AI Act Article 5(1)(F)
Definition
The EU AI Act prohibition on AI systems that infer emotions, intentions, or characteristics from biometric data in law-enforcement and related contexts. Likely covers computerised polygraph scoring systems sold or used in EU member states.
- Scope
- Applies to AI systems in law-enforcement contexts that infer emotions, intentions, or characteristics from biometric data.
- Technologies Covered
- Computerised fMRI analysis, automated polygraph scoring, and similar physiological deception-detection systems.
- Geographic Jurisdiction
- EU member states where these systems are sold or used.
Common questions
What does the AI Act Article 5(1)(f) actually prohibit?+
It bans AI systems used in law enforcement from inferring emotions, intentions, or characteristics from biometric data. This covers technologies like automated fMRI analysis and computerised polygraph scoring systems that try to detect deception or infer mental states.
Does this rule apply to polygraphs and fMRI lie detectors?+
Yes. Any computerised system that processes physiological data (like heart rate, blood pressure, brain imaging) to infer deception or emotional state falls under this prohibition when deployed in law-enforcement contexts in the EU.
Why did the EU restrict AI-based deception detection?+
These technologies lack reliable scientific grounding. fMRI and polygraph results are prone to false positives and false negatives, and automating their interpretation through AI amplifies those risks when stakes are high, such as in criminal investigations.
Related terms
- BEOS / Brain Fingerprinting
- Lawrence Farwell's EEG-based method using the P300 event-related potential to detect concealed knowledge of crime-relevant items. Addressed by the Indian Supreme Court...
- BOLD Signal
- Blood-oxygen-level-dependent signal; the measure in fMRI that captures changes in the ratio of oxyhaemoglobin to deoxyhaemoglobin as a proxy for neural activity....
- BSA § 39
- Section 39 of the Bharatiya Sakshya Adhiniyam 2023 (formerly IEA § 45), governing the admissibility of expert opinion in Indian courts. The...
- Cephos Corporation
- A US company that offered commercial fMRI lie-detection services from approximately 2006. Its technique was excluded from evidence in US v. Semrau...
- Concealed Information Test (CIT)
- The Lykken paradigm that detects concealed knowledge by measuring the orienting response to crime-specific items embedded among foils. Theoretically sounder than the...
- Control Question Test (CQT)
- The dominant polygraph examination format, comparing physiological responses to crime-relevant questions against responses to emotionally loaded control questions. Criticised for its false-positive...
- Daubert Standard
- The US federal evidentiary standard (Daubert v. Merrell Dow Pharmaceuticals, 1993) requiring that expert testimony be based on scientifically valid methods with...
- Ecological Validity
- The degree to which a measurement or test procedure produces results that generalise from the controlled laboratory context to the real-world condition...
- False Positive (PDD)
- A truthful person classified as deceptive by the polygraph examination. The rate of false positives in field conditions is non-trivially high and...
- Frye Standard
- The US legal test for admissibility of scientific evidence, originating from Frye v. United States (1923), which required that a technique be...
- Greely-Illes Critique (2007)
- Henry Greely and Judy Illes's 2007 analysis in the American Journal of Law and Medicine identifying five categories of problem with fMRI...
- No Lie MRI
- A San Diego-based company offering fMRI lie-detection services during the same period as Cephos. No court has admitted No Lie MRI evidence...
Explained in these topics
- fMRI Lie Detection and the Cephos / No Lie MRI DebateThe EU AI Act prohibition on AI systems that infer emotions, intentions, or characteristics from biometric data in law-enforcement contexts, covering computeri...
- Polygraphy, PDD Methods and the US v. Scheffer Inadmissibility FrameThe EU AI Act prohibition on AI systems that infer emotions, intentions, or characteristics from biometric data in law-enforcement and related contexts. Likely...