EU AI Act Mandates AI Disclosure — and Europe Fears 'Disclosure Fatigue'

From August 2026, the key transparency provisions of the EU AI Act have taken effect in the European Union: companies must clearly notify users when they interact with AI or view AI-generated or AI-edited content. As Wired notes, Europeans are realizing for the first time just how deeply AI is already embedded in their everyday lives.

What exactly does the EU AI Act require

Article 52 of the AI Act sets mandatory disclosure requirements for operators of AI systems. Chatbots and voice assistants must tell the user, in every interaction, that they are talking to an AI and not a human. Any content — text, images, audio or video — created or substantially altered with the help of AI must be clearly labeled. The law singles out deepfakes separately: synthetic media depicting real people require special disclosure.

• Chatbots and voice AI — notification in every interaction
• AI-generated text, images, audio, video — mandatory labeling
• Deepfakes of real people — special marking
• Penalty for violations: a fine of up to 15 million euros or 3% of global annual turnover

The requirements apply to all companies serving users in the EU, regardless of where they are registered. American and Asian platforms must comply with the new rules on the same footing as European ones.

Why are experts warning about 'disclosure fatigue'?

The industry's main worry is disclosure fatigue, or 'warning fatigue'. The average user now interacts with AI systems dozens of times a day: in email with autocomplete, in search engines, on streaming services, in navigation apps, when contacting customer support. If every one of these contacts comes with an explicit warning, the information value of those warnings quickly erodes — exactly as happened with cookie banners.

"People are about to fully realize just how much AI has already become part of their everyday activities,"

Wired writes in its coverage of the new regulation.

Regulators and UX specialists are looking for formats that actually work. One concept under discussion is a unified visual icon for AI content, similar to the HTTPS padlock in a browser's address bar: small, instantly recognizable, and requiring no reading of text in every interaction.

How companies are preparing for the requirements

Large platforms — Google, Meta*, Microsoft — are already developing and testing AI content labeling systems as part of their preparation for the EU AI Act. Some companies build notifications directly into the interface for every interaction, while others limit themselves to a banner at first login or special settings in the user profile. Whether such approaches meet the spirit of the law remains an open question: the European Commission requires clear and conspicuous disclosure, not fine print in a page footer.

Small and medium-sized businesses are under additional pressure: building the necessary infrastructure without significant resources is not easy, and the fines are the same for companies of any size.

What this means

The EU AI Act is becoming the first mass test of how society copes with the ubiquity of AI. Disclosure is the right principle, but its real effectiveness will be determined by design: how often warnings appear, how visible they are, and how readable. The coming months will show whether AI disclaimers become an effective tool or repeat the fate of cookie banners.

Frequently asked questions

How much does violating the AI disclosure rules under the EU AI

Act cost?

Violating the EU AI Act's transparency requirements carries a fine of up to 15 million euros or 3% of the company's global annual turnover, whichever is higher.

Who must comply with the AI disclosure requirements?

The requirements apply to all companies providing services to users in the EU, including foreign ones — American, Asian and others. The company's place of registration does not matter.

*Meta is designated an extremist organization and banned in Russia.