Article 3(41) AI Act defines an RBI system as ‘an AI system for the purpose of identifying natural persons, without their active involvement, typically at a distance through the comparison of a person’s biometric data with the biometric data contained in a reference database’. An AI system must therefore fulfil three conditions to be classified as an RBI system : (i) its purpose must be biometric identification; (ii) that identification must occur without the active involvement of the person concerned, typically at a distance; and (iii) the system must perform a comparison between that person’s biometric dataand biometric data contained in a reference database. Each of these conditions is described in more detail in subsections below.
i. Biometric identification system
Biometric recognition systems can be used for verification (authentication) and identification purposes14. Such systems always rely on biometric data to perform such verification and identification. They detect, capture, and transform measurable physiological and physical characteristics (such as eye distance and size, nose length, etc.) or behavioural characteristics (such as gait or voice) into machine-readable biometric data (see Section 3.1.1.i) above). Those data are available in different forms (images or templates), which are a mathematical representation of the salient features of an individual, used for verification and identification purposes.
Article 3(35) AI Act defines ‘biometric identification’ as ‘the automated recognition of physical, physiological and behavioural or psychological human features for the purpose of establishing the identity of a natural person by comparing biometric data of that individual to biometric data of individuals stored in a database.’ Such identification can be done for a variety of private or public sector purposes. Regarding law enforcement, such identification is typically done in view of an action or decision to be taken by either administrative, law enforcement or judicial authorities.
This definition excludes systems whose intended purpose is not to establish the identity of a natural person and systems which do not compare the biometric data of a natural person to biometric data stored in a database. Processing of biometric data that is not done for the purpose of identifying persons do not constitute biometric identification (e.g. collecting additional information or establishing links between criminal cases based on biometric samples).
AI systems may be used for b iometric identification in all stag es of activities . For example, biometric identificationincludes ‘initial identification’ as used in Article 26(10) AI Act(see Section 3.1.2.2. below). Biometric identification can also cover establishing the identity (including, but not limited to, the civil identity) of a natural person whose identity is already known or previously verified by a human.
An AI system that locates an individual does not necessarily constitute an RBI system. Where the system does not rely on biometric features to locate a person (or there is no comparison of the biometric data of that person with biometric data stored in a reference database), that system will not be classified as high-risk pursuant to point 1(a) of Annex III AI Act. For example, the following AI systems are not considered biometric identification systems:
- AI systems automatically searching for non-biometric identifiers ( including logos on clothing, specific items).
- Automated number plate recognition (ANPR) systems when used for locating a motorised vehicle by the means of the car number plate, by the shape, the make or the colour of the vehicle, related to a person of interest (e.g. missing persons, persons suspected of having committed a crime, subject to a European arrest warrant).
- AI systems automatically searching occurrences of the validation of the individual’s transport tickets (e.g., to establish the movements of the suspect). The ‘tracking’ is done based on the individual’s validations of a ticket/subscription and not based on biometric data.
- AI systems used to geolocate a picture involving criminal activities (such as CSAM) based on background or unique identifiers in the picture excluding biometric data.
By contrast, AI systems intended to beused to track natural persons (e.g., to determine the direction in which a suspect is moving or escaping in the view of taking a possible coercive action ) are covered by the definition of biometric identification, provided the ability to track a person relies on a biometric data and a comparison is made of the biometric data of that person with biometric data stored in a reference database . This conclusion can be deduced from the meaning of Article 5(1)(h)(iii) AI Act and Article 26(10) AI Act. Article 5(1)(h)(iii) AI Act allows the use of real-time RBI for the tracking of suspects of crimes. Article 26(10) AI Act refers to the targeted search of a person suspected or convicted of having committed a criminal offence (see also Section 3.1.2.2. below), provided all other conditions for such use are met (see also Section 3.1.2.2. below). For example, the following AI systems are considered biometric identification systems:
- Automated comparison of biometric identifiers (combination of face, height, frame, hair / eye colour, gait allowing the unique identification of a person) of one or several persons suspected of committing a crime with a live CCTV stream (when exempted from prohibitions in Article 5(1)(h) AI Act under one of the limited exceptions) or with CCTV recordings analysed after the crime and matched against a reference database of biometric templates of known convicted criminals.
- An AI system that uses gait recognition to compare the known fugitive’s stored biometric gait profile of an individual against CCTV footage feeds collected from main transport hubs close by to track the individual after a first detection on the reference material.
- An AI system is deployed during a major event (e.g., sport event, political gathering) to detect the presence of specific known individuals linked to terrorist groups in specific sites using biometric data stored in law enforcement systems (when exempted from prohibitions in Article 5(1)(h) AI Act under one of the limited exceptions).
AI systems that are intended to be used for biometric verificationfall outside the scope of the high-risk classification15 . Biometric verification ( including authentication) consists of comparing biometric data presented at a sensor with another set of previously provided biometric data stored on a device, such as a smartphone, a passport, or an ID cardor stored in a database in an encrypted form while the decryption key is solely stored at the individual’send, e.g., on a smartphone (one-to-one verification). The sole purpose of biometric verification is to confirm that a specific person is who they claim to be. For example:
- An AI system that compares an image of a traveller’sface (facial image) captured at an automated border control gate with the facial image stored in their passport ’ s chip to confirm that the holder is the legitimate document owner would not be considered RBI system. If, however, within the same scenario the biometric data of the traveller captured at the border is in addition to the verification purpose simultaneously compared against a criminal database, for example by Interpol, this would be RBI.
- An AI system comprising a mobile fingerprint scanner used to compare a live fingerprint against fingerprint templates stored on the persons’ electronic ID card or driver’s license for the purposes of checking their identity (e.g., a municipal employee or law enforcement officer requests individuals to place their finger on a mobile fingerprint scanner to verify their identity). Such system would not be considered an RBI system.
ii. Remoteness
According to Article 3(41) AI Act, remoteness implies the ability of an AI system to identify an individual without their active involvement, typically at a distance.
Point 1(a) of Annex III explicitly excludes from high-risk classification AI systems intended to be used for biometric verification the sole purpose of which is to confirm that a specific natural person is the person he or she claims to be. As clarified in Recital 15 AI Act, this may include a biometric system that confirms the identity of a natural person for the purpose of allowing them to have access to a service, unlock a device, or have security access to premises16.
For example, the following would not be considered an RBI system
- An AI system using face scanning technology to admit entry to a restricted area (e.g. power plant premises): such a system compares the face of the individual presented at the entrance camera with a reference image contained in a reference database of persons allowed to enter the building (one-to-many).
- An AI system using smartphone facial authentication linked to a person’s digital ID walletto access an online police reporting portal: such a system compares the biometric sample stored in the digital ID wallet with the biometric data captured by another device to confirm the user’s claimed identity before allowingthe submission of the report.
Recital 17 AI Act clarifies that the exclusion of biometric verification and authentication systems from the definition of RBI systems is justified by the fact that such systems are likely to have a minor impact on the fundamental rights of natural persons, as compared to RBI systems, which may be used for the processing of the biometric data of a large number of persons without their active involvement. That recital further clarifies that RBI systems are typically used to perceive multiple persons or their behaviour simultaneously to facilitate the identification of natural persons without their active involvement.
For active involvement to be present , it is insufficient that persons on whom a biometric identification system is used are informed of such use. Rather, they need to actively participate in its use, for example, by actively and consciously presenting themselves in front of a sensor that is installed in a way to foster active participation.
For example,
- An AI system used to give access to a metro station, such as biometric metro tickets, whereby persons are actively involved and consciously approach the biometric sensor to obtain access, does not fulfil the condition of remoteness.
- In contrast, an AI system used in cameras installed on the walls or ceilings of metro stations for surveillance purposes fulfils the condition of remoteness, since the persons on whom the system is used are not actively involved in their identification process.
- An AI system used for law enforcement purposesusing the biometric data of a personincluded in evidence taken from a crime scene (e.g., finger latent) without the presence of that person (or with the ir presence, but whe re the person is unconscious ) does not constitute active involvement.
- Following an arrest, a suspect is taken into police custody. In accordance with applicable national criminal procedural law, officers capture the suspect’s facial photograph as part of the standard custody processing procedure, uses an AI system to compare a ‘mugshot’ taken during this formal identification procedure against a database of biometric templates. The system does not qualify as an RBI system as it is not remote (there is active involvement of the suspect).
- An AI system used by law enforcement whereby a witness offers to verify their identity by capturing their fingerprint and comparing it solely against the biometric template stored on the person’s electronic ID cardconstitutes active involvement.
- An AI system used by a public authority to check the identity of individuals through facial recognition, against a predefined watchlist of persons suspected of involvement in terrorism. The AI system is used for the purposes of granting them entry to a restricted and fenced security perimeter within an otherwise publicly accessible space (e.g. a city square) during a major public event . Individuals wishing to enter the secured perimeter are informed in advance, through visible notices and public communication, that entry is subject to biometric checks and the facial recognition scanner is positioned to require conscious participation (e.g. standing within a marked capture zone and facing the sensor)to gain entry to the secured area. Such involvement would be considered active involvement.
Biometric identification systems that process ( also contactless, i.e., without physical contact ) fingerprints, gait, voice, DNA, keystrokes and other (biometric) behavioural signals may also constitute RBI systems, provided they fulfil the condition of remoteness described above17. Palm scanning systems will mostly constitute biometric verification or authentication systems. However, in certain cases, palm scanning systems may also be used for identifying a suspect , for example when comparing child sexual abuse material to see whether the same offender is involved.
For example, the following biometric identification can be considered remote:
- A biometric identification technology system that captures voice samples may be deployed on a person who is speaking by using a microphone to collect a biometric sample.
- A gait recognition system may be used via closed-circuit television (CCTV) and the videos automatically checked for matches with previously captured templates.
- Keystroke biometric technology may be used for findinga person typing a fraudulent message by matching the person’s distinctive typing rhythm with previously captured keystroke templates.
Body-cams used by individual law enforcement authorities for the purposes of, for example, the untargeted filming of a demonstration with hundreds of participants, will be considered to fulfil the condition of remoteness. RBI systems used in the virtual space may also be considered to fulfil that condition where the system typically functions at a distance without active involvement.
iii. Reference database
Biometric identification is not possible without a reference database containing biometric data for comparison purposes for identifying natural persons. Thus, the existence of a reference database is indispensable to perform remote biometric identification of a person . A reference database may consist of biometric data of one or many individuals.
Biometric identification against a reference database requires that the comparison is carried out against all data records containing biometric references18.
For example,
- The Automated Fingerprint Identification System of SIS (SIS-AFIS)19 database could be used as a reference database to carry out dactyloscopy searches.
- A national missing persons database could be used as a reference database for facial recognition in case of missing persons.
- A private company’s voice database could be used as a reference database.
a) Practical examples of AI systems falling within the high-risk use case of point 1(a)
- Facial/voice recognition for media archives : AI-enabled facial and voice recognition technology intended to be used to identify selected individuals (e.g., public figures) from audiovisual content by comparing that content against a reference database, such as media archives maintained by public broadcasters and national audiovisual repositories, without the individuals’active involvement.
- (Voice)print matching in databases: AI-enabled solutions intended to be used to analyse recordings or voiceprints to identify individual speakers, by comparing their biometric data against an established database of known individuals and voiceprints without their active involvement.
- CCTV cameras installed in the walls and/or ceilings of a stadium (post remote): If an incident happens during a match, the captured biometric data(i.e. facial images) is compared to a biometric database to identify offenders , without the active involvement of visitors to the stadium.
- An AI system intended to be used to identify , from an image taken from a CCTV camera installed on private property, an individual suspected of carrying out a burglary and operating without their active involvement . Unlike the prohibition in Article 5(1)(h) AI Act, the high-risk use case listed in point 1(a) of Annex III is not limited to RBI in publicly accessible spaces.
- AI systems intended to be used to compare child sexual abuse material (CSAM) found on the Internet to an established national or international database of suspects containing facial images and tattoos to identify an offender without their active involvement.
- An investigation requiring the identification on the internet, via backwards search (reverse image searches), of a terrorist living undercover, without his or her active involvement: A reference database can be a transformation of biometric template(s) taken from information available on the website or the Internet in general.
- An AI system intended to be used to compare attributes in a picture, e.g., taken from the Internet or CCTV, with pictures taken from the Internet and indexed in a database (a reference database can be a transformation of biometric template(s) taken from information available on the website or generally the Internet) to identify the person in the picture without their active involvement.
b) Practical examples of AI systems falling outside the high-risk use case of point 1(a)
- A CCTV facial recognition system at the entry to a stadium to detect blacklisted individuals forbidden to enter the stadiumvia direct access control: such a systemwould not fulfil the condition of remoteness , because natural persons need to actively participate, i.e., actively step in front of the AI system to get access to the stadium.
- AI systems intended to be used for biometric verification (authentication) or for identification that is not remote, such as: o AI based solutions to unlock a smartphone. o AI-based solution s to streamline the onboarding experience for the purposes of identifying traders. o AI-based solution s for the authentication of online exam proctoring (so-called remote proctoring). o AI-based solution s to streamline access to a service (e.g., logging in to a bank account online (login authentication) or calling customer service of a bank to receive information about specific transactions). o AI-based solution to carry out verification of identity during a roadside checkby law enforcement. Officer uses a mobile device to compare the person’s live fingerprint solely against the fingerprint stored on the ID card to verify the person’s identity.
- AI systems for smart homes/residential areas : An AI system intended to be used to authenticate or verify the identity of an individual to give access to a home or residential area by scanning facial images, irises, voice (voiceprint identification), or fingerprints at the entrance and which grants access after a successful check against a reference database made up of biometric data which the relevant individual agreed to register, would be considered to be for biometric verification purposes and would not be considered to fulfil the condition of remoteness (the individual agreed to register and consciously presents themselves at the entrance).
- Smart corridor AI biometrics system: an AI system that captures the facial images of individuals from several angles in a short corridor or passage and compares those images against a biometric database of persons that have registered with the system through which those persons can enter a concert, stadium, major public event at a city square, etc., without stopping, whereas unregistered persons are approached and ejected by security staff or by automatic door systems, depending on the application. Such systems are used solely to gain access to a service/premises and persons that registered would be considered to actively participate in the identification.
- AI systems intended to be used to confirm the identity of a natural person for the sole purpose of having access to a service, unlocking a device or having secure access to premises: e.g., an AI-enabled facial recognition system to contol access to hazardous areas (e.g., used in industrial settings to grant or deny access to machinery or areas with elevated safety risks ); an AI-based solution to control physical access to secure areas within a corporation’s office buildingsAI biometric system for secure facility access used in ; an critical electricity infrastructure or seaport/river port faci lities (employees/authorized personnel only).
- AI systems that make use of biometric data, but which are not intended to be used for biometric identification : e.g. a n AI system that uses a voice sample collected with the individual’s active involvement to carry out language/dialect analysis in order to identify the country and region from where a person originates in the context of migration and asylum processes. The aim of such a tool is not to identify the person. Rather, a voice sample is matched against a pool of accredited voice samples to substantiate country of origin. However, see Section 3.1.3 on Point 1(b) and Section 3.7 on Point 7 of Annex III below.
- AI systems filtering seized material to structure its content: e.g. the AI system filters the seized footage, e.g., 100 000 files, in preparation for the evaluation of the seized evidence, e.g., whether there is a child present in a photo or not.
- AI systems used by law enforcement authorities to analyse crime scenes: All kinds of material are collected from a crime scene including CCTV, witnesses’ material, etc. The objective is to gain investigative insights in what has exactly happened by analysing occurrences of persons and objects (e.g. via clustering, tracking or localisation based on objects such as cars, number plates, clothes, bags, weapons, drugs, and possibly as well for faces, tattoos and birthmarks). Persons are not matched against external databases, i.e. their identity is not established; and only material that has a direct link to the crime place is being used. This is the work-step prior to initial identification, and even if there might be a comparison of faces involved, in view of the closed dataset this would not yet be subject to the high-risk classification under Article 6(2) and Annex III, point (1)(a) AI Act.
- AI systems that use multispectral imaging to capture information from the surface and sub-surface (dermis) of the skin(collecting of biometric data) to capture fingerprints without carrying out biometric matching against a reference database.
- AI systems used solely for cybersecurity and personal data protection purposes: e.g. AI systems with the sole purpose of anonymising personal data , such as b iometric systems which are intended to be used solely for the purpose of enabling cybersecurity and personal data protection measures as indicated in Recital 54 AI Act.
- AI systems that use cameras (e.g., satellite, infrared/thermal) to detect the presence of humans, such as: o AI systems that provide information on seat occupancy in trains or counting people: the AI system uses infrared thermal cameras installed on the ceiling in the train compartment; captures the temperature for each of the seats; and analyses the data and provides updates to the passenger via website/mobile application. o AI systems that are used to detect people on tracks (trespassers) to prevent accidents and damages to the infrastructure : the AI system uses infrared thermal cameras installed around rail tracks, in front of the train, etc.; alerts driver/security team that there are trespassers. o AI system s using images or recordings of specific spaces captured by cameras carried by drones/satellites to verify whether there are any humans present in those spaces based on non-biometric data (e.g., unusual motions, radio frequency signals emitted by the electronic device carried by humans) in the search for missing people (e.g., people lost in vast uninhabited spaces, for instance large forests). Although such AI systems can be used to identify individuals, they do not involve biometric data. o AI systems that use infrared thermal cameras to detect the presence of humans are not classified as high-risk AI systems because their purpose is not to identify individuals, but instead to detect the mere presence of humans20. In addition, such systems do not rely on biometric data.
- AI systems that do not track an individual on the basis of biometric data : e.g., law enforcement uses AI video analytics tools to track a fleeing suspect across cameras based on clothing colour and body outline, without extracting or matching biometric features.
- AI-based large data processing systems used by law enforcementto cross-check and identify links between various criminal cases on the basis of facial imagery without using biometric data to identify the common suspects.
- As definedin Article 3(36 ) AI Act,and agreedby the biometrics community in ISO/IEC Standard 2382-37:2022 Information Technology-Vocabulary, Biometric recognition, Term 37.01.03.
- Annex III, point 1(a)AI Act.
- E.g. Ross A, Jain AK (2015) ‘Biometrics, Overview’ in Li S.Z. and Jain A.K. (eds) Encyclopedia of Biometrics, (1sted. Springer Science, New York), pp. 289-294.
- EDPB-EDPS, Joint Opinion 5/2021, p. 11; Council of the European Union, ‘Opinion of the Legal Service’, 12302/22, 12 September 2022, paragraph 33, and Recital 15 AI Act.
- In line with ISO/IEC 2382-37.
- Article 42, 43 of the Regulation (EU) 2018/1862 of the European Parliament and of the Council of 28 November 2018 on the establishment, operation and use of the Schengen Information System (SIS) in the field of police cooperation and judicial cooperation in criminal matters, amending and repealing Council Decision 2007/533/JHA, and repealing Regulation (EC) No 1986/2006 of the European Parliament and of the Council and Commission Decision 2010/261/EU and Article 32, 33 of the Regulation (EU) 2018/1861 of the European Parliament and of the Council of 28 November 2018 on the establishment, operation and use of the Schengen Information System (SIS) in the field of border checks, and amending the Convention implementing the Schengen Agreement, and amending and repealing Regulation (EC) No 1987/2006.
- However, in the context of migration, asylum and border control management, an AI system that uses infrared thermal or satelliteimagingto detect people crossing aborder in an unauthorised placewouldfall under the use case of point 7(d)of Annex IIIAI Act (Section 3.7.5 below).