EU AI Act: Companies must label AI content and deepfakes from August 2 - Code of Practice signing deadline July 22
What it really says
On August 2, 2026, the transparency obligations of the EU AI Act (Article 50) take effect - the first major enforcement wave of the AI regulation across all 27 member states. From this date, AI systems that interact with people, such as chatbots, must disclose that the user is communicating with an AI. Deepfakes - AI-generated or manipulated image, video, or audio content depicting real persons or events that could appear authentic - must be labeled as AI-produced, regardless of whether there is any intent to deceive. The European Commission has developed standardized labels: the official 'AI GENERATED' mark for fully AI-generated content and 'AI MODIFIED' for partially edited content, with localized variants including 'KI' for the German market. Before this enforcement date lies an important intermediate deadline: until July 22, 2026, at 18:00 CEST, providers and deployers of AI systems can sign the Code of Practice on Transparency of AI-Generated Content developed by the European Commission. Signatories receive a presumption of conformity - in case of a regulatory inspection, it is initially assumed that the company meets the transparency obligations, which significantly reduces enforcement risk and the burden of proof. The Code of Practice consists of two sections: Section 1 for providers of generative AI systems covers machine-readable marking and detection mechanisms, Section 2 for deployers covers visible disclosure to users. Both sections can be signed independently. Non-compliance with Article 50 carries fines of up to 15 million euros or 3 percent of global annual turnover, whichever is higher. The obligations apply not only to AI system manufacturers but to every company, self-employed professional, and public authority that uses AI professionally or provides it under its own name.
Our assessment
This development merits a yellow rating because it represents an important step forward in protecting against AI deception while also bringing significant practical challenges. The positive side: for the first time in any G7 jurisdiction, there is a binding obligation to label AI-generated content. This addresses one of the population's greatest concerns - the fear that deepfakes and AI-generated texts can no longer be distinguished from authentic content. The standardized EU labels create clarity and set a global benchmark. The requirement to label content even without deceptive intent closes a potential loophole. The challenging side: the timeline is extremely tight. Many companies, especially in the SME sector, are not yet prepared for these obligations. The technical implementation of watermarks and labels is complex, particularly for smaller companies without specialized AI teams. Questions of enforcement also remain: how will authorities handle the flood of AI-generated content on social media? And what does the obligation mean for everyday applications - must every automatically generated email reply be labeled? The European Commission's guidelines still leave room for interpretation.
Relevance for Germany
This deadline is highly relevant and urgent for German companies. First, the obligations apply to every organization that uses AI professionally, not just tech companies. A medium-sized machine manufacturer running an AI chatbot on its website must label content just like a DAX corporation. According to the Bitkom Study 2026, 41 percent of German companies already use AI - all of them must check whether their applications fall under the transparency obligations. Second, the Code of Practice deadline on July 22 gives German companies exactly 20 days to secure a regulatory advantage by signing - the presumption of conformity can be decisive during a regulatory inspection. Third, the Federal Network Agency (Bundesnetzagentur), which has only served as Germany's central AI supervisory authority since the AI Implementation Act (KI-MIG) of June 12, 2026, must ensure enforcement of these obligations - a monumental task for an agency still building its new competencies. Fourth, the standardized 'KI' label will become visible in everyday life in Germany - on websites, social media, and digital services. This can help strengthen media literacy among the population in dealing with AI-generated content and increase trust in digital information.
Fact check
The enforcement of transparency obligations under Article 50 on August 2, 2026, derives directly from the EU AI Act's legislative text and is confirmed by the official European Commission page as well as legal analyses from Haufe, IT-Administrator, Greenberg Traurig, and Bird & Bird. The Code of Practice signing deadline of July 22, 2026, at 18:00 CEST is documented by the European Commission on its official Code of Practice information page. The fine amounts of up to 15 million euros or 3 percent of global annual turnover come from the legislative text and are consistently cited across all legal analyses. The standardized labels ('AI GENERATED', 'AI MODIFIED') and localized variants are described in the Code of Practice itself. The figure that 41 percent of German companies use AI comes from the Bitkom study 'Artificial Intelligence in Germany 2026'. The Federal Network Agency's role as AI supervisory authority is established by the AI Implementation Act (KI-MIG) passed by the Bundestag on June 12, 2026.
Source
- • https://digital-strategy.ec.europa.eu/en/policies/code-practice-ai-generated-content
- • https://artificialintelligenceact.eu/transparency-rules-article-50/
- • https://www.haufe.de/recht/kanzleimanagement/kennzeichnungspflicht-fuer-ki-inhalte-gilt-ab-august-2026_222_681220.html
- • https://www.it-administrator.de/eu-ki-transparenzpflicht-kennzeichnung-deepfakes-chatbots