On August 2, new requirements of the Artificial Intelligence Act (AI Act) came into force in the European Union, requiring the labeling of content created with the help of AI.
On August 2, new requirements of the Artificial Intelligence Act (AI Act) came into force in the European Union, requiring the labeling of content created with the help of AI.
The European Commission said the new rules should help users understand when they are interacting with artificial intelligence or viewing content created by it, by requiring AI developers to clearly label it. The requirements apply to images, audio and video that mimic real people, as well as objects, places, events and texts. The rules also apply to systems that analyze emotions or use human biometric data.
Content must have clear visual markings and machine-readable labels — for this, the EU has developed a special set of icons to identify AI content. In particular, users must now be warned if they are communicating with an artificial intelligence rather than a human.
The law does not apply to systems that perform purely auxiliary functions — for example, spell checking or basic photo processing. A separate, more flexible regime applies to materials of an obviously artistic, creative, satirical or fictional nature — in such cases, other methods of informing the audience are allowed, as long as they do not interfere with the perception of the work itself.
Companies that fail to comply with the rules face fines of up to €15 million or 3% of annual global turnover, whichever is greater. For EU institutions, bodies and agencies, the penalty is lower, up to €750,000. For small and medium-sized enterprises, penalties will be applied on a proportional basis.
Recall that the AI Act came into force on August 1, 2024, but its requirements are being implemented in stages — the current content labeling rules have become the next stage of this process.
