The EU’s New AI Labeling Rules: What Changes Starting August 2, 2026?

The European Union’s new AI labeling rules, set to take effect on August 2, 2026, make it a legal requirement to clearly label deepfake images, AI chatbots, fully AI-generated texts of public interest that lack human oversight, and biometric systems. Introduced under Article 50 of the Artificial Intelligence Act, this regulation applies not only to companies based in the EU, but to all businesses globally whose AI outputs or products are consumed by EU citizens.

Which Content Must Be Labeled?

Contrary to fears that all AI usage is being targeted, the regulation focuses specifically on distinct and sensitive areas. According to Article 50(4) of the AI Act, the main categories requiring mandatory labeling are:

• Deepfake Content: Any visual, audio, or video material that mimics a real person, object, place, or event and is presented as authentic.
• Chatbots and AI Agents: Users must be explicitly made aware that they are not interacting with a human.
• Fully AI-Authored Texts: Especially content of public interest that has not undergone human supervision and editorial filtering.
• Emotion Recognition and Biometric Categorization Tools: Systems that analyze individuals' emotional states or biometric data.

Complying with these rules is the joint responsibility of both the providers who develop the AI systems and the deployers who utilize these systems in their commercial processes.

Which AI Content is Exempt?

Contrary to concerns circulating in the market, not every AI-assisted work requires labeling. The most important exemptions designed to relieve teams that use AI in their workflows are:

• Human-Reviewed Texts: Even if generated by AI, texts that have been reviewed and edited by a human—backed by an individual assuming legal/editorial responsibility—are exempt from the labeling requirement.
• General AI Assistance: Routine AI uses that do not fall under the scope of "public interest" and are not deceptive, such as coding support, drafting, or standard productivity tools, are excluded from the requirement.

However, legal experts advise advertising and marketing teams to apply labels as a precautionary measure for commercial illustrations and product visuals that closely mimic a real object or person.

Sectoral Implications and Key Considerations

For design, marketing, and software agencies operating in the global market, this regulation introduces a mandatory transparency requirement throughout the supply chain. Much like GDPR and the European Accessibility Act (EAA), using an externally licensed AI tool does not exempt companies from legal liability. Therefore, it is critical for digital product owners and agencies to document the processes through which their tools' outputs are generated, as well as their layers of human oversight.

*This news report has been prepared based on data published by Smashing Magazine.