BRUSSELS, BELGIUM / RankWire.AI / – Effective from August 2, 2026, the European Union introduced new transparency regulations for artificial intelligence. These rules, outlined in Article 50 of the EU AI Act, specify how businesses must identify certain AI-driven interactions and synthetic media, including chatbots, deepfakes, generated media, and some public-interest texts. The legislation does not require all AI-produced material to be visibly labeled but assigns different responsibilities based on the system, its output, and how it is presented by an organization.
Companies operating interactive AI systems are obligated to inform users when they are engaging with artificial intelligence. This requirement does not apply if an average user can easily recognize the artificial nature of the system. Providers of generative AI must also include machine-readable data within synthetic content—such as text, images, audio, and video—that facilitates automated detection of altered or generated material. This measure should be implemented when the available technology can reliably and effectively support marking.
Separate public disclosure obligations apply to organizations distributing synthetic media. They must clearly identify deepfake images, audio, or videos that might appear authentic and, in certain cases, label AI-generated content related to public interest topics. However, these labeling requirements do not apply if individuals review and exercise editorial control over the material, and a responsible person or organization has accepted accountability for its publication before the exemption is granted.
Public notification focuses on deepfakes and civic-related information
Technologies used for emotion recognition and biometric categorization that evaluate individuals are also subject to these rules. Operators must notify those affected when deploying these systems, except in legally authorized limited cases. The EU AI Act provides more flexible disclosure requirements for creative, fictional, artistic, and satirical works, which should maintain their normal presentation and enjoyment. Nonetheless, operators are still required to disclose deepfake content embedded within these works.
European Commission has issued guidance clarifying the scope and exceptions of Article 50. Additionally, it published a voluntary code designed for companies developing or using generative AI systems. This code allows firms to demonstrate how their labeling and marking practices align with legal standards. Those choosing not to participate must implement other effective compliance strategies. While optional icons can help raise public awareness, they do not alone guarantee compliance.
Authorities are authorized to levy hefty fines for non-compliance
Across EU member states, national market surveillance agencies will oversee enforcement, with the AI Office focusing on general-purpose AI models and major platform systems. The European Data Protection Supervisor will supervise systems used by EU institutions and agencies. Penalties for violations can reach up to 15 million euros or 3% of a company’s global annual turnover, according to the law.
Existing generative AI systems on the market before August 2 are granted a transitional period, with providers given until December 2, 2026, to comply with the machine-readable marking obligation. This extension only applies to the technical aspect of marking synthetic content and does not delay disclosures related to chatbots, deepfakes, or public-interest texts. Furthermore, the legislation does not mandate labeling of content created before the transparency rules came into effect.