While the EU Commission wants to regulate AI applications for police and justice, the current Council Presidency is in favour of as few restrictions as possible. The processing of facial images from public spaces is becoming a bone of contention.
The Portuguese EU Presidency is questioning a blanket ban on artificial intelligence (AI) for facial recognition in public spaces. The Commission had presented such a proposal on 21 April, but also named a number of exceptions and associated obligations.
In a discussion paper distributed to the member states, this approach is now criticised by the Council. Instead, it is according to the Presidency “essential to ensure that we are not not unnecessarily limiting the development and use of technological development”. The police and judicial use of AI must be “be practical, useful and improve the efficiency with which law enforcement authorities work”.
“High risk” in the area of migration, asylum, border control
The EU Commission had suggested adopting the “the first-ever legal framework” on AI systems. These are divided into different categories. Accordingly, AI-supported video games or spam filters is considered a “minimal risk” for civil rights or security, and the regulation should not intervene there. Self-learning chatbots pose a “low risk”; users should therefore only be informed about the function. An “unacceptable risk” exists for technology that classifies social behaviour or toys with voice assistants.
According to the proposal, AI systems used in critical infrastructures, human resources management or credit rating pose a “high risk”. AI-supported software in the area of migration, asylum and border control falls into the same category, for example when it checks the authenticity of travel documents. According to the Commission, “adequate mitigation and control measures” should then apply. Data records fed into the system would have to be checked for quality and all processes would have to be logged.
In principle, the Commission wants to ban “all remote biometric identification” for law enforcement in public spaces. This refers, for example, to real-time facial recognition with cameras at airports or in railway stations.
Warning of losses for security industry
At the same time, however, the Commission restricts the ban significantly and lists a large number of exceptions in which the police and judiciary are allowed to use such systems. These include cases in which applications are “strictly necessary” to find missing children or to avert a “terrorist attack”. The prosecution of serious crimes should also be possible after a competent judicial authority has issued the necessary order.
The paper from Portugal not only questions the blanket classification of various functions as “high-risk”. Also, the exceptions and obligations mentioned may “not be realistic and feasible at operational level”. The Council Presidency does not get more specific; the criticism is presumably also aimed at the prescribed judicial decision, which, according to the Commission’s proposal, could only be waived for the time being in “in a duly justified situation of urgency”.
Elsewhere, the Council paper also warns of financial losses for companies in the security technology sector. Accordingly, the regulation proposed by the Commission could hinder innovation. Therefore, Europol is now to enter into a “dialogue” with the private sector. For this purpose, the police agency operates a new “EU Innovation Centre for Internal Security”, which is to identify and push priorities in security research.
“Biometric Identity Repository”
Finally, the Presidency also addresses formal aspects of the Commission’s regulatory proposal. For example, some AI applications that are considered high-risk and are currently used by police and judicial authorities, or would be at the time of the entry into force of such a regulation, might not fall within its scope.
This is because before the EU adopts a regulation on AI, other new systems have long since gone into operation. In the “Interoperability” project, the Agency for the operational management of large-scale IT systems (eu-LISA) is currently setting up a biometric “Common Identity Repository” in which fingerprints and facial images of all large EU information systems will be combined. The so-called Prüm mechanism is also being extended from fingerprints to facial images. Soon it will be possible to query the photo of an unknown person in all EU member states at the same time. The paper of the EU Presidency also deals with this. Thus, a future AI regulation should take into account “all relevant phases” of the processing of facial images in the Prüm procedure.
The Portuguese Presidency’s surveillance initiative comes at the same time that 180 global civil rights organisations are speaking out against facial recognition in public spaces. They are calling on governments to halt all investment in the use of remote biometric recognition technologies and ban their use in publicly accessible spaces. With the same goal, the European citizens’ initiative “Reclaim Your Face” has already collected over 54,000 signatures.
Image: John Wisniewski, CC-BY-ND 2.0
Leave a Reply