Paralegal
Isabella Mahieu
GDPR, IT law, International Media Law
31 March 2026
Reading time: approx. 7 minutes |
- Joanne van Eenennaam
- IT law
What transparency obligations apply to AI use?
AI Office at European Commission publishes draft code of practice on transparency marking and labeling AI-generated content. What transparency obligations apply to AI use?
On 17 December 2025, the European Commission published the first draft for the world's first: “Code of Practice on marking and Labelling of AI- generated content”. The aim of the Code of Practice is to facilitate the effective implementation of a number of transparency obligations under Regulation (EU) 2024/168 (“AI Regulation”).
In this article, we summarize the currently existing rules regarding transparency through marking and labeling of AI systems and the use of these tools.
Statutory framework
Since August 2, 2025, certain major developers and users of certain AI systems have been subject to a number of transparency obligations. For AI systems to which the transparency obligations under the law do not explicitly apply regarding development and use, voluntary compliance with these transparency rules is encouraged.
Unless the proposed Digital Omnibus Regulation (“Digital Omnibus”) postpones the timeline for the gradual applicability of the transparency chapter of Regulation (EU) 2024/168 (“AI Regulation”) by another year or two given the pace of current developments in AI technology, the transparency obligations will apply to everyone across the EU from 2 August 2026.
Transparency obligations
The legal transparency obligations for providers and controllers of certain AI systems are laid down in Article 50 of the AI Regulation (Transparency obligations for providers and controllers of certain AI systems). In short, this concerns the following statutory obligations:
1. Direct interaction with natural persons: Providers of AI systems are required to ensure that AI systems intended to interact directly with natural persons are designed and developed in such a way that the natural persons concerned are informed that they are interacting with an AI-system, unless this is obvious from the perspective of an “ordinarily informed and reasonably observant and circumspect natural person”, taking into account the circumstances and the context of use.
2. Generative AI: Providers of AI systems, including general-purpose AI systems, that generate synthetic audio, image, video, or text content shall ensure that the outputs of the AI system are marked in a machine-readable format and are detectable as artificially generated or manipulated.
3. Emotion recognition: Users of an emotion recognition system or a biometric categorization system shall inform the natural persons exposed to it about the functioning of the system and process the personal data in accordance with privacy and data protection law.
4. Influencing public opinion:
-Deepfakes: Users of an AI system that generates or edits image, audio, or video content that contains a deepfake must make it known that the work has been artificially generated or manipulated.
-Entertainment media: Where the content is part of an apparently artistic, creative, satirical, fictional or analogue work or programme, transparency obligations are limited to disclosing the existence of such generated or edited content in “an appropriate manner which does not impede the display or enjoyment of the work”.
-Journalism: Users of an AI system that generates or edits text published to inform the public about matters of public interest must disclose that the text has been artificially generated or edited. However, this obligation does not apply when the AI-generated content has undergone human review or editorial control and when a natural or legal person bears editorial responsibility.
5. Time of information = time of first interaction or exposure to AI: The information must be provided to the natural persons concerned in a clear and distinguishable manner at the latest at the time of the first interaction or exposure. The information must meet applicable accessibility requirements, for example, for people with disabilities or other vulnerable groups such as children.
The transparency obligations under the AI Regulation are without prejudice to other transparency obligations under Union or national law for controllers of AI systems, such as information obligations under consumer protection law or data protection law. These transparency obligations do not apply to legally permitted uses of AI for the detection, prevention, investigation, or prosecution of criminal offenses, which are governed by a separate framework.
As is often the case with statutory norms, the transparency obligations are formulated in a somewhat general way and contain open-ended criteria, such as: "reasonably informed and reasonably observant and circumspect natural person," "an appropriate manner that does not impede the reproduction or enjoyment of the work," and "a process of human review or editorial oversight." The European legislator was aware that, in practice and in sector-specific contexts, a need for more concrete interpretation would arise.
The AI Office at the European Commission has therefore been tasked with encouraging and facilitating the development of codes of practice at EU level to facilitate the effective implementation of obligations regarding the detection and labeling of artificially generated or edited material. The recently published first draft is the first attempt at developing such a code of practice.
First draft Code of Practice on marking and labeling or AI- generated content
In the first draft of the Code of Practice for Marking and Labeling AI-Generated Content, the working group recommends, among other things, pursuing measures, such as a single, standardized machine-readable marking technique, in consultation with providers of generative AI systems. There are other examples of standardization mandated by the European legislator, such as the mandatory use of USB-C for radio equipment instead of proprietary technology unique to each company.
Measures to comply with transparency obligations
For controllers using the products of providers of generative AI systems, the draft code of practice emphasizes proactive disclosure of manipulated images, audio or video material with regard to deepfakes and proposes to pursue a ' common taxonomy ', in other words a standardized vocabulary for how to word these warnings.
The Code of Practice also proposes standardizing a " common icon" for AI-generated and manipulated material, similar to, for example, the © icon from US copyright law or the CE marking from European product safety law, so that in the future people can see at a glance whether AI has been used.
The proposed measures from the draft code of practice can serve as inspiration for your own transparency notifications, warnings, or labels about your AI use.
Handling uncertainty in early stages of AI rules
The first draft code of practice is the world's first example of a Code of Practice at the European Union level. It's still a preliminary draft, and feedback from society and the business community could lead to a completely different approach. Like AI technology itself, the applicable legislative and regulatory framework for AI is in its very early stages. Not all rules surrounding AI are yet clear or knowable. We previously wrote about how to navigate this uncertainty:
and
It will take several years for the applicable legal and regulatory framework to crystallize, but you can stay ahead of developments by being informed in a timely manner and recording the choices you are currently making.
Questions?
Are you already using AI within your company or organization? Would you like to develop an AI policy, AI ethics statement , or AI-related terms and conditions for your company or organization, but aren't sure where to begin? Contact our IT lawyers today; they're happy to help.
Archief
