BRUSSELS, BELGIUM / RankWire.AI / – European Union enforcement of transparency standards for AI-produced content began across member states on August 2, 2026. Under Article 50 of the EU AI Act, organizations are now mandated to include clear notices for specific AI interactions and synthetic media. These regulations pertain to both providers designing AI systems and deployers utilizing them to publish or present material. Notably, there is no requirement for a visible label on every AI-generated item. These obligations are part of the law’s phased rollout following its ratification in 2024.
Organizations offering systems that directly engage with users must notify when they are interacting with AI. Such a notification is unnecessary if the artificial nature of the interaction is obvious to a reasonably informed person. Additionally, providers of systems that generate synthetic text, audio, images, or video are required to include machine-readable marks. These marks must enable detection of AI-generated or manipulated content, where technically feasible. Standard editing tools are exempt from this requirement if they do not substantially alter the input or its meaning.
The legislation distinguishes these technical marking obligations from the visible disclosures that deployers must provide to audiences. Deployers are responsible for identifying deepfake images, audio, and video when the material might falsely appear authentic. They must also label AI-generated text published for informing the public on matters of public interest. This text labeling does not apply after human review or editorial control when a person or organization accepts responsibility for the content. The required information must be presented clearly, distinguishably, and accessibly to the audience.
Machine marking rules are separate from public labeling obligations
Article 50 additionally covers systems used for emotion recognition and biometric categorization, which require notices to the individuals exposed. Such disclosures must appear clearly and prominently no later than the initial interaction or exposure. The law also provides some relief for certain artistic, creative, satirical, or fictional works to prevent hindering normal display and enjoyment. While disclosures for deepfakes in creative works can be limited, they cannot be entirely omitted. There are also specific exceptions for certain law-enforcement uses authorized by law and protected by safeguards.
On July 20, the European Commission issued comprehensive guidelines clarifying the scope, definitions, and exceptions. Moreover, a voluntary Code of Practice was published on June 10 for providers and deployers of generative AI systems. Companies can adopt this code to demonstrate compliance with marking and labeling duties across EU member states. Non-signatories must implement alternative measures and show authorities how these measures meet legal requirements. The Commission also released optional icons, but these icons alone do not constitute legal compliance.
Enforcement primarily falls to national authorities
The primary responsibility for enforcing the new transparency regulations lies with national market surveillance agencies. The AI Office manages limited cases involving systems connected to general-purpose AI models or large online platforms. The European Data Protection Supervisor oversees systems used by EU institutions, bodies, and agencies. Penalties for violations can reach up to 15 million euros or 3% of global annual turnover, with smaller businesses facing lower limits. Member states are tasked with establishing penalty procedures and may employ warnings or other non-monetary enforcement actions.
A transitional period applies to generative AI systems introduced before August 2, 2026. Providers of these existing systems have until December 2, 2026, to comply with the machine-readable marking requirement. This transition addresses only the technical marking and detection obligations outlined in Article 50. The other transparency obligations are already enforceable across the EU. Content generated before August 2 is not subject to retroactive labeling, whereas newly created material must adhere to the relevant disclosure and marking regulations.