BRUSSELS, BELGIUM / RankWire.AI / – Effective from August 2, 2026, the European Union has implemented new transparency regulations for artificial intelligence. Article 50 of the EU AI Act sets out requirements for how organizations must identify specific AI-driven interactions and synthetic content, covering chatbots, deepfakes, generated media, and certain public-interest texts. However, the law does not mandate visible labels on all AI-created content; instead, each obligation depends on the system, its output, and how an organization presents it.
Companies utilizing interactive AI technologies are required to inform users when they are engaging with artificial intelligence. This requirement does not apply when a typical user can easily recognize the system’s artificial origin. Generative AI providers must also embed machine-readable data in synthetic texts, images, audio, and video to facilitate automated detection of altered or generated content. Such measures should be implemented when current technology can provide reliable and effective marking.
The distribution of synthetic media by organizations is subject to distinct public disclosure rules. They are obligated to label deepfake images, audio, or videos that might be mistaken for genuine. Similarly, AI-generated texts discussing public interest topics must be marked in specific cases, unless the material is reviewed and editorial control is exercised by individuals or organizations responsible for publication. Responsibility for the content must be clearly attributed before exemptions apply.
Focus on Deepfakes and Civic Information
The regulations extend to emotion recognition and biometric categorization systems used to analyze individuals, requiring operators to notify those affected unless legally authorized exceptions apply. The EU AI Act provides more lenient disclosure rules for creative, fictional, artistic, and satirical works, as long as notices do not disrupt the typical presentation or enjoyment of those works. Nonetheless, clear disclosures are still necessary if the work contains deepfake content.
The European Commission has issued guidance clarifying the scope of Article 50 and its exceptions. It also published a voluntary code for companies involved in developing or deploying generative AI, allowing participants to demonstrate how their labeling and marking practices meet legal standards. Companies opting out must rely on other effective compliance strategies. While optional icons can aid public awareness, they do not automatically signify compliance.
Enforcement and Penalties by Regulators
Enforcement will primarily be led by national market surveillance authorities across the EU, with the AI Office overseeing a select group of systems related to general-purpose AI models and key platforms. The European Data Protection Supervisor will supervise systems used by EU institutions and agencies. Violations can lead to fines of up to 15 million euros, or penalties amounting to 3% of a company’s global annual turnover.
Existing generative AI systems on the market before August 2 benefit from a limited transition period, with providers given until December 2, 2026, to comply with the machine-readable marking requirement. This extension pertains only to the technical aspect of detectable synthetic content and does not delay notices related to chatbots, deepfake disclosures, or rules involving public-interest content. The law also clarifies that organizations are not required to label material produced before the transparency rules came into effect.