The EU is introducing mandatory labeling of AI content: why the new rules will be difficult to implement

What is changing

The new rules should make the use of artificial intelligence more transparent. Before viewing, the user must understand whether the content was created or edited with the help of AI. For this, it is recommended to use both human-visible markers and technical markers that can automatically recognize online platforms.

At the same time, the requirements do not apply to artistic, satirical or fictional content, if it does not mislead the audience.

Why the implementation of the rules is questionable

The main problem remains not the requirements of the law themselves, but the lack of uniform technological standards. Today, companies use different methods of marking - digital watermarks, metadata or systems to confirm the origin of content, but they are often incompatible with each other.

In addition, after editing, compressing files or re-uploading to other platforms, such marks may disappear or change, making it difficult to verify the origin of the materials.

What challenges do experts see?

Experts point out that effective implementation of the AI ​​Act is hindered by the lack of an international standard for labeling AI content, different approaches to its recognition, and the too rapid development of artificial intelligence technologies. In addition, even if European companies fully comply with the requirements, unmarked content created outside the EU will continue to enter the European information space unhindered.

That is why the European Commission recommends using several methods of marking and checking at once, rather than relying on just one tool.

When all the norms will work

Most provisions of the AI ​​Act become mandatory from August 3, 2026, and full implementation of the law is planned for August 2, 2027. Significant fines are provided for violation of the requirements — from 7.5 to 35 million euros or up to 7% of the company's annual turnover, depending on the nature of the violation.