EU: European Parliament publishes study on biometric identification with recommendations for AI Regulation
The European Parliament published, on 2 September 2021, a study on biometric recognition and behavioural detection as requested by the Committee on Legal Affairs and Committee on Petitions. In particular, the study assesses the ethical aspects of biometric recognition and behavioural detection techniques with a focus on their current and future use in public spaces with respect to biometric, as well as making a number of suggested amendments to the European Commission's Proposal for an AI Regulation ('the Proposal') based on such assessment.
More specifically, the study outlines that the creation and storage of a unique template that identifies a particular person means transforming unique physical features of a human being into digital data, leading to a 'datafication' of humans and that since the features that uniquely identify a person are part of a person's body, their collection and use interfere with a human's personal autonomy and dignity. In addition, the study notes that the use of biometric identification methods in public spaces raise ethical issues with respect to large-scale surveillance of individuals.
Moreover, the study highlights that the development and use of categorisation systems raise ethical issues in relation to the definition of categories, the associated assumptions and the conclusions or reactions triggered by the system, leading to risks such as discrimination, stigmatisation, and the drawing of inappropriate inferences, as well as manipulation and exploitation of group-specific vulnerabilities. Furthermore, the study notes that particular attention should be paid to systems detecting human emotions, thoughts and intentions which potentially calls for a new set of 'neuro-rights' (such as the right to mental privacy and mental integrity).
In consideration of the above, the study proposes to include in the Proposal a new Title IIa that is devoted to restricted AI practices, including biometric techniques and inferences, ensuring responsible use of these techniques without stifling innovation and growth. In addition, the study suggests amendments to the several of the definitions in Article 3 of the Proposal, as well as to Title II on prohibited AI practices including, among others, that real-time remote biometric identification should be removed from Article 5 of the Proposal and transferred to a new Title IIa on 'restricted AI practices'; the list of prohibited AI practices in Article 5(1) of the Proposal should be enriched; the Commission should have the possibility to adapt the list of prohibited AI practices periodically; and clarification that prohibitions following from other laws (such as data protection or consumer protection law) remain unaffected. Lastly, the study proposes that Annex III point 1 of the Proposal should be extended to cover emotion recognition systems in (at least) the same way as biometric categorisation systems.
You can read the study here.