos/audio (deepfakes) are labelled directly on the content (E.g. an AI photo that looks like a real team or business photo.)	Art 50 Abs 4 UAbs 1, Abs 5 KI-VO
AI texts on topics of public interest: either reviewed by a person with editorial responsibility, or labelled as AI-generated (Typical business texts (About us, services) are usually not of public interest under the RTR definition; documenting the review internally is advisable.)	Art 50 Abs 4 UAbs 2 KI-VO
Labelling is accessible (e.g. text rather than just an icon) (The notice can also be read by screen readers.)	Art 50 Abs 5 KI-VO

Who is exempt

AI interaction that is obvious to a reasonably well-informed, observant and circumspect person (Art 50 Abs 1)
AI texts on matters of public interest that have been reviewed by a person or are under editorial control, and for which someone holds editorial responsibility (Art 50 Abs 4 UAbs 2)
Texts that do not serve to inform the public on matters of public interest are not covered by Art 50 Abs 4 UAbs 2
For evidently artistic, satirical or fictional works, only a limited disclosure (Art 50 Abs 4 UAbs 1)

Common mistakes

Your website chatbot greets visitors with a name like "Anna" without saying it's an AI.
You mention the AI chatbot only in your privacy policy instead of directly in the chat window.
You use an AI-generated photo that looks like your real team or your real shop, with no notice.
You label every AI-assisted text across the board, although you reviewed it yourself and it doesn't concern a topic of public interest – unnecessary, but not a breach.
You publish AI texts unreviewed, such as news on changes in the law or health topics, without labelling them.
Your AI notice is just a small icon without text that screen readers don't read out.
You believe the Digital Omnibus postponed the labelling duties – according to secondary sources they have applied since 2 August 2026.

Frequently asked questions

Do I have to label my AI-written "About us" text?

Usually not. Art. 50 Abs. 4 only requires AI texts to be labelled if they inform the public on matters of public interest – according to the RTR, for example politics, public administration or public health. Even then, the duty doesn't apply if a person has reviewed the content of the text and someone holds editorial responsibility. So really read the text through and check the facts, not just the spelling.

I have an AI chatbot on my website. What do I have to do?

Visitors must learn at the latest at the first interaction that they're writing to an AI (Art. 50 Abs. 1 and 5). The RTR stresses that a notice only in the terms and conditions or general texts isn't enough. The simplest way is a clear sentence directly in the chat window, such as "You're chatting with an AI assistant". If the AI nature is obvious to everyone, the duty doesn't apply – but you shouldn't rely on that.

Can I use AI-generated images on my website?

Yes. You have to label them if they're a deepfake – i.e. they show people, places or events so realistically that they could falsely appear to be real (Art. 50 Abs. 4). So an AI photo that looks like your real team or your real shop should be labelled directly on the image. Clearly recognisable illustrations or icons aren't covered by this duty.

Has the Digital Omnibus postponed AI labelling?

According to specialist reports, no. The Digital Omnibus on AI (Regulation (EU) 2026/1744, in force since 27 July 2026) mainly postponed the high-risk obligations. Art. 50 still applies from 2 August 2026. Only for the AI providers' technical watermarking duty (Art. 50 Abs. 2) is there a transition period until 2 December 2026 for systems already on the market – that concerns the tool makers, not your website.

What happens if I forget the labelling?

For breaches of Art. 50, Art. 99 Abs. 4 of the AI Act provides for fines of up to EUR 15 million or 3 % of worldwide annual turnover; for SMEs, the lower amount applies in each case. How strictly this will be enforced in Austria is still open. A clear notice on the chatbot and on realistic AI images takes you little effort.

This page is not legal advice. What counts are the laws in force and the information from your chamber or your legal adviser.

Request a free draft for your business

No obligation, no appointment needed. Send me a few details and I'll build your first draft.

Get a free, no-obligation quote

Sources

artificialintelligenceact.eu – Art 50 KI-VO (Text) (retrieved on 3 October 2026)
RTR KI-Servicestelle – AI Act: Transparenzpflichten (retrieved on 3 October 2026)
Baker Botts – EU AI Act Article 50 Transparency Obligations Go Live (Sept. 2026) (retrieved on 3 October 2026)
artificialintelligenceact.eu – Art 99 KI-VO (retrieved on 3 October 2026)
artificialintelligenceact.eu – Art 113 KI-VO (retrieved on 3 October 2026)
RTR – Presseinformation 02.07.2026 zu Art 50 KI-VO (retrieved on 3 October 2026)
Gibson Dunn – EU AI Act Omnibus Agreement (retrieved on 3 October 2026)
EUR-Lex – VO (EU) 2026/1744 (Digital Omnibus zur KI) – UNGEPRÜFT: Abruf technisch blockiert, Inhalt nur über Sekundärquellen (retrieved on 3 October 2026)

Last checked on 3 October 2026.