The European Commission has published new guidelines to help organisations meet transparency rules under the EU’s AI Act that start applying on 2 August 2026.
The guidance is aimed at both “providers” that build artificial intelligence systems and “deployers” that use them, setting out when users must be told they are interacting with AI and when content has been generated or altered by AI, the Commission explained in a statement on Monday.
Under the AI Act, providers will have to design AI systems so they inform users when they are directly interacting with an AI system, and add machine-readable marks so AI-generated or manipulated content can be detected.
Deployers will have to inform users when they are exposed to “deep fakes” — AI-generated or altered media that appears real — as well as when they see AI-generated content on matters of public interest without human review or editorial control.
The rules also cover situations where people are exposed to emotion recognition systems or biometric categorisation systems.
What the guidelines cover
The Commission said the guidelines clarify which organisations must comply, explain key concepts and set out exemptions and examples.
Examples of “directly interactive” AI systems include chatbots, while “synthetic content” includes partially or fully AI-generated text.
An example of an exception is standard editing such as spelling and grammar correction.
The guidelines also set out ways to demonstrate compliance, including following a code of practice, which the Commission said offers a “simple and practical” route for organisations.
The guidelines are intended to complement a separate Code of Practice on Transparency of AI-generated content, which was drafted by independent experts with input from hundreds of stakeholders.
The Commission and the AI Board have confirmed the code is a voluntary way providers and deployers can use to demonstrate compliance with the AI Act.
Most rules in the AI Act, including enforcement powers for the Commission and national market surveillance authorities, start to apply on 2 August 2026.
AI systems placed on the market before August 2026 will have to comply with marking and detection obligations from 2 December 2026.
