If it’s AI, say so: new transparency rules upcoming

The AI Act, which entered into force on 1 August 2024, provides for a phased implementation of its obligations. Following the first phase in February 2025 and the second phase in August 2025, the third phase will commence on 2 August 2026. In the meantime, the application of certain obligations has already been postponed.

In our previous two newsflashes, we already discussed the first two phases of the AI Act: on the one hand, the promotion of AI literacy and the prohibition of certain AI practices (Phase 1), and on the other hand, the rules on sanctions, governance and general-purpose AI models (Phase 2). While the AI Act was, in principle, set to become fully applicable in August 2026, the entry into application of a number of obligations have since been postponed.

1. Transparency obligations

On 2 August 2026, the transparency obligations set out in Article 50 of the AI Act will become applicable to providers and deployers of certain AI systems. These obligations are intended to ensure that individuals are clearly informed when they are interacting with AI or being exposed to AI-generated content. Through these requirements, the EU seeks to reduce the “black box” nature of AI systems, enabling individuals to make more informed choices and strengthening trust in these technologies. In practical terms, the transparency obligations require the following:

Providers of AI systems must:

• For AI systems intended to interact directly with natural persons:
  • inform the natural persons concerned that they are interacting with an AI system, unless this is obvious from the perspective of a reasonably well-informed, observant and circumspect natural person, taking into account the circumstances and the context of use.
• For general-purpose AI systems that generate synthetic audio, image, video or text content:
  • ensure that the outputs of the AI system are marked in a machine-readable format and are detectable as artificially generated or manipulated (e.g., through “watermarking”).

Deployers of AI systems must:

• For emotion recognition systems or biometric categorisation systems:
  • inform the natural persons exposed to the system about its operation and ensure that the processing of personal data complies with the General Data Protection Regulation (GDPR).
• For AI systems that generate or manipulate image, audio or video content constituting a deepfake:
  • disclose that the content has been artificially generated or manipulated.
  • where the content forms part of an apparently artistic, creative, satirical, fictional or analogous work or programme, the transparency obligation is limited to disclosing the existence of generated or manipulated content in an appropriate manner that does not interfere with the display or enjoyment of the work.
• For AI systems that generate or manipulate text intended for publication to inform the public on matters of public interest:
  • disclose that the text has been artificially generated or manipulated.

The transparency obligation must be fulfilled no later than the first interaction with, or exposure to, the AI system or content, in a clear and distinguishable manner.

In this context, the EU has introduced a series of pictograms that providers and deployers of generative AI systems may use to comply with the transparency obligations set out in Article 50 of the AI Act. In addition, on 10 June 2026, the EU published a “Code of Practice on the Transparency of AI-Generated Content”. This code is expected to be supplemented by guidance from the European Commission on the precise scope of the transparency obligations. It already contains practical guidance for providers and deployers, including rules on the marking and detection of AI-generated or manipulated content, as well as the labelling of deepfakes and AI-generated text.

2. Postponement of remaining obligations

In principle, the remaining obligations under the AI Act were also due to become applicable on 2 August 2026. However, the European Parliament partially revised this timeline through the Omnibus VII package adopted on 16 June 2026, postponing a number of compliance deadlines. The Omnibus VII package also extends the more flexible rules currently available to SMEs to “small mid-cap enterprises” (generally, companies with fewer than 750 employees).

Key message

As of 2 August 2026, companies that develop, provide or deploy certain AI systems must comply with specific transparency obligations, the core requirement being that individuals must be clearly informed, in advance, when they are interacting with AI. This includes, among other things, the explicit labelling of chatbots, deepfakes, and AI-generated or AI-manipulated content. These obligations must be fulfilled no later than the first interaction with, or exposure to, the AI system or content. At the same time, the application of a number of other AI Act obligations has been postponed, including the rules governing certain categories of high-risk AI systems.