EU AI Act Transparency Rules Start August 2: New Commission Guidelines on Chatbot Disclosure and AI Content Labeling

The European Commission published guidelines on July 20, 2026 to help providers and deployers of AI systems comply with the transparency obligations of the EU’s AI Act¹. The guidelines cover interactive AI systems such as chatbots, and the marking and labeling of AI-generated content — the obligations set out in Article 50 of the regulation.

These transparency obligations start to apply on August 2, 2026¹². With the deadline only two weeks away, the guidelines give companies offering AI services in the EU a concrete reference for what they need to do, and by when.

What the Guidelines Require

The Commission explains the purpose of the transparency obligations as helping people recognize when they are interacting with AI or when content has been generated or altered by AI, reducing the risk of deception and manipulation¹. The obligations are split between “providers,” who develop and supply AI systems, and “deployers,” who use them in their operations.

Providers: Design for Disclosure, Add Machine-Readable Marks

Providers of AI systems must design them so that users are informed when they are interacting directly with an AI system¹. Systems that converse with people — chatbots and AI agents — are the typical targets.

Providers must also add machine-readable marks to AI-generated or manipulated content so that it can be detected¹. Notably, this goes beyond human-visible labels: the requirement assumes marking in a format that software can automatically identify.

Deployers: Disclose Deepfakes and More

Deployers of AI systems must inform people in cases such as the following¹:

When publishing deepfakes (AI-generated or manipulated images, audio, or video)
When publishing AI-generated text on matters of public interest without human review
When people are subject to emotion recognition systems or biometric categorization

Henna Virkkunen, the Commission’s Executive Vice-President for Technological Sovereignty, said: “These guidelines help providers and deployers to comply with their obligations under the AI Act, while also helping citizens to know when they are interacting with AI.”²

Timeline: August 2 Start, December 2 for Existing Systems

The transparency obligations apply from August 2, 2026, and from that date the Commission and national authorities gain the power to impose sanctions for violations². Systems placed on the market before August 2 have until December 2, 2026 to comply with the labeling and related requirements².

Alongside the guidelines, the Commission is offering tools to support compliance. The Code of Practice on Transparency of AI-Generated Content is a voluntary framework, but signing it can serve as evidence of compliance¹². The Commission also provides the AI Act Support Service, a digital helpdesk for businesses and public bodies, and common EU icons for labeling AI-generated content². A Q&A on the Article 50 obligations and a fact page have been published as well¹.

After the Code of Practice, the Interpretive Guidelines Complete the Set

The AI Act is entering into force in stages, and the transparency obligations are the next milestone. As groundwork for these guidelines, the Commission had been preparing a code of practice for labeling AI-generated content in June 2026, which we covered in our article on the EU code of practice. With this publication, both the voluntary code and the official guidelines interpreting the obligations are now in place.

The push for AI transparency is not limited to the EU. Google has started disclosing which ads are made with AI across Search and YouTube, Illinois enacted the first US law requiring annual third-party AI audits, and Australia announced its national AI framework, the Australian Standards for AI. Rules that make AI use visible and verifiable are advancing in parallel across governments and platforms.

Serving EU Users? Check Your Scope Now

The obligations are not limited to companies based in the EU — they concern anyone offering AI systems or AI-generated content to users in the EU. If your service includes a chatbot for EU users or distributes AI-generated images or text there, you likely need to check whether you are in scope.

Three questions are a practical starting point: (1) Does your conversational AI clearly tell users it is an AI? (2) Can you embed machine-readable marks in AI-generated or altered content? (3) Do any of your use cases involve deepfakes, emotion recognition, or biometric categorization? For borderline cases, the published guidelines, the Q&A, and the AI Act Support Service are the first places to consult¹. With less than two weeks before the rules apply, companies serving EU users should take stock now.