BRUSSELS, BELGIUM / RankWire.AI / – European Union transparency directives concerning AI-generated content became effective across the bloc on August 2, 2026. Under Article 50 of the EU AI Act, clear notifications are now mandated for specific AI interactions and synthetic media. These requirements apply to both providers developing AI systems and deployers utilizing them to distribute or showcase content. However, a visible label on every AI-created item is not obligatory. These obligations are part of the phased rollout following the law’s adoption in 2024.
As of early August, providers whose systems engage directly with users must inform them when they are interacting with AI. If the artificial origin of the interaction is obvious to a reasonably informed individual, such notice is not necessary. For systems that generate synthetic text, audio, images, or video, providers are also required to include machine-readable indicators. These markers should enable detection of AI-manipulated or generated content whenever technically feasible. Standard editing tools remain outside this requirement if they do not significantly alter input or its meaning.
The legislation distinguishes between the technical machine-readable markings and the visible disclosures that deployers must provide to audiences. Deployers are responsible for marking deepfake images, audio, and video that could deceive viewers into believing they are authentic. They are also required to label AI-generated public interest texts. However, this labeling does not apply once the content has undergone human review or editorial oversight, with the responsible individual or organization accepting editorial responsibility. Such disclosures must be presented clearly, in a way that is easy to distinguish and accessible to the public.
Technical Marking Requirements and Public Labeling Are Kept Separate
In addition, Article 50 addresses systems used for emotion recognition and biometric categorization, which also demand notices for individuals exposed. These disclosures must be made prominently and no later than the first interaction or exposure. To avoid impeding artistic, satirical, or fictional works, the law limits some labeling duties for such creative outputs. For deepfakes, disclosures in creative contexts can be limited but not eliminated entirely. Certain law enforcement applications, authorized by law and under safeguards, are granted specific exemptions from these requirements.
European Commission provided comprehensive guidelines on July 20 to clarify the scope, definitions, and exceptions. Additionally, a voluntary Code of Practice was issued on June 10, aimed at providers and deployers of generative AI. Companies can adopt this code as a means to demonstrate compliance with the marking and labeling obligations within the EU. Organizations that do not sign up must implement other appropriate measures and demonstrate how they meet legal standards. The Commission also introduced optional icons, but these alone do not fulfill legal requirements.
Enforcement Mainly Handled by National Authorities
National authorities responsible for market oversight primarily enforce the new transparency rules. The AI Office supervises limited cases involving general-purpose AI models or major online platforms. The European Data Protection Supervisor oversees systems used by EU institutions, agencies, and bodies. Penalties for violations can reach up to 15 million euros or 3% of global annual turnover, with smaller fines applicable to smaller firms. Enforcement procedures, including warnings and other measures, are set by member states.
A transitional period is in effect for generative AI systems introduced before August 2, 2026. Providers of such existing systems have until December 2, 2026, to implement the machine-readable marking requirement. This transition pertains solely to the technical marking and detection duties outlined in Article 50. Other transparency obligations, which already apply across the EU, are unaffected. Content created before August 2 does not need retroactive labeling, but new content must adhere to the applicable disclosure and marking rules.