Chatbots, Deepfakes and AI Content: What the New Transparency Rules under the AI Act Mean for Companies

27.10.2026 12:02

On August 2nd, 2026, Article 50 of the European Artificial Intelligence Act has come into force. It concerns not only technology companies that develop AI but also affects regular businesses that operate a chatbot on their website or use content created by artificial intelligence in their marketing. Below, we will summarise what the new rules require, where exceptions apply and what the best next steps are for individuals and businesses.

When the Rules Come Into Force

Article 50 of Regulation (EU) 2024/1689 of the European Parliament and of the Council (the Artificial Intelligence Act, hereinafter only the “AI Act”) applies starting August 2^nd, 2026. Article 50 imposes obligations on providers and users (so-called deployers) of AI systems, whose aim is to make sure individuals are alerted when they are communicating with artificial intelligence or consuming content generated by AI.

This year’s amendment, the so-called AI omnibus (Regulation (EU) 2026/1744, effective from 27 July 2026), postponed obligations for high-risk systems, but left transparency rules in place. Only providers of generative AI which came onto the market before August 2^nd 2026 received extra time: they must ensure machine-readable labelling of outputs by December 2^nd 2026. Content created before August 2^nd 2026 does not need to be labelled retroactively; in the case of texts, the date of publication is the deciding factor.

Whom This Concerns

The AI Act distinguishes between two roles. A provider is a person who develops or has an AI system developed and places it on the EU market or makes it available under their own name or trademark (whether for a fee or free of charge), regardless of whether the provider is located in the EU. A deployer is anyone who uses an AI system as a tool while pursuing a professional activity; the only exception is purely personal, non-professional use. A single corporation can have both roles: if, for example, you operate a chatbot on your website under your own brand, built on a third party’s model, you may have obligations as both a provider and a deployer.

When One Must Inform Users

Article 50 establishes obligations in four distinct situations:

Chatbots and other systems communicating with people (provider´s obligation). Systems intended for direct interaction with people must be designed so that the person concerned knows that they are communicating with an AI system, unless this is obvious from the situation.
Content generated by AI (provider´s obligation). Audio, image, video and text outputs of AI systems must be labelled in a machine-readable format (e.g. in metadata or with a digital watermark) and detectable as artificially created or modified. This is therefore not a visible “label”. The obligation does not apply if AI performs an assistance function in standard editing or does not substantially change the input data or their meaning.
Emotion recognition and biometric categorisation (deployer´s obligation). People must be informed about the presence of a system that can recognise emotions or about a biometric categorisation system, and their personal data must be processed in accordance with the GDPR. In the workplace and in schools, emotion recognition (except for health and safety reasons) has already been prohibited since February 2025.
Deepfakes (deployer´s obligation). If an image, sound, or video created or modified by AI represents a so-called “deepfake”, that is, resembles existing persons, objects, places or events and could appear to be real, the person who created or modified it using AI must state that it is an artificially created content. For clearly artistic, satirical or fictional works, a notice that does not disrupt their perception is sufficient.
Texts on matters of public interest (deployer´s obligation). Text created or modified using AI, which is published for the purpose of informing the public about matters of public interest, must be labelled as artificially created. This does not apply if the text has been reviewed by a human, and someone bears editorial responsibility for its publication.

All such information must be provided in a clear and distinguishable manner no later than at the first interaction with the system or the first contact with the content, and must meet accessibility requirements.

Who Is Monitoring and What Are The Risks

Supervision is carried out by national market supervisory authorities, in the Czech Republic among others by the Czech Telecommunications Office; in some cases, the European Artificial Intelligence Office is competent. For violations of Article 50, the AI Act provides for fines of up to EUR 15 million or 3% of annual worldwide turnover, whichever is higher. For small and medium-sized enterprises, the lower of the two values applies.

The European Commission’s guidelines on Article 50 from July 2026 and the Code of Practice for Transparency of AI-Generated Content, which also contains examples of appropriate labelling, serve as guidance for interpretation. Both documents are non-binding, but it can be expected that supervisory authorities will rely on them.

Our Recommendations:

Make a list of ways in which you use AI in your company (website, customer support, marketing, HR, internal tools).
For each tool, determine whether you are a provider, a deployer, or both.
Add a clear notice to chatbots and voice assistants stating that they are AI.
Set internal rules for labelling images, videos and audio created by AI and for editorial review of texts.
Verify with suppliers of AI tools how they handle machine-readable labelling of outputs and reflect this in contracts.
If you operate your own generative tool placed on the market before August 2^nd 2026, prepare for the December 2^nd 2026 deadline.

Not sure whether or how the new rules concern you? We will be happy to help you review your use of AI tools, help set up internal rules, and amend contracts with suppliers.

Useful links

The update was prepared by JUDr. Adam Felix, LL.M., Ph.D.