EU AI Content Transparency Code
Code of Practice on Transparency of AI-Generated Content (EU AI Act Article 50)
Voluntary code of practice under the EU AI Act that sets out how providers of generative AI systems can meet the Article 50(2) duty to mark AI-generated or manipulated content in a machine-readable, detectable way, and how deployers can meet the Article 50(4) duty to label deepfakes and AI-generated text published on matters of public interest. The Commission and the AI Board have assessed it as an adequate means of demonstrating compliance.
Jurisdiction
European Union
Enacted
Jun 10, 2026
Effective
Aug 2, 2026
Enforcement
None for the code itself (voluntary). The underlying Article 50 obligations are enforced by national market surveillance authorities and, for systems based on general-purpose AI models from the same provider, the AI Office.
Final code published by the European Commission on June 10, 2026, after a multi-stakeholder drafting process facilitated by the AI Office with two working groups. On July 8, 2026 the Commission concluded that the code adequately covers the obligations in Articles 50(2), (4) and (5) of the AI Act, and on July 9, 2026 the AI Board adopted its adequacy assessment. The Article 50 obligations it supports apply from August 2, 2026. About 190 companies and organisations had signed by the end of July 2026.
European Commission, Shaping Europe's digital future: Code of Practice on Transparency of AI-generated Content (with code PDF and Commission opinion)Why It Matters
The EU-wide practical framework for AI-content marking and labelling under the AI Act, playing the role for Article 50 that the General-Purpose AI Code of Practice plays for GPAI model obligations. Signatories can rely on its measures to demonstrate compliance in every Member State, though the Commission states adherence is not conclusive evidence of compliance; non-signatories must show their own measures are adequate, assessed case by case by national market surveillance authorities.
Recent Developments
Published June 10, 2026 and assessed as adequate by the Commission (July 8, 2026) and the AI Board (July 9, 2026), ahead of the August 2, 2026 application date of Article 50. The Commission also published guidelines on the scope of the Article 50 transparency obligations and a set of EU icons deployers may use for labelling. The AI Office will consider facilitating formal updates to the code at least every two years.
At a Glance
Applies to
Harms addressed
Requires
Who Must Comply
- Providers of generative AI systems, including general-purpose AI systems, that sign the code
- Deployers of AI systems generating deepfakes or text published to inform the public on matters of public interest that sign the code
Safety Provisions
- Providers: machine-readable marking of AI-generated or manipulated audio, image, video and text output, with non-removal of markings
- Providers: detection mechanisms for their markings and clear, accessible disclosure of detection results
- Providers: marking and detection solutions must be effective, reliable, robust and interoperable as far as technically feasible
- Providers: compliance processes, testing, verification, monitoring, training and cooperation with market surveillance authorities
- Deployers: design and placement specifications for disclosing deepfakes and AI-generated text on matters of public interest, with optional EU icons
- Deployers: internal compliance processes, awareness and literacy, and review and cooperation with authorities
- Deployers: tailored disclosure for artistic, creative and similar works, and an exception path for published text under human review and editorial responsibility
Exemptions
Human review and editorial responsibility
AI-generated or manipulated text published on matters of public interest need not be labelled where it has undergone human review or editorial control and a natural or legal person holds editorial responsibility
- • Human review or editorial control
- • Editorial responsibility held by a person
Artistic, creative, satirical or fictional works
Deepfake disclosure for evidently artistic, creative, satirical, fictional or analogous works may be made in a manner that does not hamper display or enjoyment of the work
- • Evidently artistic, creative, satirical or fictional work
Compliance & Enforcement
Key Dates
Aug 2, 2026
AI Act Article 50 transparency obligations (marking and detection, deepfake and public-interest text labelling) apply; signatories may rely on the code to demonstrate compliance
Penalties
None for non-adherence to the code. Breach of the underlying Article 50 obligations falls under Article 99 of the AI Act, up to EUR 15 million or 3% of worldwide annual turnover, whichever is higher.
View on map
European Union
Focus Areas
Cite This
APA
European Union. (2026). Code of Practice on Transparency of AI-Generated Content (EU AI Act Article 50).
Related Regulations
EU Digital Omnibus (AI)
Amendments to the EU AI Act (Regulation (EU) 2026/1744, in force July 27, 2026) that delay high-risk AI system obligations by up to 16 months, make compliance timing conditional on availability of harmonised standards and support tools, and add a new Article 5 prohibition on AI systems for generating non-consensual intimate imagery (nudification tools) and CSAM.
EU CRA
Mandatory cybersecurity requirements for all products with digital elements placed on the EU market, including AI software. Requires security by design, vulnerability handling, incident reporting to ENISA, software bills of materials, and CE marking for market access.
Italy AI Act
First EU member state comprehensive national AI law complementing the EU AI Act. 28 articles covering AI governance principles, sector-specific rules for healthcare, employment, justice, and public administration, criminal provisions, copyright protections, and a EUR 1 billion AI investment fund.
DE Digital Violence Bill (GgdG)
German government bill creating a standalone Digital Violence Act (GgdG) that gives victims court-ordered disclosure of anonymous offenders' identities, data preservation, content removal and temporary account suspension against platforms, and adding Criminal Code offences for sexualised deepfakes, reputation-damaging deepfakes, identity misuse and covert tracking.
Singapore OSRA
Creates a dedicated Online Safety Commission (OSC) with powers to order takedowns and disable access, establishes statutory torts providing victims of online harms with direct civil remedies against platforms and perpetrators.
SD Deepfakes Act
Prohibits disseminating deepfakes about candidates within 90 days of election with intent to cause injury. Class 1 misdemeanor with up to 1 year imprisonment and $2,000 fine. Affirmative defense for content with AI manipulation disclosure. Civil remedies available to AG, candidates, and depicted individuals.
Last updated October 4, 2026. Verify against primary sources before relying on this information.