Strasbourg: The Commission opened a public consultation on 8 May 2026 on draft guidelines for AI transparency obligations under the AI Act, marking a critical step in the operationalisation of the regulation ahead of its full application on 2 August 2026. The guidelines address Article 50 obligations, which require providers and deployers of certain AI systems to inform users that they are interacting with an AI system and to mark synthetic content generated by AI.
The consultation comes at a moment of considerable political and technical pressure. The Council and Parliament reached political agreement on 7 May 2026 on amendments to the AI Act under the Digital Omnibus package, which extends selected compliance deadlines, sharpens the rules on AI-generated intimate content, and clarifies the relationship between AI Act obligations and sectoral safety legislation. The transparency guidelines therefore operate within a framework that has been adjusted weeks before its main entry into force.
The guidelines clarify the practical meaning of transparency obligations for chatbots, deepfake detection, AI-generated text on matters of public interest, and emotion recognition systems. For chatbots, the requirement is that users be clearly informed they are interacting with an AI system, except where it is obvious from the circumstances. For synthetic content — text, image, audio or video generated or substantially modified by an AI system — providers must implement reliable, robust and interoperable technical solutions to mark the content as artificial.
The watermarking deadline has been shortened from the original six months to three months under the Digital Omnibus, fixing a new date of 2 December 2026 for compliance. Under the grandfathering provisions agreed in the political compromise, generative AI systems placed on the market or put into service before 2 August 2026 must comply with watermarking requirements from 2 December 2026 onward. Penalties for non-compliance reach 15 million euros or 3 per cent of worldwide annual turnover, whichever is higher.
The Code of Practice for general-purpose AI, developed under the AI Office’s coordination, provides a structured framework for demonstrating compliance with Article 50 obligations. Providers may also choose alternative methods of demonstrating compliance, but the Code has been treated by enterprise users as a baseline reference. The Commission published its draft Code of Practice in parallel with the transparency guidelines, with both instruments now subject to stakeholder feedback before their finalisation.
The consultation invites input on operational aspects that have generated considerable industry concern: the technical robustness of available watermarking solutions, the interoperability between marking systems from different providers, the cost and feasibility for smaller developers, and the interaction between AI Act transparency and existing rules under the Digital Services Act, the General Data Protection Regulation and intellectual property law. The consultation closes in late summer, with the Commission expected to publish final guidelines before the December watermarking deadline.
The broader picture remains that the AI Act is the world’s first comprehensive horizontal regulation of artificial intelligence. The high-risk obligations, the prohibited practices regime, the obligations on general-purpose AI model providers, and the transparency rules together establish a regulatory perimeter that other major jurisdictions are observing closely. The transparency guidelines, although operationally narrower than other elements, will define the everyday experience of European users interacting with AI systems and will set the practical baseline for how the regulation manifests in consumer-facing applications.




