On 2 August 2026, the transparency obligations for certain AI systems under Article 50 of the AI Act will enter into force. While attention in recent months has focused primarily on the postponed obligations for high-risk AI, it is equally important to consider the new transparency obligations that will apply from that date.
Through the transparency obligations, the European legislator aims to ensure that natural persons can recognise when they are interacting with artificial intelligence or when certain content has been generated or manipulated by AI. The aim is to reduce the risk of deception, manipulation and disinformation.
To support companies in applying these new obligations, the European Commission published detailed guidelines shortly before they entered into force. These guidelines clarify when Article 50 applies, which exceptions apply and how companies can comply with their transparency obligations in practice.
In this regard, the AI Act distinguishes between providers and deployers of AI systems.
More specifically, providers of AI systems intended to interact directly with natural persons must design these systems in such a way that users are informed when they are interacting directly with an AI system. The AI Act does not prescribe how this information must be provided. According to the European Commission’s guidelines, however, the notification must be clear, easily distinguishable from other information and provided no later than at the time of the first interaction. The form of the notification may vary depending on the context, for example through a textual notice, a visual label or a spoken message. Providers must also take the intended target audience into account.
However, this information obligation does not apply where it is obvious to a reasonably well-informed, observant and circumspect natural person that they are interacting with artificial intelligence. In addition, providers of AI systems that generate or manipulate synthetic text, image, audio or video content must ensure that the output is marked in a machine-readable format so that AI-generated or manipulated content can be detected.
Deployers are also subject to transparency obligations. Deployers must clearly disclose when they use deepfakes, when they publish AI-generated or manipulated text concerning matters of public interest without meaningful human review or editorial control, and when they use emotion recognition or biometric categorisation systems.
For the sake of completeness, we note that the AI Act does not apply to natural persons who use AI systems exclusively in the course of purely personal, non-professional activities. The transparency obligations under Article 50 are therefore primarily aimed at providers and deployers of AI systems in a professional context.
For AI systems placed on the market before 2 August 2026, the obligations concerning machine-readable markings must be complied with from 2 December 2026. No transitional regime applies to the other transparency obligations: from 2 August 2026, these apply immediately to all AI systems falling within the scope of Article 50, regardless of when they were placed on the market. Non-compliance may result in fines of up to EUR 15 million or 3% of worldwide annual turnover, whichever is higher.
Do you have questions about the transparency obligations under Article 50 of the AI Act or would you like to determine whether your company is required to comply with these obligations? Moore Law is happy to assist companies with the legal analysis and practical implementation of the new AI regulations. Read our previous article as well: AI Act update: postponement of high-risk obligations and clarification of transparency requirements | Moore Law
