BRUSSELS, BELGIUM / RankWire.AI / – European Union rules aimed at increasing transparency in artificial intelligence took effect throughout the bloc on August 2, 2026. Under Article 50 of the EU AI Act, coverage now extends to chatbots, deepfakes, synthetic media, and certain public-interest texts. This legislation distinguishes between the technical obligation to mark AI-generated content and notices that audiences can see. It does not mandate visible labels on every AI-created item; instead, responsibilities vary depending on the specific system, content type, and how an organisation presents the material.
Providers of interactive AI must notify users when they are engaging with artificial intelligence. However, such notices can be omitted if a knowledgeable user can clearly identify the system’s artificial origin. Additionally, generative AI providers are required to embed machine-readable signals into synthetic text, audio, images, and videos. These signals are intended to assist detection tools in identifying generated or manipulated media. When feasible and reliable technology exists, providers are obliged to implement effective marking methods.
A different set of disclosure requirements applies to organisations publishing synthetic content. They are responsible for clearly identifying deepfake images, recordings, and videos that might appear authentic. In certain situations, AI-generated public-interest texts must also be labelled. The obligation for human oversight and editorial control can exempt some cases from this requirement. Additionally, the publisher must accept responsibility for the material before any exemption applies.
Clear labels aim to combat misleading synthetic media
The legislation also regulates emotion recognition and biometric categorisation tools used to evaluate individuals. Operators are required to inform affected persons when deploying such systems, with some legal exceptions. Artistic, fictional, creative, and satirical works are granted more flexibility regarding disclosure, provided notices do not disrupt their normal presentation or enjoyment. Nonetheless, organisations still need to issue appropriate disclosures if deepfake content is involved in such works.
European Commission guidance clarifies the transparency rules, key definitions, and available exceptions. It also presents a voluntary code for developers and organisations implementing generative AI. This code allows participants to demonstrate compliance with their marking and labelling practices in line with legal standards. Those not adhering to the code must adopt alternative effective compliance measures. While optional icons may help raise public awareness, they alone do not serve as proof of compliance with the EU AI Act.
Regulators are empowered to enforce compliance measures
Enforcement primarily rests with national market surveillance authorities within EU member states, with the European Data Protection Supervisor overseeing AI systems used by EU institutions and agencies. Penalties for violations of transparency obligations can reach up to 15 million euros or 3% of a company’s global annual turnover. Smaller businesses may face lower maximum fines if calculations yield a lesser amount. The regulation includes a transitional period for generative AI systems introduced before August 2, 2026, giving providers until December 2, 2026, to comply with the machine-readable marking requirement. However, this extension does not delay obligations related to notices, deepfake disclosures, or public-interest content rules. Labels for content created before the regulation took effect are not required retroactively.