BRUSSELS, BELGIUM / RankWire.AI / – Effective from August 2, 2026, the European Union has enforced new transparency regulations for artificial intelligence. According to Article 50 of the EU AI Act, companies now have specific obligations concerning the identification of certain AI-driven interactions and artificially generated content. The legislation addresses chatbots, deepfakes, media produced via AI, and some publicly relevant texts. However, it does not mandate visible labels on all AI-generated material. Instead, the requirements vary depending on the system, the nature of the output, and how organizations present it.
Operators of interactive AI systems are required to notify users when they are engaging with artificial intelligence. This obligation does not apply if a typical user can easily recognize the system’s artificial origin. Additionally, providers of generative AI must embed machine-readable data within synthetic media—texts, images, audio, and video—to facilitate automated detection of altered or generated content. This marking should be applied when existing technology can reliably and effectively implement it.
Rules regarding the public disclosure of synthetic media are separate. Distributors must identify deepfake images, audio, or video that might be mistaken for authentic. They are also required to label AI-generated content related to public interest topics under certain conditions. The labeling obligation does not apply if the material is reviewed and editorially controlled by individuals. Furthermore, a specific person or organization must accept responsibility for the publication before this exemption is granted.
Public disclosure obligations mainly target deepfakes and civic information
The regulations also extend to biometric categorization and emotion recognition technologies that evaluate individuals. Those deploying such systems must inform the affected parties unless explicitly authorized by law. The EU AI Act provides more leniency for creative, fictional, artistic, and satirical works, allowing disclosures to be made in a way that preserves the work’s presentation and enjoyment. Nonetheless, operators are still required to disclose deepfake content within these works when applicable.
The European Commission has issued guidance clarifying the scope of Article 50 and its exceptions. It has also introduced a voluntary code for organizations involved in developing or deploying generative AI. This code offers a framework for demonstrating compliance with labeling and marking standards. Those opting out of the code must rely on other effective compliance measures. While optional icons can help raise public awareness, they do not alone fulfill legal requirements.
Authorities can levy substantial penalties for violations
Enforcement will primarily be overseen by national market surveillance authorities across the EU, with the AI Office supervising a select group of systems related to general-purpose AI models and major platforms. The European Data Protection Supervisor will monitor systems used by EU institutions and agencies. Penalties for non-compliance can reach up to 15 million euros or 3% of a company’s global annual turnover, whichever is higher.
A transitional period is available for generative AI systems already on the market before August 2, 2026. Their providers have until December 2, 2026, to adhere to the machine-readable marking requirement. This extension pertains solely to the technical aspect of detectable synthetic content and does not delay the implementation of chatbot notices, deepfake disclosures, or rules concerning public-interest texts. Additionally, organizations are not required to label content created prior to the enforcement of these transparency rules.