On 2 August 2026, the transparency requirements under Article 50 of Regulation (EU) 2024/1689 of the European Parliament and of the Council (the AI Act) became applicable.
These rules establish obligations relating to the use of AI systems and distinguish between four main regimes.
1. Transparency of Chatbots
The first regime concerns AI systems intended to interact directly with natural persons, typically chatbots or voice assistants. Providers of such systems must ensure that they are designed and developed in a way that informs individuals that they are interacting with an AI system rather than a human, unless this is already obvious from the circumstances and context of use to a reasonably well-informed, observant and circumspect person.
An exception applies to AI systems authorised by law to detect, prevent, investigate or prosecute criminal offences. However, this exception does not apply where such systems are available to the public for the purpose of reporting criminal offences, in which case the information obligation applies again.
2. Labelling of AI-Generated Content
The second regime requires providers of AI systems (including general-purpose AI systems) generating synthetic image, audio, video or text content to ensure that their outputs are marked in a machine-readable format and detectable as artificially generated or manipulated, using effective, interoperable, robust and reliable technical solutions, insofar as technically feasible.
This obligation applies to providers of AI tools (e.g. developers of generative AI models), rather than their users.
The obligation does not apply where AI performs only an assistive function for standard editing or does not substantially alter the input data provided by the user or its meaning. It also does not apply where the use is authorised by law for the purposes of detecting or prosecuting criminal offences.
3. Emotion Recognition and Biometric Categorisation Systems
The third regime requires entities deploying emotion recognition systems (which identify or infer individuals' emotions or intentions based on biometric data) or biometric categorisation systems (which assign individuals to categories such as sex, age or ethnic origin based on biometric data) to inform the natural persons exposed to such systems about their operation and to process personal data in accordance with applicable data protection legislation.
Outside the law enforcement context (where the obligation does not apply, provided that safeguards for the rights of third parties are maintained), these systems may be used, for example, in medicine or marketing, and potentially also in call centres in the future.
The information obligation does not extend to purely ancillary biometric categorisation features that are intrinsically linked to another service, such as social media filters or virtual product try-on features in e-commerce.
4. Transparency of Deepfake Content and Texts on Matters of Public Interest
The fourth regime distinguishes between two separate obligations.
The first applies exclusively to deepfake content, i.e. image, audio or video material generated or manipulated by artificial intelligence that closely resembles existing persons, objects, places or events and could falsely appear authentic to the person concerned.
Entities deploying such content are required to clearly and distinctly disclose that it has been artificially generated or manipulated.
Where the content forms part of an evidently artistic, creative, satirical or fictional work, this obligation is limited to disclosing the existence of such generated or manipulated content in a manner that does not interfere with the display or enjoyment of the work.
The second obligation concerns texts generated or manipulated by artificial intelligence that are published for the purpose of informing the public on matters of public interest, typically news or journalistic content.
This obligation therefore does not apply to visual advertising materials or to texts that do not serve an informational purpose in relation to matters of public interest.
What Does This Mean in Practice?
The AI Act introduces different transparency requirements depending on the type of AI system, the nature of the generated content and the role of the entity developing or using the AI system.
When using AI for content creation, it is particularly important to distinguish between the general obligation of AI providers to label synthetic outputs through technical (machine-readable) means and the obligation of deployers (typically marketing or PR agencies, rather than necessarily the client commissioning a campaign) to publicly disclose the artificial origin of content, but only in cases expressly specified by the Regulation — primarily deepfake content and texts concerning matters of public interest.
Outside these specifically defined cases, the AI Act itself does not impose an obligation to visibly disclose the AI-generated nature of content in graphics or accompanying text.
At the same time, the AI Office, established by a separate decision of the European Commission as a unit organisationally integrated into the European Commission, is tasked with supporting the development of voluntary codes of practice for detecting and labelling artificially generated content. These may, in the future, extend transparency standards beyond the minimum legal requirements.
It also does not apply where the text has undergone human editorial review and a natural or legal person (typically a publisher) holds editorial responsibility for its publication, or where its use is authorised by law for criminal law enforcement purposes.