Article 5(1)(h) AI Act - real-time remote biometric identification (RBI) systems for law enforcement purposes

Recital 32 AI Act acknowledges the intrusive nature of real-time RBI systems in publicly accessible spaces for law enforcement purposes to the rights and freedoms of persons concerned, to the extent that it may affect the private life of a large part of the population, evoke a feeling of constant surveillance, and indirectly dissuade the exercise of the freedom of assembly and other fundamental rights. Technical inaccuracies of AI systems intended for the remote biometric identification of natural persons can lead to biased results and entail discriminatory effects. Such possibly biased results and discriminatory effects are particularly relevant with regard to age, ethnicity, race, sex or disabilities. In addition, the immediacy of the impact and the limited opportunities for further checks or corrections in relation to the use of such systems operating in real-time carry heightened risks for the rights and freedoms of the persons concerned in the context of, or impacted by, law enforcement activities.

However, where the use of such systems is strictly necessary to achieve a substantial public interest and where the situations in which such use may occur are exhaustively listed and narrowly defined, that use outweighs the risks to fundamental rights (Recital 33 AI Act). To ensure that such systems are used in a ‘responsible and proportionate manner’, their use is subject to the safeguards and the specific obligations and requirements in Article 5(2)-(7) AI Act.

Several cumulative conditions must be fulfilled for the prohibition in Article 5(1)(h) AI Act to apply:

(i) The AI system must be a RBI system, (ii) The activity consists of the ‘use’ of that system, (iii)in ‘real-time’, (iv) in publicly accessible spaces, and (v) for law enforcement purposes.

The second condition, i.e. the ‘use’ of the AI system, has been already analysed in section 2.. of these Guidelines. The other conditions listed above are further described and analysed below.

9... The Notion of Remote Biometric Identification

Biometric recognition technologies detect, capture, and transform measurable 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 7...d) above). These data are available in different forms: images or templates, which are a mathematical representation of the salient features of an individual, used for recognition purposes. Biometric recognition technologies are used for verification and identification purposes.

According to Article 3(41) AI Act, a RBI system is [a]n 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.

This definition covers only the identification functionality of biometric recognition systems, which implies the absence of active involvement of the persons concerned (i.e. no active participation) and results in the capture of the characteristics of those persons typically at a distance. For identification performance, the captured biometric data are compared with biometric data already stored in a reference database (such as a repository, e.g., a criminal database containing facial images or templates of suspects). a) Identification purposes only

The notion of ‘biometric identification’ is defined in Article 3(35) of the AI Act 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.

Recital 15 AI Act further clarifies that such human features may comprise the face, eye movement, body shape, voice, prosody, gait, posture, heart rate, blood pressure, odour, keystroke characteristics,

AI systems used for following natural persons can also be included in the definition of biometric identification, for example to see in which direction a suspect escapes. This can be concluded from Article 5(1)(h)(iii) AI Act, that allows for the localisation of suspects of crimes. The localisation is possible when a person is being followed.

AI systems that are intended to be used for biometric verification fall outside the scope of the prohibition in Article 5(1)(h) AI Act. Biometric verification (or authentication) consists of comparing data presented at a sensor with another set of previously recorded data stored on a device, such as a smartphone, a passport, or an ID card. The purpose of biometric verification is to verify that a specific person is who they claim to be. An example of biometric verification is the comparison of a traveller’s face scanned at an e-gate with the facial image contained in their passport.

b) Remoteness

According to Article 3(41) AI Act, remoteness implies the ability of biometric systems to identify individuals 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.

The use of biometric systems to confirm the identity of a natural person for the sole purpose of having access to a service, unlocking a device, or having security access to premises is excluded from the concept of ‘remote’ (Recital 15 AI Act). This modality is used, for example, in access control.

For example, a face identification system is deployed to enter a restricted area (e.g., power plant premises) through face scanning technology; the 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.

Recital 17 AI Act clarifies that this exclusion from the scope of the prohibition is justified by the fact that such systems are likely to have a minor impact on 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 in order to facilitate significantly the identification of natural persons without their active involvement. For active involvement, it is not sufficient that persons are informed about the presence of cameras, but they need to step actively and consciously in front of a camera that is installed in a way fostering active participation.

For example, - RBI systems that are used in cameras installed at walls or ceiling of metro stations for surveillance purposes. Such a system fulfils the condition of remoteness. - Systems that are used to give access to the metro station, such as biometric metro tickets, where persons are actively involved and consciously approach the biometric sensor to obtain access, do not fulfil that condition.

Biometric recognition systems that process (contactless) fingerprints, gait, voice, DNA, keystrokes and other (biometric) behavioural signals may also constitute RBI systems.

For example, - A voice biometric technology system may be deployed to identify a person speaking. The microphone then collects the biometric sample. - A gait recognition system may be used via CCTV and the videos are automatically checked for matches with previously captured templates.

- Keystroke biometric technology may be used to identify the person typing a fraudulent message.

The fact that these systems are given as examples of RBI systems does not imply that they are prohibited under Article 5 AI Act.

In the case of body-cams capable of RBI used by individual law enforcement agents, the untargeted filming during, for example, a demonstration with hundreds of participants will be considered to fulfil the condition of remoteness. c) Reference database

Identification is not possible without a reference database containing biometric data for comparison purposes. Thus, the existence of a reference database is indispensable to perform the comparison for identification purposes.

For example, in the case of missing persons, the Schengen Information System database could be used as the reference database for facial recognition purposes (once operational).

9... Real-time

Real-time means that the system captures and further processes biometric data ‘instantaneously, near-instantaneously or in any event without any significant delay.’ All the processing steps, i.e. the capture, comparison, and identification of biometric data, occur simultaneously or almost simultaneously, which may include a ‘limited short delay’ to avoid the prohibition being circumvented through the retrospective use of RBI systems. The notion of ‘without a significant delay’ is not defined in the AI Act; it will have to be assessed on a case-by-case basis. As the devices used for real- time or post-remote identification are increasingly one and the same with different functionalities, the distinction is temporal. Generally speaking, a delay is significant at least when the person is likely to have left the place where the biometric data was taken.

Real-time systems in general are used at a given place to facilitate a quick reaction and not to retrospectively identify persons. They provide the user of the system with a means to track the movements of persons under surveillance and to monitor them. a) An AI system screens all incoming visitors to a concert venue: Real-time RBI b) A system films all incoming visitors to a concert. An incident happens at the concert. After the concert, the identification system is operated on the video material in order to identify the offender: Post-RBI.

When a law enforcement authority covertly takes a picture of a person via a mobile device and submits it to a database for immediate search, depending on the circumstances, this may fall under the prohibition of Article 5(1)(h) AI Act. 9... In publicly accessible spaces

Article 3(44) AI Act defines publicly accessible spaces as any publicly or privately owned physical space accessible to an undetermined number of natural persons, regardless of whether certain conditions for access may apply, and regardless of the potential capacity restrictions.

Recital 19 AI Act lists several elements that characterise such spaces: - Accessibility to an undetermined number of persons, independently of the potential capacity or security restrictions, such as purchasing a ticket or title of transport, prior registration or having a certain age. The possibility of getting access to a space through an unlocked door does not mean that the space is publicly accessible if indications or circumstances suggest the contrary (such as a sign restricting access). Moreover, the access to a space can be limited to certain persons, as defined by law, linked to public safety or security, or to the decision of the person having the relevant authority over the space.

For example, publicly accessible spaces are in principle:

- a concert venue for which participants pay an entrance fee. - an event location where a trade fair is organised targeting participants over the age of 50.

A space closed by a gate, even if the gate is unlocked, such as the gated entrance of a fenced residential area of several houses, will normally not be considered a publicly accessible space. By contrast, a park in a gated residence with public opening hours without any access restrictions during those hours will generally constitute a publicly accessible space during those hours and a closed space outside those hours. - Irrelevance of ownership, i.e. a space does not need to be in public ownership for it to be considered as a publicly accessible space.

For example, the space may be owned by a private entity, a public entity, or a public entity and managed by a private party, without impacting the nature of the space.

- No specific activity for which the space is used; a publicly accessible area is not necessarily a space linked to a public service. Moreover, a space linked to a public service may include non-publicly accessible spaces, i.e. the offices of the civil servants working at a municipality.

For example, publicly accessible spaces may be used for commerce, such as shops, restaurants, cafés, etc.; for services, such as banks, professional activities (a doctor’s office as well as an accountant’s office), hospitality (e.g., a hotel), etc.; for sport, such as swimming pools, gyms, stadiums, etc.; for transport, such as bus, metro and railway stations, airports, means of transport, etc.; for entertainment, such as cinemas, theatres, museums, concert and conference halls, etc.; or for leisure or otherwise, public roads and squares, parks, forests, playgrounds.

The following spaces do not constitute publicly accessible spaces within the meaning of Article 5(1)(h) AI Act:

- online spaces, since they do not constitute a physical space within the meaning of Article 3(44) AI Act.

For example, chat rooms, social media, online platforms, etc., are therefore excluded from the scope of the prohibition.

- Certain spaces meant to be accessed by a limited number of persons, such as factories, companies and workplaces with entry control or limitations to relevant employees or service providers, since they are intended to be accessed only by relevant employees and service providers.

For example, a workplace accessible with a badge is in principle not considered a publicly accessible space, whereas an office without entry controls falls may be.

- prisons and border control are not publicly accessible spaces

For example, a border crossing point is not a publicly accessible space, while the street leading to the border crossing point or a forest in the vicinity normally is.

Some spaces can have a dual function. For example, an airport is generally considered a publicly accessible space as regards its common areas, but the area dedicated to border control (where the customs officials stand and passports or ID checks occur) is excluded from the scope of the prohibition.

As clarified in Recital 19 AI Act, assessing whether a space is accessible to the public should be done based on a case-by-case analysis.

9... For law enforcement purposes

The prohibition in Article 5(1)(h) AI Act applies to the use of RBI systems for law enforcement purposes, irrespective of the entity, authority, or body carrying out the law enforcement activities.

Law enforcement is defined in Article 3(46) AI Act as the ‘activities carried out by law enforcement authorities or on their behalf for the prevention, investigation, detection or prosecution of criminal offences or the execution of criminal penalties, including safeguarding against and preventing threats to public security.’ These purposes are the same as those listed in Article 1 LED. Thus, any interpretation of those purposes in relation to LED may also be relevant for the purpose of interpreting the notion of ‘law enforcement’ used in the AI Act.

Law enforcement purposes comprise the investigation, detection, and prosecution of criminal offences. They also comprise activities in relation to the prevention of criminal offences, including safeguarding against and the prevention of threats to public security, before any crime has actually been committed. For instance, the police may take ‘coercive measures at demonstrations, major sporting events or riots’ in the context of crime prevention. Finally, those activities comprise the execution of penalties, such as the execution of sentences.

According to Article 3(46) AI Act, law enforcement activities may be performed by law enforcement authorities or on their behalf. Law enforcement authorities are further defined in Article 3(45) AI Act in the same manner as national competent authorities are defined in the LED. That definition covers law enforcement authorities and entrusted bodies or entities (which may be private parties):

(a) any public authority competent for the prevention, investigation, detection or prosecution of criminal offences or the execution of criminal penalties, including the safeguarding against and the prevention of threats to public security; or

For example, such public authorities include police authorities and criminal justice authorities (such as prosecutors) when they carry out a law enforcement task.

(b) any other body or entity entrusted by Member State law to exercise public authority and public powers for the purposes of the prevention, investigation, detection or prosecution of criminal offences or the execution of criminal penalties, including the safeguarding against and the prevention of threats to public security;

Under the AI Act other entities, bodies, or persons, may exercise law enforcement activities after being entrusted by Member States law entrusting them public authority and public powers for the purposes listed above.

‘On behalf of’ means that a law enforcement authority has delegated the performance of a law enforcement activity (or part of it) to another entity or person, including private parties, or has requested in specific cases another entity or person to act to support law enforcement activities. In both cases, the law enforcement authorities must instruct on all major aspects and supervise the other entity, as this requirement is inherent to the notion of acting ‘on behalf’ of a person.

Delegation of tasks to other bodies may include, for example, - Public transport companies requested by law enforcement authorities to ensure security on the public transport networks under their instructions and supervision.

- Sports federations requested by law enforcement authorities to act under their instructions and supervision to provide security at sporting events. - Banks that are requested by law enforcement authorities to conduct certain actions to ‘counter certain crimes in specific cases’ under the instructions and supervision of law enforcement authorities. These activities fall within the definition of “for the purpose of law enforcement” since those entities act ‘on behalf’ of law enforcement authorities. If those entities act on their ‘own behalf’ when detecting and countering crimes (such as fraud, money laundering), they will not be considered to fall under the prohibition of Article 5(1)(h) AI Act.

Only when those other bodies or entities have been entrusted with a specific law enforcement task will their activities fall under the definition of ‘law enforcement’.

The AI Act provides three exceptions to the general prohibition on the use of real-time RBI in publicly accessible spaces for law enforcement purposes. Article 5(1)(h)(i) to (iii) AI Act exhaustively lists three objectives for which real-time RBI may be authorised, while Article 5(2) to 5(7) AI Act lays down the conditions and safeguards for such authorisation. Article 5(1)(h)(i)-(iii) AI Act does not in itself constitute a legal basis for the real-time use of RBI systems in publicly accessible spaces. Rather, only a domestic Member State law that fulfils, in particular, the requirements in Article 5(2)- (7) AI Act can allow the use of real-time RBI, as provided by Article 5(2) AI Act. Consequently, in the absence of Member State legislation authorising the use of real- time RBI for one or more of those objectives, such use is prohibited as from 2 February 2025.

9... Rationale and objectives

The objectives set out in Article 5(1)(h)(i)-(iii) AI Act aim to allow the use of certain AI and investigative tools for law enforcement purposes. These objectives are: (i) the targeted search of victims of three specific serious crimes and missing persons [protection];

(ii) the prevention of imminent threats to life or physical safety or a genuine threat of terrorist attacks [prevention]; and (iii) the localisation or identification of suspects and offenders of certain serious crimes as listed in Annex II [prosecution/investigation].

In those scenarios, the Union legislature has balanced the security needs of society against the risk that real-time RBI systems pose to the fundamental rights of individuals subject to those systems. According to Recital 33 AI Act, the objectives for which the use of real-time RBI systems for law enforcement purposes in publicly accessible spaces is allowed must be strictly, exhaustively, and narrowly defined, and appear when there is a ‘strict necessity’ to achieve ‘a substantial public interest’ which ‘outweighs the risks’ posed to fundamental rights. Any other use of real-time RBI systems in publicly accessible spaces for law enforcement purposes which is not listed in Article 5(1)(h)(i)-(iii) AI Act is prohibited.

For instance, the use of real-time RBI systems by the police to identify a shoplifter and compare their facial images against criminal databases is prohibited, as it does not fall under any of the objectives listed in Article 5(1)(h)(i)-(iii) AI Act. 9... Targeted search for the victims of three serious crimes and missing persons

According to Article 5(1)(h)(i) AI Act, the use of real-time RBI in publicly accessible spaces for law enforcement purposes is allowed, subject to strict necessity and the conditions in Article 5(2)-(7) AI Act, for the targeted search of victims of abduction, trafficking in human beings, or sexual exploitation of human beings, as well as the search for missing persons.

a) Targeted search for victims of three types of crimes

The scenario described in Article 5(1)(h)(i) AI Act seeks to assist law enforcement authorities to search for victims of three serious crimes.

A targeted search would involve the localisation and identification of victims. Three types of crimes

The targeted search for specific victims of three serious crimes are covered by the scenario listed in Article 5(1)(h)(i) AI Act: the abduction of, trafficking in, and sexual exploitation of human beings. If, for example, a child is kidnapped and there are concrete indications that the kidnapper intends to bring the child from one place to another by car, the police may use a real-time RBI system for the targeted search of that child, but it must define a perimeter of deployment and duration of use to identify the child. b) Searching for missing persons

The first scenario also covers the search for a missing person.

A distinction may be made between missing children and missing adults, since the voluntary disappearance of a missing adult will not always trigger a search. The applicable rules regarding missing children vary considerably from one Member State to another. In any event, Article 5(1)(h)(i) AI Act only allows the use of a real-time RBI system to search for missing persons for law enforcement purposes.

The disappearance of an adult does not always lead to a search of that person by police, as adults have the right to disappear. A search could be linked to the legal status of the person (‘under curatorship’), their health condition (a mental illness), the existence of a suicidal note, but also the departure without personal belongings. If the circumstances of the disappearance are a cause for concern, the disappearance may be filed with the police so that a search can start.

In some Member States, the search for a missing person may occur under an administrative procedure and not for law enforcement purposes. For example, where a vulnerable person is missing, but there is no suspicion of a crime or any other law enforcement purpose, the use of real-time RBI systems to search for that person would not be deemed to be for law enforcement purposes and would therefore fall under the rules for such use under the GDPR.

9... Prevention of imminent threats to life or terrorist attacks

Article 5(1)(h)(ii) AI Act lists the second scenario in which the use of real-time RBI in publicly accessible spaces for law enforcement purposes is allowed, subject to strict necessity and the conditions established in Article 5(2)-(7) AI Act the prevention of a specific, substantial and imminent threat to the life or physical safety of natural persons or a genuine and present or genuine and foreseeable threat of a terrorist attack.

a) Specific, substantial and imminent threat to life or physical safety of natural persons

In application of Article 2 of the Charter, which guarantees the right to life, the Union and its Member States must safeguard and, thus, protect the lives of individuals. The criteria in Article 5(1)(h)(ii) AI Act concerning the threat to life to allow for the use of real-time RBI systems in publicly accessible spaces require the existence of (1) a specific, (2) substantial and (3) imminent threat to the life or physical safety of (4) natural persons. The threat does not need to be limited to identified individuals or a group, as it relates to natural persons in general.

Recital 33 AI Act clarifies that an imminent threat to life or the physical safety of natural persons may also include an imminent threat to critical infrastructure ‘where the disruption or destruction of such critical infrastructure would result in an imminent threat to the life, or the physical safety of a person, including through serious harm to the provision of basic supplies to the population or to the exercise of the core function of the State.’

For example, A serious disruption and destruction of critical infrastructure (e.g., a power plant, water supply, or a hospital) may result in an imminent threat to the life or the physical safety of a person when there is serious harm of cessation of basic supplies to the population (deprivation of electricity or drinkable water for a long period, in a particularly warm or cold weather, etc).

What constitutes an imminent threat to life or the physical safety of natural persons is ultimately defined and assessed at the level of the Member State based on its national laws, in accordance with EU law, in particular taking into account the key elements and rationale of Article 5 AI Act. This will have to be laid down/referred to in the laws Member States must adopt to make use of the exceptions to the prohibition on the use of real-time RBI for law enforcement purposes in publicly accessible spaces.

An imminent threat to life or physical safety is a threat that can occur at any moment and requires ‘immediate action to be taken.’ A substantial threat to physical safety relates to serious bodily injuries.

A specific threat means that the threat is clearly defined, individualised and concrete, in that it should not be hypothetical or relate to certain dangers in general. For example, the police are informed that a former student plans a deadly attack at his former university as he seeks revenge on several former classmates. The police receives information about the imminence of the attack, the targeted school, and the weapons he plans to use to execute his plans.

A specific threat needs not be intentional. Non-intentional actions could also result in a threat to life or physical safety.

b) A genuine and present or genuine and foreseeable threat of a terrorist attack

This part of the second scenario described in Article 5(1)(h)(ii) AI Act is comprised of several elements: the existence of a threat of a terrorist attack and the characteristics of the threat, which must be genuine and present or genuine and foreseeable. Threat of a terrorist attack

The assessment concerning the existence and seriousness of the threat is made at national level when assessing the actual circumstances of a measure to be taken to safeguard national security, and more specifically, in case of a terrorist attack. The terrorist threat level is defined at national level and varies from one Member State to another. For example, the Netherlands has established five levels of threats, Belgium four, France three, and Sweden five. However, the concept of ‘a genuine and present or genuine and foreseeable threat’, as used in Article 5(1)(h)(ii), is an autonomous notion of Union law and should therefore be assessed, in principle, independently of national definitions. The threat relates not to terrorism in general, but specifically to a threat of a terrorist attack.

Characteristics of the threat: genuine and present or genuine and foreseeable

The threshold of seriousness that a threat needs to reach to allow for the use of real- time RBI systems in publicly accessible spaces for law enforcement purposes was inspired by the CJEU’s case law on data retention and passenger name record measures aimed at safeguarding national security, in particular, against terrorist attacks. According to the CJEU, in those contexts, ‘a threat to national security must be genuine and present, or at the very least, foreseeable, which presupposes that sufficiently concrete circumstances have arisen.’ Prevention

Contrary to Article 5(1)(h)(i) and Article 5(1)(h)(iii) AI Act, the scenario described in Article 5(1)(h)(ii) does not specify that the use of real-time RBI is permitted to locate or identify a concrete person. Its purpose is the prevention of a particular threat. Accordingly, the scenario may also cover the use of real-time RBI to detect and follow ‘terrorists on the move’, i.e. several persons linked to the same threat, if there are concrete indications that those persons plan to commit a terrorist attack, but it is not clear where.

Real-time RBI to prevent a terrorist attack in a park The police are informed that a person is running around a park looking for people to attack with a knife while he is screaming violent extremist slogans usually linked to terrorist attacks and terrorist groups. If the Member State has authorised the use of real-time RBI in the scenario described by Article 5(1)(h)(ii) AI Act, law enforcement authorities may use real-time RBI to identify and locate the person in and around the park to prevent the attack, provided the other conditions of Article 5(2)-(7) AI Act have been met.

9... Localisation and identification of suspects of certain crimes

Article 5(1)(h)(iii) AI Act allows the real-time use of RBI in publicly accessible spaces for ‘the localisation and identification of a person suspected of having committed a criminal offence, for the purpose of conducting a criminal investigation or prosecution or executing a criminal penalty for offences referred to in Annex II and punishable in the Member States concerned by a custodial sentence or a detention order for a maximum period of at least four years.’

Annex II AI Act provides an exhaustive list of serious crimes for which the use of real- time RBI may be authorised for the aforementioned objective. Those criminal offences are:

- terrorism, - trafficking in human beings, - sexual exploitation of children, and child pornography, - illicit trafficking in narcotic drugs or psychotropic substances, - illicit trafficking in weapons, munitions or explosives, - murder, grievous bodily injury, - illicit trade in human organs or tissue, - illicit trafficking in nuclear or radioactive materials, - kidnapping, illegal restraint or hostage-taking, - crimes within the jurisdiction of the International Criminal Court, - unlawful seizure of aircraft or ships, - rape, - environmental crime, - organised or armed robbery, - sabotage, - participation in a criminal organisation involved in one or more of the offences listed above.

a) Localisation and identification

A Member State may authorise the use of real-time RBI in publicly accessible spaces for law enforcement purposes to locate and identify a suspect of a criminal offence to conduct a criminal investigation, prosecute that person for the crime committed, or execute an existing sentence.

b) Suspects and Perpetrators

Article 5(1)(h)(iii) AI Act covers two categories of individuals: suspects and perpetrators. A suspect is a person with regard to whom there are serious grounds for believing that they have committed a criminal offence, and sufficient evidence of that person’s involvement in the offence has already been gathered. A perpetrator is a person who is accused or convicted of having committed a criminal offence. The same conditions (crime listed in Annex II and maximum punishment of at least four years) apply to locate or identify the accomplice of the crimes listed in Annex II AI Act. c) List of serious crimes

Only serious crimes justify the use of real-time RBI systems in publicly accessible spaces for law enforcement purposes.

The first five offences listed in Annex II AI Act are the same as the ‘euro crimes’; listed in Article 83 TFEU, while the other offences constitute priorities for law enforcement cooperation. Some of them (e.g., kidnapping, illicit trafficking in nuclear or radioactive materials) may be linked to terrorism.

Although all the criminal offences listed in Annex II may trigger the issuance of a European Arrest Warrant (‘EAW’) against a suspect or perpetrator, the use of real-time RBI to locate and identify a suspect for one of these serious criminal offences does not require that an EAW has been issued.

Moreover, to use real-time RBI for this purpose, the respective criminal offence must be punishable in the Member State concerned by a custodial sentence or a detention order for a maximum period of at least four years.

During a busy festival in a city, police authorities deploy live facial recognition technologies to monitor the area around the festival and identify wanted individuals with outstanding arrest warrants for illegal drug trafficking and sexual offences. At different entrances to the festival, the police use live video footage of people passing in front of a mobile camera to compare their faces with a watchlist of faces of wanted individuals.

First, concerning the offence types, RBI can be used in case of illegal drug trafficking. However sexual offences are not on the list of offences, unless they relate to the sexual exploitation of children, child sexual abuse material, or rape. The police are not allowed to deploy real-time facial recognition technologies in a broad, untargeted manner, i.e. in the hope of finding wanted criminals and taking them off the streets.

The case is different if the police have received a physical description with a photograph of a wanted individual that is subject to a European Arrest Warrant for drug trafficking and they have reasons to believe that he will be present at the festival. In those circumstances, deploying real-time facial recognition technologies to identify a targeted individual may be covered by Article 5(1)(h)(iii) AI Act.

After a serious terror attack at a Christmas market with 12 deaths, the police uses real-time facial recognition technologies to identify the offender and to see to where he is escaping. In that context they also use the real-time facial recognition technologies of the nearby train station and at destination stations of the trains leaving from there shortly after the attack. In the case of a terror attack, such use can be permitted under Article 5(1)(h)(iii) AI Act.

A link between Article 5(1)(h)(i) and Article 5(1)(h)(iii) AI Act may be made for the crimes covered by the scenario described in Article 5(1)(h)(i) AI Act. While real-time RBI systems may be deployed to find a victim or a missing person, those systems may also be used to locate and identify the perpetrator or suspect of trafficking in human beings, sexual exploitation as far as it concerns children (as listed in Annex II), and kidnapping (as far as the abduction mentioned in Article 5(1)(h)(i) AI Act qualifies as kidnapping as listed in Annex II AI Act). A link may also be made between Article 5(1)(h)(ii) and (iii) AI Act: real-time RBI systems may be used to prevent a threat falling within the scope of Article 5(1)(h)(ii) and, if that threat materialises, those systems may be used to identify/locate the perpetrator ‘on the move’.