The essentials
Not every use of AI triggers the same duty.
Article 50 combines several transparency duties. Providers must design certain systems so people can recognise AI interactions and artificial or manipulated content can be detected through machine-readable marking. Deployers must provide notices or labels in specific use cases.
Design the system correctly
Especially relevant are direct AI interactions and systems that generate or manipulate synthetic audio, image, video or text content.
Inform people correctly
Especially relevant are emotion recognition, biometric categorisation, deepfakes and certain AI text on matters of public interest.
Duties at a glance
Four questions reveal the first action points.
Do people interact directly with AI?
People must generally be informed no later than the first interaction, unless this is obvious to a reasonably well-informed and observant person.
Does the system generate or manipulate content?
Providers of relevant systems must mark outputs in a machine-readable format so artificial generation or manipulation can be detected.
Are people being analysed?
Deployers of emotion recognition or biometric categorisation must inform exposed people. GDPR compliance must also be assessed separately.
Are deepfakes or public-interest texts published?
Disclosure duties apply, subject to statutory exceptions and adjusted requirements, including for editorially reviewed text or artistic works.