Under the AI Act, the authorities may only use real-time biometric recognition in three circumstances: to search for missing persons or victims of abduction and human trafficking, to prevent an imminent terrorist threat, or to identify the perpetrators of a limited list of particularly serious offences
Brussels – How far can artificial intelligence go in monitoring citizens? This is the question that has reignited the European debate on facial recognition following the European Commission’s warning in Italy. “The AI Act prohibits real-time biometric identification in public spaces, barring strictly limited exceptions,” explained a spokesperson for the EU executive during the usual daily press briefing. “We do not want artificial intelligence to be used to monitor people’s daily movements in accessible public spaces, so that is a clear prohibition.” This principle is once again taking centre stage as the Meloni government pushes ahead with the legislative process for the national decree, which is intended to implement the new European rules on artificial intelligence.
Facial recognition was one of the most divisive issues in the AI Act, the world’s first regulatory framework dedicated to artificial intelligence. During the negotiations, the European Parliament had called for a total ban on biometric identification in public spaces. At the same time, several Member States have pushed to ensure greater scope for action by law enforcement agencies. The final compromise retained the ban as the general rule, while introducing some strictly limited exceptions.
From 2 February 2025, when the AI Act came into force, authorities may use real-time biometric recognition only in three circumstances: to search for missing persons or victims of abduction and trafficking, to prevent an imminent terrorist threat, or to identify the perpetrators of a limited list of particularly serious offences, such as terrorism, murder, or human trafficking. Even in these cases, prior authorisation from a judicial or an independent authority is required, except in duly justified emergencies. The AI Act also prohibits the creation of biometric databases obtained by indiscriminately collecting images from the internet or from CCTV systems.
The dispute with Italy has arisen precisely on the boundary between these exceptions and the general ban. The legislative decree through which the government intends to bring national legislation into line with the AI Act is now under scrutiny by parliamentary committees. Still, certain provisions have raised doubts as to their compatibility with the European regulation. Opposition parties have demonstrated outside Palazzo Chigi under the slogan “No to Italy’s Big Brother”, accusing the government of seeking to introduce a surveillance system incompatible with the safeguards established by the European Union.
Article 8 of the decree essentially reiterates the exceptions provided for in the AI Act, permitting real-time facial recognition only in exceptional casesand subject to authorisation by the public prosecutor, with a fast-track procedure in urgent situations. More controversial, however, is Article 10, which regulates the use of artificial intelligence in existing CCTV systems. The text provides that, following the commission of a crime, software may be used to identify individuals already under suspicion. However, it also introduces the possibility of collecting and storing, for seven days, the biometric data of people passing through locations deemed sensitive – such as squares, stadiums or public events – so that this data can be compared at a later date should a crime be committed.
It is precisely this last passage that has sparked the debate. According to the government, data collection and data recognition are two distinct stages: biometric processing would only take place after an offence has been committed. However, this interpretation does not satisfy the Data Protection Authority, which has called for the text to explicitly state that data analysis may only take place after a crime has been committed, and has deemed the definition of the locations where pre-emptive collection would be permitted to be too broad, raising concerns about the risk of mass surveillance.
According to MEP Brando Benifei (PD/S&D), rapporteur for the AI Act, the decree risks “allowing the mass collection of biometric data in public spaces, contrary to the spirit of European legislation.” Benifei has announced an urgent question to the European Commission asking what measures it intends to take against the Italian government to ensure citizens’ rights are upheld.
Having set limits and restrictions on the use of artificial intelligence, Brussels will need to ensure that the delicate balance between security and fundamental rights is upheld in national legislation as well. The leader of the PD delegation to the European Parliament, Nicola Zingaretti, shares this view, arguing that “security cannot become a pretext for curtailing fundamental freedoms and rights.” Europe has been clear: indiscriminate facial recognition in public places is prohibited. The real challenge is to use artificial intelligence to protect people, not to turn them into subjects under constant surveillance.”
Eunews. English version by the Translation Service of Withub
