BRUSSELS, BELGIUM / RankWire.AI / – European Union rules mandating clarity for certain AI-produced materials took effect across the bloc on August 2, 2026. Article 50 of the EU AI Act now imposes responsibilities on both developers and users of artificial intelligence systems. These regulations encompass chatbots, synthetic media, deepfakes, and specific content related to public interests. Not every AI-generated item must carry a visible warning. Instead, the law dictates different requirements depending on the system, type of content, and publication method.
When AI systems interact directly with individuals, they are required to reveal that users are engaging with artificial intelligence. This obligation does not apply if the system’s nature remains obvious to a well-informed and attentive individual. Providers of generative AI must also incorporate machine-readable markers into synthetic text, audio, images, and video. These markers are intended to assist detection tools in identifying generated or manipulated content. The rule applies when the technology can reliably mark outputs without rendering them unusable or significantly altering their normal function.
Separate disclosure obligations exist for organizations publishing or presenting synthetic media. They must identify deepfake images, recordings, and videos that might appear authentic. They are also required to label AI-generated text that informs the public on matters of public interest. However, this text labeling requirement does not apply after thorough human review and editorial oversight. The responsible individual or organization must accept accountability for the published material. Notices must stay clear, accessible, and visible during the initial interaction or exposure.
Deepfake disclosures are mandatory under the new regulation
Article 50 also addresses emotion recognition and biometric categorization tools used on individuals. Operators must inform people when such systems analyze them, with limited legal exceptions. Artistic, satirical, fictional, and creative works are subject to narrower disclosure rules under this regulation. Notices for these works should not disrupt their normal display or enjoyment. Nonetheless, the deepfake disclosure obligation remains relevant and applicable. Certain law enforcement applications may qualify for specific exemptions if authorized by law and safeguarded appropriately.
The European Commission has issued guidance and a voluntary code to promote uniform compliance among member states. Providers and operators can adopt this code to demonstrate how their marking and labelling systems satisfy legal standards. Participation is voluntary, but the transparency requirements are compulsory. Companies not adhering to the code must implement other effective measures and provide explanations to regulators. While optional icons can aid audiences in recognizing synthetic content, using an icon alone does not guarantee full compliance with the AI Act.
Member state authorities oversee enforcement of transparency rules
National market surveillance agencies bear primary responsibility for enforcing these transparency standards. The AI Office monitors limited cases involving certain systems linked to general-purpose AI models and large online platforms. The European Data Protection Supervisor supervises systems used by EU institutions and agencies. Violations can result in fines up to 15 million euros or 3% of the company’s global annual turnover. Smaller companies are subject to lower ceilings under the regulation. Member states can also impose warnings and other corrective actions through their enforcement systems.
A transitional period applies for generative AI systems introduced before August 2, 2026. Providers of these systems have until December 2, 2026, to comply with the machine-readable marking requirement. This extension addresses only that specific technical obligation. It does not delay the enforcement of chatbot notices, deepfake disclosures, or rules for public-interest content. Content created before August 2 does not need retroactive labelling. All new content and interactions from the enforcement date must adhere to the applicable transparency obligations.