The Code of Practice on Transparency of AI-generated content was published in June 2026. Providers and deployers can sign to as a way to comply with the obligations for transparency of AI-generated content.
Who is eligible to sign up to the code?
Providers and deployers of generative AI systems subject to the obligations laid down in Article 50(2) and/or Article 50(4) of the AI Act may sign up to the Code of Practice on Transparency of AI-Generated Content. This includes:
- Providers of generative AI systems who develop, or have developed, an AI system capable of generating synthetic audio, image, video or text content. This includes but is not limited to general-purpose AI systems, and existing or planned operations to place such AI systems on the EU market or put them into service under their own name or trademark, whether for payment or free of charge.
- Deployers of generative AI systems who are subject to Article 50(4) of the AI Act, in particular those who use generative AI systems under their authority for professional purposes and are affected by the obligations concerning the disclosure of deepfakes or AI-generated or manipulated text published with the purpose of informing the public on matters of public interest.
While not directly bound by the obligation under Article 50(2) of the AI Act, Section 1 of the Code may also be signed by:
- Providers of generative AI models, who develop or have developed an AI system that is capable of generating synthetic audio, image, video or text content. This includes but it is not limited to general-purpose AI systems and existing or planned operations to place such AI system on the EU market or put it into service under their own trademark, whether for payment or free of charge.
- Technology providers of marking and detection solutions, whodevelop or have developed technical tools, services or infrastructure for the marking, provenance, watermarking and/or detection of AI-generated or manipulated content and place such tools on the EU market, whether for payment or free of charge.
How does a provider or deployer sign up to the code?
Download the signature form (DOCX), complete it and submit it by 22 July 2026, 18:00 CEST, via the email: CNECT-AIOFFICE-CODE-OF-PRACTICE-TRANSPARENCY@ec.europa.eu. You can also read further instructions.
What is the deadline for submitting a signatory form?
To be included in the list of initial signatories that will be published before the AI Act’s general date of entry into application of 2 August 2026, signatories must submit their completed forms by 22 July 2026, 18:00 CEST.
Who should sign the code’s signature form?
The form should be signed by a senior executive with sufficient authority to bind the provider or deployer as a signatory to the Code of Practice on Transparency of AI-generated content.
Can a provider become a signatory to the code after 22 July 2026?
In principle, providers and deployers sign the code at any time, including after 22 July 2026, simply by submitting the signature form (DOCX) and submitting it via e-mail to CNECT-AIOFFICE-CODE-OF-PRACTICE-TRANSPARENCY@ec.europa.eu.
However, the AI Office encourages providers and deployers to sign the code until 22 July 2026. Article 50(2), (4) and (5) of the AI Act become applicable from 2 August 2026.
If a provider or a deployer does not sign the code, after 2 August 2026 they will need to demonstrate compliance through other means. Their adequacy will be assessed by the competent market surveillance authorities.
Why are assessments of the Commission and Board necessary?
Article 50(7) of the AI Act mandates the AI Office to facilitate the drawing up of EU codes of practice to support the effective implementation of the obligations regarding the detection and labelling of artificially generated or manipulated content. That provision also states that the Commission may adopt implementing acts to approve such codes in accordance with the procedure laid down in Article 56(6) of the AI Act.
In accordance with Article 56(6) of the AI Act, the AI Office and the AI Board shall assess the adequacy of codes of practice and publish their assessments. The commission may also approve a code of practice with an implementing act, giving it general validity within the EU.
The Commission and AI Board assessments will establish:
- whether adherence to the code adequately ensures compliance with the transparency obligations for AI-generated content under Article 50(2), (4) and (5) of the AI Act
- and whether providers and deployers can rely on the code to demonstrate compliance with those obligations.
Without such an assessment, adherence to the code will not have the same legal and supervisory value to demonstrate compliance with those obligations. Such an assessment also promotes legal certainty and ensures consistent and predictable implementation of those obligations throughout the EU. This is the case, regardless of the place of establishment and operation of the provider or the deployer and the competent supervisory authority.
The procedure for the adequacy assessment of codes of practice has been amended by the AI Omnibus. Under that procedure, a single adequacy assessment with an opinion of the Commission after consultation of the AI Board will produce the same legal effects.
Why are signatures requested before the adequacy assessment from the Commission and the AI Board?
The signatures are conditional on a positive adequacy assessments by the Commission and the AI Board, as indicated in the Signature Form.
Can an organisation choose which commitments they would like to sign up to?
The code is structured in two separate sections, reflecting the respective transparency obligations of providers and deployers of generative AI systems under Article 50 of the AI Act. Separate sections of the code may be signed individually, while individual commitments may not.
Signatories may sign up to the relevant one or the two sections, depending on their role and the obligations applicable to them. Each section of the code should be signed as a whole, since they contain all the commitments with which the respective provider or deployer needs to comply.
Providers of generative AI systems within the scope of Article 50(2) of the AI Act are invited to sign up to section 1 of the code. It concerns the marking of AI-generated or manipulated content in a machine-readable format and related detection mechanisms.
Deployers of generative AI systems within the scope of Article 50(4) of the AI Act are invited to sign up to section 2 of the code. It refers to the disclosure and labelling of deepfakes and certain AI-generated or manipulated text published with the purpose of informing the public on matters of public interest.
An organisation acting both as a provider and a deployer of in-scope generative AI systems is invited to sign up to the entire code (both Section 1 and Section 2).
What are the benefits of signing up to the code?
The code contains practical measures that providers and deployers may rely on to show how they meet their obligations related to marking and labelling of AI-generated content.
Signing the code gives organisations a structured and recognised framework for effective implementation of the transparency obligations, reducing uncertainty about expected practices.
The code helps to avoid fragmented or inconsistent approaches by setting out common practices for marking, detection and labelling of AI-generated or manipulated content.
Signatories will also benefit from clarity and increased legal certainty across the EU. Subject to a positive assessment by the Commission and the AI Board, signatories may rely on the code to demonstrate compliance with their obligations under Article 50(2),(3) and (5) of the AI Act, regardless of their place of establishment and operation and competent market surveillance authority.
Signatories that adhere to the code – deemed adequate by the Commission and AI Board – also demonstrate transparency and gain public trust in compliance with obligations for AI-generated content in the AI Act. This helps to reduce risks of deception and manipulation in the information ecosystem.
What are the consequences of not signing the code? how do non-signatories demonstrate compliance?
Signing the code of practice is voluntary. Not signing it does not constitute non-compliance with the AI Act.Any enforcement consequences would relate to non-compliance with obligations in Article 50 of the AI Act, not the absence of adherence to the code.
Non-signatories remain responsible for complying with the transparency obligations under Article 50 of the AI Act and must be prepared to demonstrate compliance through other adequate means.
To monitor their compliance, competent authorities will likely need more detailed information. Non-signatories should be able to explain and document how their chosen measures ensure compliance with the obligations laid down in Article 50 of the AI Act.
Non-signatories may have to carry out a gap analysis, comparing their measures with those set out in the code.
Non-signatory providers and deployers within the scope of Article 50(2) and (4) may be subject to a larger number of requests for information or access from competent authorities with regard to their marking and labelling practices under Article 50(2) and (4) of the AI Act, respectively.
What if it is hard to comply with the AI Act given the code is published only shortly ahead of August 2026?
Article 50 of the AI Act applies from 2 August 2026. From that date, all AI systems falling within the scope of that provision, placed on the EU market or put into service in the EU, must comply, regardless of when they were first placed on the market or put into service. The timing of publication of the code does not change the date of application laid down in the AI Act itself.
However, two points should be considered:
- The AI Omnibus proposal agreed by the EU co-legislators, envisages a targeted grandfathering rule for the marking and detection obligations laid down in Article 50(2) for generative AI systems placed on the market or put into service before 2 August 2026. If adopted, this would allow a transitional period for providers of those existing systems to bring them into conformity by 2 December 2026.
- AI-generated or manipulated outputs, including deep fakes falling within the scope of Article 50(2) and (4) of the AI Act which were generated and already made available before 2 August 2026, do not need to be marked or labelled retroactively.
Can a signatory withdraw their signature?
Signatures can be withdrawn. However, the benefits of signing up and adhering to the code would no longer apply to the provider and/or deployer. To withdraw a signature, a letter signed by an executive at the same level as the one who signed up to the code is necessary.
This letter should be sent by email to CNECT-AIOFFICE-CODE-OF-PRACTICE-TRANSPARENCY@ec.europa.eu, with CNECT-AIOFFICE@ec.europa.eu in copy.
Related content
This code of practice supports compliance with the AI Act transparency obligations related to marking and labelling of AI-generated content.