Remote ID
AI matches camera footage against a database to recognise a person.
Annex III point 1
Biometric AI needs early scoping because identification, categorisation and emotion recognition can each trigger a different route.
Remote biometric identification.
Biometric categorisation using sensitive or protected attributes.
Emotion recognition, especially in work and education contexts.
Check first whether Article 5 prohibits the use case.
Facial recognition in public or semi-public spaces
Matching against a reference database
Emotion, stress or attention classification
Biometric categories that steer access or treatment
Article 5 prohibition check
GDPR/DPIA and biometric data
Human oversight and error margins
Bias and representativeness questions
AI matches camera footage against a database to recognise a person.
The system infers protected or sensitive attributes from biometric signals.
A tool estimates stress, attention or behaviour from face or voice signals.
This page is a practical guide for first classification and preparation. For legal decisions, a full system and context review remains necessary.
Annex III point 4
Could the AI affect a person’s job opportunity, evaluation or working conditions?
High-risk AI in HR and employmentAnnex III point 5
Could the AI make access to an essential service easier, harder or more expensive?
High-risk AI in essential servicesAnnex III point 3
Does the AI support learning, or does it help decide something about a learner?
High-risk AI in educationAnnex III point 2
Does the AI influence safety, availability or priority for a critical service?
High-risk AI in critical infrastructureStart with the Gap Intake and share which AI systems, suppliers and processes are in scope. Then we can sharpen the classification and first evidence route.