EU Commission publishes Code of Practice for labelling AI-generated content - deepfakes and AI text must be marked from August 2026
What it really says
On June 10, 2026, the European Commission published the final Code of Practice for marking and labelling AI-generated content. The code implements Article 50 of the EU AI Act and contains concrete practical measures that providers and deployers of generative AI systems should follow. From August 2, 2026, the corresponding transparency obligations of the AI Act become legally binding. The code prescribes two technical mechanisms: digitally signed metadata (based on the C2PA standard) and imperceptible watermarks (such as Google's SynthID). Different labelling requirements apply to different media types: AI-generated or AI-manipulated videos and photos must be visibly marked as 'AI-generated' or 'AI-edited,' with labels at the beginning of videos and at regular intervals. AI-generated audio content such as podcasts must include a spoken label at the start. For texts, an AI label should appear above the text or near the headline. Deepfakes and AI-generated text on matters of public interest must be clearly labelled. The list of signatories will be published in July 2026. Those who sign the code can use it to demonstrate compliance with AI Act obligations.
Our assessment
This Code of Practice is an important and concrete step in combating the spread of unlabelled AI content. For people worried about deepfakes and manipulated content, this is good news: the EU is establishing clear rules effective from August 2026. The technical approach using invisible watermarks and metadata is particularly positive - these are harder to remove than visible labels and enable automatic detection. However, the code itself is voluntary - though the underlying AI Act obligations are binding. The real challenge lies in enforcement: criminals and malicious actors will not comply with labelling requirements. The code primarily targets reputable providers like OpenAI, Google, or Adobe. Against deliberately spread deepfakes for disinformation, it helps only indirectly by standardizing detection tools. Nevertheless, a unified EU standard for labelling represents clear progress compared to the previous fragmented landscape.
Relevance for Germany
This Code of Practice is directly relevant for Germany. From August 2026, German users will see standardized labelling on AI-generated content - whether images, videos, audio, or text. This affects social media, news platforms, and AI chatbots equally. Given the ongoing debate about disinformation in Germany, the deepfake labelling requirement is particularly significant. The EU Commission's Representation in Germany explicitly presented the code. For German companies that deploy or offer generative AI systems, the code means concrete compliance requirements: they must implement technical marking mechanisms before obligations take effect on August 2. The C2PA standard and watermarking technologies like SynthID will become the new industry standard in the German market as well.
Fact check
The publication of the Code of Practice on June 10, 2026 is documented by the official EU Commission press release. The EU Commission's Representation in Germany separately presented the code in German. The technical requirements (C2PA metadata, watermarks) and labelling obligations for different media types are consistently described in official documentation and IPTC reporting. The date for binding applicability of Article 50 obligations on August 2, 2026 follows from the official AI Act timeline. The FAQ page confirms that the signatory list will be published in July 2026.
Source
- • https://digital-strategy.ec.europa.eu/en/news/commission-publishes-code-practice-marking-and-labelling-ai-generated-content
- • https://germany.representation.ec.europa.eu/news/kennzeichnung-von-ki-generierten-inhalten-eu-kommission-veroffentlicht-verhaltenskodex-2026-06-10_de
- • https://iptc.org/news/eu-ai-transparency-code-of-practice-june-2026/
- • https://digital-strategy.ec.europa.eu/en/faqs/signing-code-practice-transparency-ai-generated-content