EU Rules Will Require Labeling AI-Generated Content Starting August 2

On August 2, 2026, the next phase of the EU's artificial intelligence regulation, known as the AI Act, comes into force. This time the new obligations no longer apply solely to the big tech companies developing AI models, but spill over into media, advertising, e-commerce, schools and anyone who uses chatbots in daily life.

The new rules require providers and users of AI systems to label content that artificial intelligence created from scratch or substantially modified, if a recipient could mistake it for authentic material. This covers text as well as images, audio and video, including deepfakes depicting real people, events or places.

In practice, this means advertising companies, online stores using AI-generated product descriptions, newsrooms publishing AI-assisted content, and schools using AI-based educational tools will all have to implement mechanisms for disclosing the origin of such material. The obligation also extends to ordinary citizens in contexts where AI helps draft documents with significant legal or financial consequences.

What the labels look like

The European Commission has already published a set of three official icons as part of its code of practice on labeling AI-generated content. The first signals that AI was involved in creating the material at all, the second marks content generated entirely by AI without human involvement, and the third applies to authentic material that has been modified by AI, for example swapping a face in a photo or virtually staging a room for a real estate listing. The icons are available free of charge in black, white and transparent versions, in SVG and PNG formats.

Media newsrooms received an important exemption: labeled AI-generated content does not require additional disclosure if it has been verified by journalists and the publisher takes editorial responsibility for it. This does, however, require keeping documentation confirming that such verification actually took place, which for smaller newsrooms could mean additional administrative burden.

Penalties for non-compliance

Failing to meet transparency obligations carries fines of 7.5 to 15 million euros, or 1 to 3 percent of a company's global annual turnover, whichever is higher. For more serious violations, including prohibited practices, the fines rise to 35 million euros or 7 percent of turnover. This puts the penalty level close to that known from the EU's GDPR, intended to push companies to treat labeling obligations as seriously as personal data protection.

The European Parliament also introduced an absolute ban on AI-generated non-consensual intimate content and material depicting the sexual abuse of children. Compliance with this ban is to be fully enforced starting December 2, 2026, a few months after the general labeling obligations take effect.

Poland's angle

In Poland, the new EU rules coincide with a national law on artificial intelligence systems, currently awaiting the president's signature after clearing votes in the Sejm (the lower house of Poland's parliament). According to official announcements, a Commission for AI Development and Security is to be established, responsible for overseeing the implementation of EU rules at the national level. Mirosław Wróblewski, head of the Personal Data Protection Office (UODO), has previously pointed out the difficulties law enforcement faces in responding to cases involving AI-generated intimate images, showing that the EU regulation alone will not solve enforcement problems on the ground.

For the average user of ChatGPT or another chatbot, the new rules mainly mean a clear division of responsibility: artificial intelligence can help draft an appeal against an official decision, a complaint to an institution, a benefits application, or a letter to a homeowners' association, but it is the human who remains responsible for the content of the document, especially when money, health, employment or other administrative decisions are at stake.

Companies and institutions have until August 2, 2027 to fully implement the entire AI Act, but August 2026 is seen as the moment when the rules stop applying only to the developers of large models and start genuinely affecting the daily work of newsrooms, advertising agencies, online stores and public offices that use AI tools.

Sources: Masz ChatGPT? Nowe przepisy UE mogą cię zaskoczyć (gazetaprawna.pl), EU Icons for labelling AI-generated content (digital-strategy.ec.europa.eu), Deepfakes, Chatbots, AI-Generated Text: European Commission Details Transparency Obligations Under the AI Act (gtlaw.com).