The European Union’s landmark Artificial Intelligence Act reached a major milestone on August 2 as Article 50 transparency obligations took effect across all 27 member states. The rules require providers and deployers of certain AI systems to inform users when they are interacting with AI rather than a human being.
Under the new framework, chatbots and interactive AI systems must clearly disclose their machine origins before or during a conversation. Deepfakes, including AI-generated or manipulated images, videos, and audio, must carry visible labels and machine-readable marks to help viewers identify synthetic content.
The European Commission’s AI Office, working alongside national enforcement authorities, will oversee compliance. Non-compliance carries penalties of up to 15 million euros or 3% of a company’s total worldwide annual turnover, whichever is higher. Small and medium-sized enterprises face a lower threshold.
Article 50 also targets AI-generated text published to inform the public on matters of interest. News articles and other informational content produced by AI must be clearly labelled as such, a provision designed to protect the integrity of the information ecosystem against deception and manipulation.
The transparency rules complement earlier provisions of the AI Act that entered into force in 2025, which classified AI systems by risk level and imposed stricter requirements on high-risk applications in areas such as law enforcement, education, and critical infrastructure.
Industry groups have raised concerns about compliance costs, particularly for smaller firms deploying generative AI tools. The Commission published a Code of Practice on transparency of AI-generated content to guide companies through the transition, though critics argue the guidance remains vague on edge cases.
Privacy advocates have welcomed the rules but warn that enforcement will be the real test. Several EU member states are still building the national regulatory bodies needed to investigate complaints and impose fines, raising questions about whether the rules will have teeth from day one.
The AI Act is the world’s first comprehensive legal framework for artificial intelligence. Its phased rollout continues through 2027, with requirements for high-risk AI systems set to apply next. The August 2 transparency provisions represent the first direct obligations felt by everyday AI users and developers across Europe.
Sources: European Commission, EU AI Act Portal, IBTimes UK
discussion