EU begins labelling AI-generated content: who is affected by the new transparency rules under the AI Act

Transparency requirements under the European Union’s Artificial Intelligence Act began to apply on 2 August. Users must be informed when they are interacting with AI, while deepfakes and certain other algorithmically generated materials must carry clear and machine-readable labels.

The new rules apply to providers and companies that use artificial intelligence systems. Their purpose is to help people distinguish authentic content from synthetic material and reduce the risks of deception, fraud and disinformation.

Users must be clearly informed when they are interacting with a chatbot, virtual assistant, AI agent or digital avatar rather than a real person.

Providers of generative AI systems must make it possible to automatically identify images, video, audio and text that have been generated or altered using artificial intelligence. Such materials must therefore contain machine-readable markings.

Which content must be labelled

Clear labelling is required for deepfakes — generated or manipulated images, audio and video that convincingly depict real people, objects, places or events.

AI-generated texts concerning matters of public interest must also be labelled when they are published without human review or editorial control. The rules therefore do not require every text prepared with the assistance of artificial intelligence to be marked as AI-generated.

People must also be informed when emotion-recognition or biometric-classification systems are being used. Exceptions apply, among other things, to standard editing tools when AI does not substantially alter the original material or its meaning.

What penalties are possible

Compliance will be monitored by national market-surveillance authorities, the European AI Office and the European Data Protection Supervisor when the systems are used by EU institutions.

Companies may face fines of up to €15 million or 3% of their total worldwide annual turnover. EU institutions, bodies and agencies may be fined up to €750,000. Penalties imposed on small and medium-sized businesses must take proportionality into account.

In practice, the new requirements affect AI developers, online platforms, media organisations, advertising agencies, public authorities and other organisations that publish synthetic content or use automated systems to communicate with people.

Based on materials from: European Commission, European Commission guidelines on Article 50, EU Artificial Intelligence Act