EU AI Act обязывает раскрывать ИИ: Европа боится «усталости от дисклеймеров»

Starting from August 2026, key provisions of the EU AI Act on transparency have come into force in the European Union: companies are required to explicitly notify users when they interact with AI or view AI-generated or AI-edited content. As Wired notes, Europeans are for the first time fully realizing how deeply AI is already embedded in their daily lives.

What the EU AI Act specifically requires

Article 52 of the AI Act (EU AI Act) introduces mandatory disclosure requirements for AI system operators. Chatbots and voice assistants must inform users at each interaction that they are communicating with AI, not a human. Any content — text, images, audio, or video — created or significantly altered using AI must be explicitly labeled. The law specifically highlights deepfakes: synthetic media files depicting real people require special disclosure.

• Chatbots and voice AI — notification at each interaction
• AI-generated text, images, audio, video — mandatory labeling
• Deepfakes featuring real people — special marking
• Violation of requirements: fine of up to €15 million or 3% of global annual turnover

The requirements apply to all companies serving EU users, regardless of the country of registration. American and Asian platforms must comply with the new rules on equal terms with European ones.

Why experts warn about "disclosure fatigue"?

The industry's main concern is disclosure fatigue. The average user today interacts with AI systems dozens of times a day: in email with autocomplete, in search engines, streaming services, navigation apps, and when contacting customer support. If each of these interactions is accompanied by an explicit warning, their informational value quickly diminishes — just as happened with cookie banners.

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

Wired notes in its coverage of the new regulation.

Regulators and UX specialists are looking for formats that actually work. One concept being discussed is a unified visual icon for AI content, similar to the HTTPS lock in the browser address bar: small, instantly recognizable, without the need to read text at each interaction.

How companies are preparing for the requirements

Major platforms — Google, Meta, Microsoft — are already developing and testing AI content labeling systems as part of their EU AI Act preparations. Some companies are embedding notifications directly into the interface at each interaction, while others limit themselves to a banner on first login or special settings in the profile. The question of whether such approaches comply with the spirit of the law remains open: the European Commission demands clear and noticeable disclosure, not fine print in the page footer.

Small and medium-sized businesses have come under additional pressure: building the necessary infrastructure without significant resources is difficult, and 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: the frequency of warnings, their visibility and readability. The coming months will show whether AI disclaimers will become an effective tool or will repeat the fate of cookie banners.

Frequently Asked Questions

How much does violating EU AI Act AI disclosure rules cost?

For violating the EU AI Act transparency requirements, a fine of up to €15 million or 3% of the company's global annual turnover is imposed — whichever amount is greater.

Who do the AI disclosure requirements apply to?

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

*Meta has been recognized as an extremist organization and is banned in Russia.