Many companies are discussing AI-related labels. However, the real challenge of the AI Act often lies elsewhere.
EU AI Act: most companies are preparing for the wrong obligations
From 2 August 2026, the general transparency obligations of the EU AI Act, particularly those set out in Art. 50 of Regulation (EU) 2024/1689, will be fully applicable. The regulation itself has already come into force on 1 August 2024; the application of the various regulatory areas follows a staggered timetable.
Since then, we have been receiving more and more questions from marketing, sales, and compliance departments:
- Do AI-generated images on the website need to be disclosed?
- Do AI-generated product illustrations in a catalogue need to include a mention?
- Does all marketing material created with AI need to be marked accordingly?
The short answer is as follows:
Maybe. But, all too often, the wrong question is being asked.
In our view, the real challenge of the EU AI Act is not the technology.
The biggest challenge is documentation.
What the AI Act actually requires
The central standard: Art. 50 AI Act
For companies that use AI-generated content in the field of marketing and sales, Article 50, paragraph 4, of Regulation (EU) 2024/1689 is the determining provision. According to this provision, deployers of an AI system that generates or manipulates image, audio, or video content constituting a deepfake are required to disclose that the content has been artificially generated or manipulated.
Important: this obligation does not apply without limit. Article 50, paragraph 4, of Regulation (EU) 2024/1689 expressly provides for exceptions, notably for works that are clearly artistic, creative, satirical, or fictional; in these cases, appropriate disclosure is sufficient, provided it does not detract from the presentation of the work.
Furthermore, Article 50, paragraph 2, of Regulation (EU) 2024/1689 imposes obligations on providers of AI systems that generate synthetic content: they must ensure that the outputs are recognisable, in a machine-readable format, as being artificially generated or manipulated, to the extent that this is technically possible and reasonable.
Your role: deployer, not provider
Most companies that use AI tools for marketing, sales, or communication act as deployers in the sense of the regulation, and not as providers. The obligations differ significantly. Deployers generally have more limited transparency and disclosure obligations than providers, but nonetheless concrete.
It is important to distinguish this role from that of the importer of high-risk AI systems: anyone who places such systems on the market is subject, under Art. 23 of Regulation (EU) 2024/1689, to extensive obligations of verification, documentation, and retention, including the obligation to keep the EU declaration of conformity available in accordance with Art. 47 of Regulation (EU) 2024/1689 for ten years.
Not every use of AI triggers a marking obligation
In practice, we currently observe two extremes:
Variant 1: companies completely ignore the subject.
Variant 2: companies begin, as a precaution, to mark every image generated by AI.
Neither of these two approaches is generally wise.
The concrete assessment depends on the purpose of use, the nature of the content, and the circumstances of the particular case. What is decisive, in particular, is the question of whether:
- Is it deepfake content likely to mislead within the meaning of Art. 50, paragraph 4, of Regulation (EU) 2024/1689?
- Or is it purely decorative graphics, clearly stylised, that do not give the appearance of reality?
- Is the content integrated into a context that already allows for understanding that it is not real content?
Parallel legal regime: do not forget the UWG
Regardless of the AI Act, German law on fair competition imposes an independent obligation to avoid misleading presentations. According to § 5, paragraph 1, UWG, any person who engages in a misleading commercial practice that is likely to lead the consumer or any other market participant to make a commercial decision that they would not have made otherwise acts unfairly.
According to § 5, paragraph 4, UWG, visual representations can also constitute misleading information when they are likely to replace false or misleading statements.
Practical consequence: anyone who uses, in a product catalogue, AI-generated visualisations that suggest a non-existent product state or certain non-existent features risks not only a violation of the AI Act but also cease-and-desist actions and fines under German competition law. Both regimes must be complied with independently of each other.
Typical examples from practice
Website with AI-generated illustrations
A company uses AI-generated illustrations, mood visuals, or graphic visualisations on its website.
The determining question is often not:
"Should we label each image individually?"
but rather:
"Are we able to document in a traceable manner which content is generated by AI, with which systems they were created, and whether they are likely to mislead in the sense of the regulation?"
In many cases, a transparent mention in the context of website communication may be more relevant than dozens of individual markings, provided that it meets the requirements of Art. 50 of Regulation (EU) 2024/1689 in the case of concrete use.
Printed product catalogue
A manufacturer uses visualisations or AI-generated renderings in a product catalogue. The following fundamental questions arise first and foremost:
- What content has actually been generated by AI?
- What systems have been used?
- Are customers likely to be misled by the presentation regarding the product's characteristics, both in the sense of Art. 50, paragraph 4, of Regulation (EU) 2024/1689 and with regard to § 5 UWG?
- How is the decision-making basis documented?
In many cases, this documentation is already lacking.
The most common weakness: the lack of traceability
During audits and internal controls, it regularly appears that:
Companies are already using many AI systems, but cannot reliably demonstrate:
✅ which AI solutions are used
✅ which content is generated by AI
✅ which data is processed
✅ which providers are involved
✅ which risks have been assessed
✅ which decisions have been made
It is precisely here that the greatest risks in terms of governance and compliance arise, both in relation to the AI Act, the GDPR, and the UWG.
Sanctions: what are the stakes?
Regulation (EU) 2024/1689 provides for significant sanctions. Violations of transparency obligations, for example those set out in Art. 50 of Regulation (EU) 2024/1689, may be punished by fines of up to 15 million euros or 3% of the total global annual turnover of the previous financial year, whichever is higher. This may be supplemented by possible cease-and-desist actions under § 5 UWG as well as risks related to data protection.
Our recommendation
Instead of hastily placing labels on every image or catalogue, companies should first create a solid documentation base.
This includes, in particular:
- Inventory of the AI systems used (clarification of the role of deployer or provider)
- Documentation of use cases and processed data
- Risk assessment and transparency obligations in accordance with Art. 50 of Regulation (EU) 2024/1689
- Review of risks in relation to competition law in accordance with § 5 UWG
- Traceability of responsibilities and decisions
- Establishment of an AI governance process
Only on this basis is it possible to seriously assess what transparency and information obligations actually exist in the specific case.
Are you sure that your use of AI is already documented in accordance with the AI Act?
Many companies are currently focusing intensively on markings and labels. However, our experience shows that most organisations do not yet have the necessary documentation regarding their AI applications.
TMWP supports companies in the following areas:
- compliance with the AI Act
- AI governance
- data protection and AI
- transfers to third countries
- internal policies and documentation
- audit of marketing, commercial and communication processes
👉 Book an appointment now for an initial discussion without obligation and create transparency before transparency obligations are enforced.
TMWP EUwise Advisors S.A.S.U.
Governance. Risque. Compliance.
Legal notice / disclaimer
This article is intended solely to provide general information and does not constitute legal advice within the meaning of the German Legal Services Act (Rechtsdienstleistungsgesetz, RDG). The content has been prepared with care, but may not comprehensively reflect the legal situation in every specific case. In particular, this article does not replace an individual legal analysis of a specific case by qualified advisors.
Regulation (EU) 2024/1689 (EU AI Act) as well as applicable national law are subject to ongoing interpretation and development by authorities, courts, and the European Commission. At the time of publication of this article, there are still no definitive administrative guidelines or decisions from the highest courts in essential areas of the AI Act. The legal positions presented may evolve based on future developments.
TMWP EUwise Advisors S.A.S.U. disclaims all responsibility for decisions made based on this article without individual legal advice having been obtained in the specific case.
We are at your disposal for an assessment tailored to your specific situation.