The EU Artificial Intelligence Regulation (AI Act) is coming into force in stages, with August 2, 2026, marking another important date in this process. New transparency requirements for AI-generated content and systems will apply from that date. At the same time, the application deadlines for high-risk AI systems have been postponed to allow for the time needed to adapt these systems to the new requirements.
Two new transparency obligations
From August 2, 2026, entities sharing so-called deepfakes—defined as realistic images, audio, or video material depicting people, objects, places, or events that have been artificially generated or altered in a way that could mislead as to their authenticity—must inform recipients of the use of AI. If the content is part of a work of artistic, creative, satirical, or similar nature, the obligation is limited to disclosing the use of AI in a way that does not disrupt the reception of the work itself. A similar leniency applies to texts published in the public interest if they have undergone editorial review and a specific natural or legal person is responsible for the publication. Providers of content-generating systems must ensure that the materials are recognizable as being produced by AI, which in practice means implementing machine-readable tags. A transition period until December 2, 2026, is provided for systems introduced to the market before August 2, 2026, and materials published earlier do not need to be retroactively tagged.
Informing users about interactions with machines: When a user interacts with an AI system without this being obvious, especially chatbots, virtual assistants, and automated customer service systems, they must be clearly informed of this, at the latest upon first contact, in a way that is understandable and accessible even to people with special needs. A similar obligation applies to the use of emotion recognition systems and biometric categorization, although in such cases, the AI Act applies in parallel with the GDPR, not instead of it.
Before we consider ourselves fully ready – what should we remember?
First, the deadlines for high-risk AI systems have been postponed. Regulations for systems listed in Annex III to the AI Act, covering applications in education, employment, access to essential services, justice, and migration, among others, will not enter into force until December 2, 2027, instead of the originally planned August 2, 2026. High-risk systems that are part of products covered by EU security regulations have been given even more time; the requirements for them will now enter into force on August 2, 2028, rather than in 2027, as originally planned.
Secondly, extending deadlines doesn't mean it's worth pausing preparations. Adapting high-risk systems typically requires technical, organizational, and contractual changes, and their implementation is time-consuming. The additional time is intended to allow for the development of standards and procedures necessary for uniform application of requirements across the EU, not to postpone the work itself.
Third, from August 2, 2026, the European Commission will have full enforcement powers over general-purpose AI models. It can request information and documentation, conduct model assessments, and impose risk mitigation measures and sanctions.
How to prepare for change
Organizations using content-generating systems should first verify whether the materials they publish may be subject to the obligation to mark them as deepfakes and whether existing disclosure mechanisms meet the new technical requirements. Companies using chatbots and virtual assistants should verify that users are clearly informed of their interaction with the machine from the very first moment. Where emotion recognition or biometric categorization systems are used, it is worth balancing the disclosure obligations arising from the AI Act with the requirements of the GDPR, ensuring that both regimes are met simultaneously. Finally, although the new deadlines for high-risk systems seem distant, it is worth starting compliance work now, given the complexity of the technical and organizational changes they will require.
To sum up
From August 2, 2026, businesses using AI systems must expect new transparency obligations – both for AI-generated content and for users interacting with the machine. At the same time, the AI Omnibus has provided more time to adapt high-risk systems, but this does not eliminate the need to begin preparations now. Given the complexity of the new requirements and the European Commission's enhanced enforcement powers, it is advisable to consult with a specialist in new technologies law regarding your organization's scope of obligations.
This article is for informational purposes only and does not constitute legal advice.
Legal status as of August 3, 2026.
Author:
