The commercial sector supports the European Commission's fundamental goal of concretising the transparency obligations of the European Artificial Intelligence Regulation (AI Regulation) through the guidelines, thereby contributing to the protection of the information ecosystem. However, the guidelines still leave some questions unanswered in their implementation. From the perspective of the commercial sector, clear, unambiguous, and implementable requirements are needed to ensure that companies can meet the obligations safely and with reasonable effort.

Background

Article 50 of the AI Regulation formulates transparency obligations for certain AI systems. Providers and operators must ensure that users can recognise when they are interacting with AI (e.g. chatbots), when content has been artificially generated or manipulated (e.g. synthetic content like deepfakes), and that corresponding labels are applied. Specific information obligations also apply to systems for emotion recognition or biometric categorisation, so that affected persons are informed of their use. On 08.05.2026, the EU Commission published a draft of the guidelines on transparency obligations under Article 50 AI regulation. Initially, these obligations were to come into force on 02.08.2026.

DIHK demands

• The guidelines should not extend the transparency obligations of the AI Regulation to other actors without a legally binding basis.
• Double or contradictory transparency obligations should be avoided, especially with regard to the comparable requirements of the European Medical Device Regulation and In Vitro Diagnostics Regulation.
• Regular reminders and the proposed multimodal approach should be limited to high-risk cases to avoid "banner blindness".
• The obligation to label and identify AI-generated or manipulated content requires a clear de minimis threshold. Industrial AI applications with purely internal or machine-centred purposes should be completely excluded from the transparency obligations under Article 50 (2) AI regulation.
• The guidelines should include practical labelling methods that prevent overloading the display area and comply with common business standards.
• The definition of deepfakes seems unsuitable for commercial activities, as a distinction between product-related and non-product-related information would be required here.
• To reduce the barriers posed by fines for small and medium-sized enterprises, the guidelines should clarify that the competent authorities, when imposing a penalty, must consider the company's profit margins, the company's influence on the development of the respective AI system, and the transparency of the AI system itself.

What matters to companies

You should consider the following:

• Under current law, the transparency obligations of the AI regulation are expected to apply from 02.08.2026. As part of the negotiations on the Digital Omnibus Regulation on AI, it is being considered to postpone the deadline for AI systems placed on the market before 02.08.2026 by four months to 02.12.2026.
• Companies must inform users when they interact with an AI system. According to the EU Commission's guidelines, it will not be permissible to merely point this out in the terms and conditions, the URL or machine-readable formats.
• Providers of AI systems must ensure that synthetic content of an AI system is labelled in a machine-readable format and recognisable as artificially generated or manipulated.
• Deepfakes must be disclosed.

Download

The DIHK statement on the Commission's draft guidelines on transparency obligations under Article 50 AI regulation is available in English:

(only available in German)