Point 6(b) of Annex III classifies as high-risk AI systems intended to be used as polygraphs, also known as lie detectors, or similar tools. Such systems are intended to be used to assess whether a person is telling the truth by measuring physiological responses, such as heart rate, blood pressure, respiration, and skin conductivity.
Systems similar to polygraphs include voice stress analysis that examine vocal patterns for signs of deception; eye-tracking and pupillometry that monitor eye movements and pupil dilation linked to cognitive stress; brain-imaging techniques that detect brain activity associated with lying; and behavioural analysis tools that use AI to interpret micro-expressions, body language, or speech cues.
Such systems pursue a different purpose from the emotion recognition systems falling within the use case listed point 1(c) of Annex III Polygraphs are designed to infer deception and in law enforcement context are mostly used during questioning, whereas emotion recognition systems analyse biometric cues to infer emotional states (e.g. sadness or anger), typically without direct physical contact, making them relevant for such tasks as risk prediction or prioritisation of video material (see Section 3.1.4. above and Section 7 in the Guidelines on prohibited artificial intelligence practices). AI systems intended to be used as polygraphs or similar tools may include emotion recognition functionalities, since fear may serve as a proxy for lie detection. In such a case, the AI system may be considered to fall within the scope of the use case listed in point 1 of Annex III.
a) Practical examples for AI systems falling within the scope of point 6(b)
- AI systems intended to be used to analyse facial micro-expressions to assess the credibility of answers during an interrogation. Likewise, AI systems measuring eye-tracking and pupillary responses while a suspect answers questions. The role of such systems in decision-making and profiling excludes them from benefitting from the exceptions listed in Article 6(3) AI Act.
b) Practical examples for AI systems falling outside of the use case point 6(b)
- AI systems intended to be used to monitor police officers’ fatigue or stress levels (via wearable sensors or facial analysis) for occupational health and safetyreasons. Such systems should also be considered to fall outside the use case listed in point 1(c) of Annex III , as further clarified in Recital 18 AI Act (see Section 3.1.4 ); being intended to be used for medical and safety reasons, they are also excluded from the prohibited practice listed in Article 5(1)(f) AI Act.
- AI systems intended to be used by law enforcement authorities that transcribe interviews and flag changes in tone (e.g. higher tone, slower/faster speaking pace), without drawing conclusions about the truthfulness of statements made. It should, however be checked whether a particular use case would not amount to emotion recognition system and thus fall under point 1 of the Annex III of the AI Act.
- AI systems that flag irregularities in how different polygraph examiners interpret similar cases, highlighting inconsistencies for law enforcement authorities. Such system is not a polygraph itself, but is intended to be used to interpret the work of polygraph examiners.