EU’s AI Labeling Mandate Nears: Deepfakes and Chatbots Face Tightest Scrutiny
Published on 07/21/2026 at 19:52 | Redaktion boerse-global.de
Just over a third of European businesses have started preparing for a sweeping new transparency law that will force them to mark any AI-generated content — or risk penalties of up to €15 million. With the compliance deadline set for 2 August 2026, the clock is ticking for companies that still lack the technical infrastructure to watermark and sign their synthetic media.
The rules, rooted in Article 50 of the EU AI Act, apply to text, images, and audiovisual material. Content must carry machine-readable labels such as digital watermarks or signed metadata. The goal is to let users know instantly whether they are interacting with an AI or viewing artificially created work. A voluntary code of conduct published by the European Commission in June 2026 will standardise the icons and symbols used for these labels.
Deepfakes face the most stringent requirements. Any image, video, or audio that realistically mimics a real person or event and is not obvious to the viewer as artificial must be flagged. The code defines a deepfake as content whose artificial nature is not immediately apparent to an ordinary observer.
Exemptions for Human Oversight
Not all AI output needs a label. If a human takes editorial responsibility — by reviewing and editing the text before publication — the labeling obligation disappears. This exemption covers marketing copy, product descriptions, and artistic, satirical, or fictional works. But the rules tighten sharply when the topic touches public interest areas such as politics, health, or the environment. In those cases, any text produced entirely by AI without human oversight must be labelled from August. Chatbots on company websites also have to be clearly identifiable as AI.
Staggered Deadlines and a Ban on Nudifier Apps
Existing AI models already in use get until 2 December 2026 to comply. Systems placed on the market after 2 August must meet the rules immediately. On the same date, a ban enters force on “nudifier” applications that create sexual depictions of people without their consent.
High-risk AI systems have a later cut-off: 2 December 2027. Product-related high-risk systems have until 2 August 2028. In Germany, the Federal Network Agency (Bundesnetzagentur) will be the market-surveillance authority.
Liability Risks Multiply: Courts Already Setting Precedent
The financial sting goes beyond administrative fines. Companies face civil liability too. In May 2026, the Higher Regional Court of Hamm confirmed that businesses are responsible for errors made by their AI chatbots. At the same time, a legal opinion published in early July 2026 warns that professionals who refuse to use AI might breach their duty of care — if a reasonable practitioner would have employed the technology to prevent mistakes.
The Higher Regional Court of Vienna has also weighed in, protecting a publicist’s personality rights against non?obvious AI manipulation.
Despite these mounting risks, most firms are unprepared. According to recent reports, only about 35% of companies have started implementing the required labeling processes. With the first deadline just over a year away — and penalties reaching €15 million or 3% of global annual turnover — the window for catching up is narrowing fast.
Disclaimer regarding our articles: No investment advice, no buy or sell recommendation. Information on prices, companies, and markets is provided without guarantee; changes are possible at any time. Stock market transactions can lead to substantial losses. Our articles are created and reviewed in whole or in part automatically with the support of AI.
