The European Commission has published guidelines explaining how providers and deployers of artificial intelligence systems should comply with transparency requirements under the EU AI Act. The obligations will apply from 2 August 2026. AI providers will be required to design systems so that users are informed when they are interacting directly with AI. They must also apply machine-readable markings to AI-generated or manipulated content to support its detection.
Organisations deploying AI systems will be required to notify people when they are exposed to deepfakes, emotion-recognition systems or biometric-categorisation tools. Disclosure will also be required for AI-generated content concerning matters of public interest when that content has not undergone human review or editorial control.
The guidelines clarify which systems and organisations fall within the requirements and provide examples covering chatbots, AI agents, synthetic content, deepfakes and partially or fully AI-generated text. They also describe exceptions for standard editing functions, including spelling and grammar corrections.
Providers and deployers may demonstrate compliance through measures including adherence to the voluntary Code of Practice on Transparency of AI-Generated Content. The Commission and the European AI Board have determined that the code can be used as a means of showing compliance with the relevant provisions of the AI Act.
The code was prepared by independent experts following input from stakeholders. It forms part of a broader set of Commission resources that includes the new guidelines and the AI Act Service Desk, an information service intended to help organisations understand their regulatory obligations.
Most AI Act rules, including enforcement powers for the Commission and national market-surveillance authorities, take effect on 2 August 2026. AI systems placed on the market before that date must comply with marking and detection requirements from 2 December 2026.
