AI labeling requirement from August 2026 — What applies under Art. 50?

Frequently asked questions

When does the AI labeling requirement apply?

From 2 August 2026. There's an exception for makers of AI tools like image generators: if their tool was already on the market before that date, they have until 2 December 2026 to add the technical marking.

Who must label — provider or user?

Both, but differently. The makers of AI tools (e.g. providers of image generators) must build their software so AI content is technically detectable. You, as the operator publishing the content, additionally make sure people can recognize deepfakes, certain AI texts, and chatbots.

Is a caption or description enough?

Usually not, for social media and visual content. The notice needs to be visible on the content itself, not just in the caption. For chatbots, it's enough if the notice appears at first contact.

Are product descriptions exempt?

Yes, usually. Normal product and advertising copy is typically exempt. Unless the text itself informs on a matter of public interest, such as health or politics. Texts that a human has reviewed and that have a named responsible person are also usually exempt.

What happens if you fail to comply?

Fines apply, tiered by severity. The highest tier. Up to €35 million or 7% of global annual turnover. Is for prohibited AI practices like social scoring. For breaches of labeling duties like the ones on this page, it's up to €15 million or 3% of turnover. The tier relevant to this checker. Giving false information to authorities can additionally cost up to €7.5 million or 1%. Whichever amount is higher applies. This checker does not replace case-specific legal review.