(4 min read)
The EU AI Act’s transparency obligations for providers and deployers of generative AI become applicable on 2 August 2026, although the AI Omnibus has introduced a grace period until 2 December 2026 for systems placed on the market or put into service before 2 August 2026. The European Commission has recently published a voluntary Code of Practice on marking and labelling AI-generated content, which sets out practical steps to help organisations meet their transparency obligations. Read our overview of the key points.
The European Commission has recently published a voluntary Code of Practice on marking and labelling AI-generated content, which sets out practical steps to help providers and deployers of generative AI systems meet their transparency obligations under Article 50 of the EU AI Act. The obligations become applicable on 2 August 2026, but the AI Omnibus has introduced a grace period until 2 December 2026 for systems placed on the market or put into service before 2 August 2026.
The Code of Practice is divided into two sections plus an Annex, with each section setting out four commitments and measures to ensure compliance with each commitment.
Section 1: Providers - Rules for marking and detection of AI-generated and manipulated content
- Commitment 1: Marking of AI-generated or manipulated content – this sets out how to comply with the obligation on providers of AI systems generating synthetic audio, image, video or text content to ensure that the system’s outputs are marked in a machine-readable format.
- Commitment 2: Detection of markings of AI-generated or manipulated content – this covers compliance with the obligations to ensure that the system’s outputs are detectable as artificially generated or manipulated and provide the required information in a clear and distinguishable manner.
- Commitment 3: Measures to meet the requirements for marking and detection solutions – this addresses the obligation to ensure that such solutions are effective, interoperable, robust and reliable.
- Commitment 4: Testing, verification and compliance – this explains how to fulfil and demonstrate compliance with their transparency obligations.
Section 2: Deployers - Rules for labelling of deepfakes and AI-generated and manipulated text
- Commitment 1: Disclosure of deepfakes and published text – this covers the requirement for deployers of an AI system that generates or manipulates image, audio or video content constituting a deep fake to disclose that the content has been artificially generated or manipulated.
- Commitment 2: Internal processes – this sets out processes to enable deployers to effectively fulfil and demonstrate compliance with their transparency obligations.
- Commitment 3: Disclosure for artistic, creative and similar works – this explains how the disclosure obligation applies to deepfakes that form part of evidently artistic, creative, satirical, fictional or analogous work.
- Commitment 4: Human review and editorial control for published text – this provides that media service providers under the European Media Freedom Act can apply their existing procedures and professional standards. Other deployers must implement appropriate policies.
Annex 1: Publicly available EU icons
Annex 1 sets out a set of EU icons that deployers of generative AI systems may use to comply with their obligation to label their AI-generated or AI-modified content clearly. The icons are available in slightly different variations, but the core icons are:
- A basic icon for use when AI was involved in the creation of deepfake content or published text or when a custom text label or interactive second layer is implemented, eg a video labelled with the text “voices generated by” plus the icon
- An AI-generated icon for use when the entire deepfake content or text is fully generated by AI with no human-created elements or human editorial control (apart from prompting), eg fully AI-generated videos and fully AI-composed art
- A partially AI-modified icon for use when pre-existing, human-made content is partially modified with AI turning it into a deep fake or text on matters of public interest, eg a photograph is modified by changing the subject’s face or adding content
Adequacy assessment and Commission guidelines
The Code is currently undergoing an adequacy assessment by the AI Office and the AI Board. The Commission will publish complementary guidelines to:
- Address the legal interpretation of Article 50 of the AI Act and aspects not covered by the measures envisaged in the Code
- Clarify the scope and application of the transparency obligations under Article 50 of the AI Act, including:
- Which providers, deployers and AI systems are covered?
- What types of AI-generated or manipulated content fall within the scope of the rules?
- How should the obligations be applied in practice, including for AI interaction, machine-readable marking, deepfakes and AI-generated text published on matters of public interest?
Signature of the Code
Once the Code has received a positive assessment, providers and deployers who sign it can rely on its measures to demonstrate compliance with the relevant provisions of the AI Act. The AI Office encourages them to sign by 22 July 2026, in which case they will be included in the published list of initial signatories.
Providers and deployers who do not sign will need to demonstrate that the measures they use to achieve compliance are adequate, which will require individual assessment by market surveillance authorities.
