Article 50 of the EU AI Act has applied since 2 August 2026. It sets out when people must be told that they are talking to an AI or looking at AI-generated content. For many companies, this is the first part of the AI Act that reaches everyday work: the website, customer service, marketing. Austria's regulator RTR announced the start in July.
As of 28 October 2026
What does Article 50 cover?
The article splits its obligations between two roles. Providers develop an AI system and place it on the market. Deployers use an AI system in the course of their own activities. Which role a company plays depends on the system. The wording is in the text of Article 50.
Chatbots: a provider obligation
AI systems that interact directly with people must be designed so that those people are informed, unless it is obvious from the circumstances. This obligation sits with the provider. A company that deploys an off-the-shelf chatbot therefore relies primarily on the vendor to build in the notice. Whether a company that builds its own chatbot, or offers one under its own name, becomes a provider has to be checked case by case. Whatever the role, one plain sentence in the first chat window removes any doubt: "You are chatting with an AI assistant."
Deepfakes: a deployer obligation
With deepfakes, the roles flip. Deployers who generate or manipulate image, audio or video content that constitutes a deepfake must disclose that it has been artificially generated or manipulated. Human review does not lift this obligation. Here the duty lies with the company itself, whichever tool produced the image.
AI-generated text on matters of public interest
Text published to inform the public on matters of public interest must be disclosed as AI-generated. The exception applies only when two conditions are met: a human has reviewed the text or exercised editorial control, and a natural or legal person holds editorial responsibility. RTR explains this exception on its page about deployer obligations (in German). Whether an ordinary company blog counts as "public interest" is a matter of interpretation. A company that takes editorial responsibility and documents its review is in a stronger position either way.
Machine-readable marking and the transition period
Under Article 50(2), providers of generative AI systems must mark synthetic output in a machine-readable format. This is the only obligation with a transition period. The Digital Omnibus, Regulation (EU) 2026/1744, published in the Official Journal on 24 July 2026, gives systems placed on the market before 2 August 2026 until 2 December 2026. Systems placed on the market since 2 August get no extra time. Every other Article 50 obligation has applied since 2 August 2026 without any transition: the chatbot notice, deepfake disclosure and the rule for text.
To help with implementation, the Commission published a Code of Practice on marking and labelling AI-generated content on 10 June 2026. It is voluntary and covers paragraphs 2, 4 and 5; by the end of July 2026 it had around 190 signatories. Experts disagree on what legal effect signing it has.
What applies to AI-generated images?
In May 2026 the Commission put draft guidelines on Article 50 out for consultation, and on 20 July 2026 it published its guidelines on the transparency obligations. According to an analysis by the law firm Reed Smith, four rules of thumb follow from them:
- Advertising that uses AI-generated content almost always needs a standard label.
- An AI-generated background behind a real product photo is not covered.
- An image that makes a product look better than it really is counts as a deepfake.
- Content created before 2 August 2026 does not need to be labelled retroactively.
Companies should check these points against the original guidelines before turning them into internal rules.
On flb.co.at, labelling works like this: every post states that it was written with AI assistance and editorially reviewed. AI-generated images carry a visible "AI-generated" label, and the same note appears in the alt text for screen readers. A field in the CMS drives the label, so it does not depend on the editor's memory.
What are the penalties?
Under Article 99, breaches of the Article 50 transparency obligations can be fined up to 15 million euros or 3 percent of worldwide annual turnover, whichever is higher. For small and medium-sized enterprises and, since the Digital Omnibus, small mid-caps too, the lower of the two amounts applies. The consolidated text on EUR-Lex is authoritative.
Where can companies in Austria get help?
The first port of call is RTR's AI Service Desk (KI-Servicestelle). It has existed since spring 2024 and provides information and guidance. Its legal basis gives it neither supervisory powers nor the power to fine. According to RTR's FAQ, national responsibility for enforcing Article 50 in Austria has not yet been assigned, and the sources reviewed name no designated market surveillance authority. This article is not legal advice for individual cases.
What should an SME check now?
- List every place where AI talks to customers or produces content: chatbot, phone assistant, image, video and text tools.
- Record for each system whether your company is the provider or the deployer.
- For systems you buy in, ask the vendor in writing how they implement the chatbot notice and machine-readable marking.
- Check that the AI notice is visible at first contact, on mobile as well.
- Decide when an image, a video or a voice is labelled as AI-generated, and enforce the rule technically in your CMS.
- Name a person with editorial responsibility for published text and document the review.
- If you provide a generative system yourself that was on the market before 2 August 2026, schedule machine-readable marking before 2 December 2026.
- Make one person responsible for following the guidelines and the code, and use RTR's AI Service Desk for questions.
The AI consulting page describes how such a review works with outside support.