The European Union’s most visible transparency provisions under the Artificial Intelligence Act began applying on August 2, 2026. The European Commission said its AI Office and competent national authorities would begin enforcing the law from the same date. This phase focuses on ensuring that people know when they are dealing with an AI system and can recognize content that has been generated or altered by artificial intelligence.
What must users be told?
Under the Commission’s July 31 announcement, chatbots and other systems designed for direct interaction must inform people that they are communicating with AI rather than a human. The disclosure should appear clearly and distinctly from the start of the first interaction. Systems that operate only in the background, communicate from machine to machine, or do not directly contact people can fall outside this specific requirement.
The rules also seek to make generative AI output detectable. Providers of systems that create synthetic audio, images, video or text must add effective, reliable and interoperable machine-readable marks to their output. Artificially generated or manipulated images, sound and video that could appear authentic must also carry a disclosure that people can perceive. The Commission’s guidance says an embedded technical marker alone does not replace a visible or audible notice where a disclosure to the audience is required.
Where is the line for published text?
AI-generated or AI-manipulated text published to inform the public about matters of public interest can also be covered. The Commission lists politics, public services, justice, fundamental rights, public security, health, environmental protection, consumer safety and economic or scientific developments relevant to public debate among the examples. There is an exemption when text has undergone meaningful human review or editorial control and a publisher assumes legal responsibility. A superficial spelling or grammar check is not enough to qualify as human review.
Transition period and enforcement
Article 50’s transparency obligations apply from August 2. A limited transition period runs until December 2, 2026 for systems placed on the market before the August deadline, and only for the obligation to add machine-readable marking and detection features to AI-generated content. Material created before August 2 does not have to be labelled retroactively, although the Commission encourages voluntary labelling where it is practical.
National market-surveillance authorities will handle most enforcement. The AI Office has a narrower role for certain systems built on general-purpose AI models and for AI systems integrated into very large online platforms or search engines designated under the Digital Services Act. According to the Commission’s guidance, penalties can reach 15 million euros or 3 percent of a company’s total worldwide turnover in the preceding financial year. Proportionality may be considered for small and medium-sized companies.
Who falls within the rules?
The Commission’s Article 50 guidance defines a provider as a person, company or public body that develops an AI system, has one developed, and places it on the EU market or puts it into service under its own name. Being located outside the European Union does not by itself remove a provider from the law when the system’s output is used in the EU. Systems that interact directly with people must meet the relevant disclosure requirements before they are marketed or put into service.
A deployer is a person or organisation using an AI system under its authority for professional purposes; personal, non-professional use is excluded from that definition. A legal entity such as an advertising company remains the deployer when employees or contractors operate the system under its responsibility. Deployers must inform people exposed to emotion-recognition or biometric-categorisation systems and clearly label covered synthetic content in the situations described by Article 50.
The Commission has also endorsed a voluntary Code of Practice designed to help providers and deployers demonstrate compliance with the marking and labelling duties. More than 180 organisations had joined the code when the enforcement announcement was published. Companies that do not sign can use other adequate methods, but the Commission says they may face more requests for information because regulators have less visibility into how they meet the obligations.
