Direct interaction
Providers of systems interacting directly with people must generally inform them of the AI interaction unless it is obvious. A website assistant should have a clear notice at the start.
Deepfakes and text
Deployers must disclose artificially generated or manipulated deepfakes, with adapted disclosure rules for evidently artistic, creative, satirical or fictional works. The text rule concerns publication to inform the public on matters of public interest. An exception applies where human review or editorial control takes place and a person or entity holds editorial responsibility.
Different roles, different duties
Providers of generative systems also have machine-readable marking duties, subject to statutory exceptions. A label does not replace image rights or a lawful basis for personal-data processing. Identify your role, classify the content, document any exception and record approval.
Practical next step
Use clear labels such as “AI assistant” or “AI-generated illustration”, not a hidden clause. This is general guidance, not individual legal advice. Check the primary text before launching a new use case.
What to prepare for the first conversation
Record provider/deployer role, interaction or publication, content type, possible deepfake status, public-interest purpose, human controls and editorial responsibility. Then specify where the notice appears and who checks it. Technical labelling follows this assessment. Obtain legal review where uncertain instead of assuming an exception.
Discuss a project ↗