Last updated:
Article 50 of the EU AI Act (often written as Art. 50 AI Act) is the transparency article. It requires people to be told when they are talking to an AI system, AI-generated content to be marked, and deepfakes and AI-written public-interest text to be labelled. It has applied since 2 August 2026.
Not to be confused with Article 50 of the Treaty on European Union, the article used for Brexit.
Article 50 is the part of the EU AI Act that applies to everyday AI: chatbots on websites, voice agents, AI avatars, image and video generators, and AI-written content. Unlike the high-risk rules, it is about telling people, not about approving systems. The full text of Article 50 is available online.
| Paragraph | Who it applies to | What it requires |
|---|---|---|
| 50(1) | Providers of AI systems that interact with people (chatbots, voice agents, avatars) | Design the system so people are told they are interacting with AI, unless it is obvious |
| 50(2) | Providers of AI that generates audio, images, video or text | Mark outputs in a machine-readable format so they are detectable as AI-generated |
| 50(3) | Deployers of emotion recognition or biometric categorisation | Inform the people exposed to it |
| 50(4) | Deployers who publish deepfakes or AI-written text on matters of public interest | Disclose that the content is AI-generated or manipulated |
| 50(5) | Everyone above | Give the information clearly and distinguishably, at the latest at first interaction or exposure, and accessibly |
A provider builds or supplies the AI system. A deployer is the business using it. If you embed a vendor's chatbot on your site, the main 50(1) duty sits with the provider, but what your visitors actually see is often in your hands, so check both.
People must be told they are interacting with an AI system, unless that would be obvious to a reasonably well-informed, observant and circumspect person, taking the context into account. Law-firm summaries of the Commission's guidelines say this exception is read narrowly: general public awareness that chatbots exist is not enough, and what counts as obvious depends on who the system's audience actually is.
In practice that means a plain statement where the conversation starts, such as "You're chatting with an AI assistant", or a spoken statement at the start of a voice call. See our guide to chatbot disclosure.
Providers of generative AI must mark outputs in a machine-readable way and make them detectable as AI-generated, as far as technically feasible. Assistive tools that do not substantially alter the input, such as grammar correction, are exempt. The voluntary Code of Practice, published on 10 June 2026, sets out practical ways to do this.
There is a transition period: providers of systems placed on the market before 2 August 2026 have until 2 December 2026 to implement marking. This applies to 50(2) only. See the 2 December deadline.
Deployers of these systems must inform the people exposed to them. Data protection law, including the GDPR, continues to apply alongside.
Deployers who publish deepfakes, meaning image, audio or video content that resembles real people, objects, places or events and would falsely appear authentic, must disclose that it is AI-generated or manipulated. For evidently artistic, creative, satirical or fictional work, the disclosure can be limited to a way that does not spoil the work.
Deployers who publish AI-generated text to inform the public on matters of public interest must disclose it, unless the text has gone through human review or editorial control and someone holds editorial responsibility. See deepfake disclosure and AI-generated content labelling.
The information must be clear and distinguishable, given at the latest at the first interaction or exposure, and must meet applicable accessibility requirements. Law-firm summaries of the guidelines say that a notice buried in the terms of service, a vague label such as "assistant", or a machine-readable mark alone that people cannot see, will not do.
The Act does not name a language for disclosures. A disclosure the visitor cannot read is hard to call clear. More in AI disclosure and language.
National market surveillance authorities in each EU country enforce Article 50. Designation of these authorities has been uneven across member states, so check the position in the countries you serve.
Under Article 99(4), breaches can lead to fines of up to €15 million or 3% of total worldwide annual turnover, whichever is higher. For SMEs and start-ups, Article 99(6) applies the lower of the two amounts.
It can. The AI Act covers providers that place AI systems on the EU market, and providers and deployers outside the EU where the system's output is used in the EU. A company based in the US, UK or New Zealand with a chatbot serving EU visitors should not assume it is out of scope. UK business? See Does the EU AI Act apply to the UK?
A scan of your public website can check some of this: whether a chat, voice or avatar widget from a known vendor shows an AI disclosure, how clearly a disclosure it finds is presented, whether it is in the page's language, and whether images carry an AI marker with no visible label. It cannot see behind logins or consent banners, hear a voice call, see whether your outputs are machine-marked, or judge how you run your business. You can run a free scan, or read how the scan works in full. A clean result is not a compliance clearance, and this page is not legal advice.
It is the transparency article of the EU AI Act. It requires disclosure when people interact with AI, machine-readable marking of AI-generated content, and labels on deepfakes and AI-written public-interest text. It has applied since 2 August 2026.
On 2 August 2026. Providers of generative systems placed on the market before that date have until 2 December 2026 for the machine-readable marking duty in 50(2).
In most cases, yes, unless it is genuinely obvious to your audience that they are talking to an AI. The disclosure should appear where the conversation starts, in plain words.
No. They are non-binding guidance published on 20 July 2026, intended to guide how authorities apply Article 50. Only the EU courts can give binding interpretations.
Up to €15 million or 3% of worldwide annual turnover, whichever is higher. For SMEs and start-ups, the lower of the two applies.
It can. The AI Act covers providers that place AI systems on the EU market, and providers and deployers outside the EU where the system's output is used in the EU. A company based in the US, UK or New Zealand with a chatbot serving EU visitors should not assume it is out of scope.
No. Article 50 has no conformity assessment or certificate. The Code of Practice is voluntary.
Article50.io is an automated scanner for the Article 50 transparency rules of the EU AI Act. It scans a website for potential Article 50 transparency gaps and generates a report of what the scan found and what it could not check, with remediation guidance, implementation instructions and disclosure wording you can paste in.
The free scan shows your single most severe finding in about 30 seconds — no signup, public pages only.
Automated technical guidance, not legal advice. Citations refer to Regulation (EU) 2024/1689.