From the journal

European Commission Publishes Draft EU AI Act Transparency Guidelines on Deepfakes, EU, May 2026

On 8 May 2026, the European Commission published draft guidelines on the transparency obligations under Article 50 of the EU AI Act. The guidelines cover disclosure requirements for providers of AI chatbots, systems generating synthetic audio, image, video, or text, emotion recognition systems, and deployers of deepfake content. A public consultation closed on 3 June 2026. Final guidelines are expected by 2 August 2026, when the Article 50 obligations become applicable.

3 min read

On 8 May 2026, the European Commission published draft guidelines on the transparency obligations under Article 50 of Regulation (EU) 2024/1689, the EU AI Act. A public consultation on the draft closed on 3 June 2026. The guidelines are non-binding until adopted in final form. The Commission is expected to adopt final transparency guidelines before 2 August 2026, the date on which the Article 50 obligations become applicable across the EU.

Article 50 of the EU AI Act imposes four transparency obligations. Article 50(1) requires providers of AI systems that interact with natural persons to notify users they communicate with an AI, unless this is obvious from context. Article 50(2) requires providers of AI systems generating synthetic audio, image, video, or text to mark that output as AI-generated. Article 50(3) requires deployers of emotion recognition and biometric categorisation systems to notify persons within range. Article 50(4) requires deployers of deepfake content to disclose its AI-generated or AI-manipulated nature. An exception in Article 50(4) applies to evidently artistic, creative, satirical, or fictional content; disclosure is still required in those cases but must not hamper display or enjoyment of the work.

Providers of conversational AI products deployed in the EU must implement user-facing notifications before the Article 50(1) obligation becomes applicable on 2 August 2026. AI content generators, including marketing technology vendors producing synthetic images or video, must embed machine-readable labelling or provenance data in output to satisfy Article 50(2). Deepfake deployers, including advertisers using AI-generated likenesses, must disclose the AI-manipulated nature of content even where the depicted person is fictional and no deception is intended.

The draft guidelines confirm that the Article 50(1) chatbot notification obligation does not apply where the AI nature is obvious from context. The exception for artistic and creative deepfakes under Article 50(4) is narrowly drawn: the creative character must be evident to the ordinary viewer. National competent authorities retain discretion to assess adequacy of disclosure methods, and the final guidelines may refine the standards for machine-readable labelling under Article 50(2).

Licentium advises technology companies, advertisers, and content platforms on EU AI Act transparency compliance. If your products generate or deploy synthetic content or operate conversational AI in the EU, we can assist with compliance gap analysis and disclosure implementation. Work we undertake includes EU AI Act Article 50 compliance mapping, synthetic content labelling strategy, deepfake regulatory analysis, and public consultation submissions.

Source: European Commission, Draft Guidelines on Transparency Obligations under the EU AI Act (Article 50), 8 May 2026

AI Regulatory

More from the journal

See all

OCC Interpretive Letter 1192: National Bank Act Preempts State Money Transmitter Licensing for Digital Asset Activities, May 2026

The OCC published Interpretive Letter 1192 on May 12, 2026, confirming that the National Bank Act preempts state money transmitter licensing laws for national banks conducting digital asset activities. The Letter arose from Fidelity Digital Assets' request for confirmation that Iowa's money transmitter law does not apply to its national trust bank operations. The OCC held that state money transmitter laws requiring national banks to obtain state licenses, or vesting states with visitorial authority over national banks, are preempted.

House Ways and Means Committee Holds Digital Asset Tax Legislative Hearing, June 9, 2026

The House Ways and Means Committee held a full-committee legislative hearing on June 9, 2026, examining eight bills and discussion drafts on digital asset taxation. The hearing, the Committee's first legislative hearing on digital asset tax rules in years, covered proposals addressing mining and staking income classification, de minimis transaction reporting exemptions, charitable deduction rules for crypto donations, and extension of wash-sale and anti-abuse provisions to digital assets.

Trump Signs Executive Order Promoting AI Innovation and Security, June 2, 2026

President Trump signed the Executive Order 'Promoting Advanced Artificial Intelligence Innovation and Security' on June 2, 2026, directing federal agencies to establish a voluntary cybersecurity arrangement for frontier AI models and create an AI cybersecurity clearinghouse. The Order bars mandatory federal licensing or permitting requirements for AI model development and release, and directs the Attorney General to prioritize enforcement against AI-enabled criminal activity.

Ready to launch without the regulatory guesswork?

Book a 30-minute consultation. We'll map your AI or licensing path and tell you exactly what's required, in plain language.