Skip to main content
    KI-Beratung Deutschland
    🇩🇪DE🇬🇧EN
    AI regulation

    AI labelling requirements: what applies from 2 December 2026

    Christoph Herting7 min read

    The transparency obligations in Article 50 of the EU AI Act have applied since 2 August 2026. A transition period for labelling generated content was provided for AI systems already on the market before that date. It ends on 2 December 2026, when two new prohibitions also take effect. Here is who is affected and what companies should check. Information as of 15 November 2026. Source: European Commission timeline.

    What ends on 2 December 2026

    Article 50(2) requires providers of AI systems generating text, images, audio or video to mark outputs in a machine-readable format. The content must be identifiable as artificially generated or manipulated. This also applies to general-purpose AI systems.

    For systems placed on the market before 2 August 2026, the transition period ends on 2 December 2026. Systems introduced from 2 August 2026 must meet the obligation from the outset.

    The obligation falls on the provider: the company developing and offering an AI system under its own name. It does not apply where a system only assists with standard editing or does not substantially alter the input.

    What companies using AI tools need to consider

    Most companies are deployers, not providers: they use AI tools made by others. Article 50(4) has applied to deployers since 2 August 2026.

    Anyone creating a deepfake with AI, meaning realistic images, audio or video of people, places or events, must disclose that the content is artificially generated or manipulated. For clearly artistic, satirical or fictional works, an appropriate notice that does not disrupt the work is sufficient.

    AI-generated text published to inform the public on matters of public interest must also be labelled. This does not apply where a human has reviewed the text and a person or company takes editorial responsibility.

    Chatbots and phone assistants: disclose the use of AI

    Under Article 50(1), AI systems communicating directly with people must be designed so people are informed that they are speaking or writing to AI. The exception is where this is already obvious to a reasonably attentive person in the circumstances.

    For a website chatbot or AI phone assistant, a clear notice at the start of the conversation is the prudent approach. Review that notice in existing systems now, rather than waiting for the next change.

    Two new prohibitions from 2 December 2026

    From 2 December 2026, AI systems generating sexual deepfakes of people without their consent or depictions of child sexual abuse are additionally prohibited. Regulation (EU) 2026/1744 added these prohibitions to the AI Act.

    This will not change day-to-day work for most companies. However, companies offering image or video generators, or integrating them into products, should obtain written confirmation of their provider's safeguards.

    What to check by 2 December

    First, list all AI systems you use or offer. Second, clarify your role as provider or deployer for each system. Third, check machine-readable labelling for your own content-generating systems placed on the market before 2 August 2026.

    Fourth, add an AI disclosure to chatbots and phone assistants. Fifth, define who reviews AI-generated images, videos and text before publication and when a notice is necessary.

    Sixth, train employees who work with AI. Article 4 requires providers and deployers to take measures supporting staff AI literacy. Documented AI training is a suitable measure.

    Who is responsible in Germany

    The AI Market Surveillance and Innovation Promotion Act (KI-MIG) regulates supervision. Under section 2, the Federal Network Agency (Bundesnetzagentur) is generally the competent market surveillance authority. Specific responsibilities remain, including for regulated financial activities. Source: KI-MIG, Federal Law Gazette 2026 I No. 223.

    What follows in 2027

    The obligations for high-risk AI under Annex III apply from 2 December 2027, including in candidate selection and creditworthiness assessments. High-risk systems in products under Annex I follow on 2 August 2028.

    See EU AI Act 2027: deadlines and obligations for companies for next year's deadlines and obligations.

    Frequently asked questions about AI labelling

    Do I have to label every AI-generated text? No. As a deployer, you label deepfakes and published texts on matters of public interest. For those texts, the obligation does not apply if a human has reviewed them and someone takes editorial responsibility.

    What does machine-readable labelling mean? Software must be able to read the marking and identify the content as AI-generated. The obligation falls on the AI system provider, not a company using someone else's tool.

    Does this apply to our website chatbot? Yes, regarding the AI disclosure under Article 50(1), which has applied since 2 August 2026.

    Who monitors compliance in Germany? Generally, the Federal Network Agency as market surveillance authority under the KI-MIG.

    Support with implementation

    In an AI compliance assessment, we inventory your AI systems, allocate roles and obligations and define next steps. The assessment costs from €4,990 plus VAT. Case-specific legal review remains with your legal adviser. For an initial assessment, book a free initial consultation.

    Sources

    European Commission AI Act Service Desk: implementation timeline, Article 50, Regulation (EU) 2026/1744 and KI-MIG. Accessed on 15 November 2026. This article does not replace legal advice.