On December 8, the European Union signed its first comprehensive framework for artificial intelligence. Politically hailed as a historic milestone by negotiators, this agreement also drew warnings from civil society groups that crucial technical questions remained unresolved. The finer details, as is often the case, would prove decisive.
Weeks later, while the law is still awaiting final approval, an internal document reviewed by Politico raised concerns about the potential for what it described as “irresponsible and disproportionate” use of AI technologies, including biometric identification and facial recognition. German lawmaker Svenja Hahn labeled the loophole not included in the original treaty as an attack on European human rights, suggesting it could resemble the authorities seen in some authoritarian states.
A legal text prepared by the Spanish Presidency of the Council on December 22 indicated that real-time biometric surveillance would be banned. Yet the document also left room for exceptions where police and the military could deploy the technology, provided there is a formal legal authorization to counter threats such as terrorism or violent crime.
As Hahn notes, late-stage changes to European rules could permit law enforcement to use this technology on existing video recordings without prior judicial approval, extending beyond what the three primary law enforcement agencies had previously accepted. The balance between safety needs and civil liberties remains a central point of contention for the EU.
police pressure
The leaked findings, obtained by El Periódico de Catalunya from Grupo Prensa Ibérica, suggest that only administrative authorization would be required to extend the use of biometric identification and tracking across populations. This would speed up the deployment of such systems based on unique physical patterns, including facial features, fingerprints, and iris scans.
Ella Jakubowska, senior policy adviser at European Digital Rights, commented that the push for broader use appeared to come strongly from law enforcement circles, particularly in France. She noted that some EU member states struggled to accept a retroactive facial recognition framework, arguing it would be more restrictive than a real-time regime and could hinder routine police work.
“It has no place in democracy”
European negotiators have been drafting the detailed provisions of the AI regulation for weeks. This Friday, representatives from the European Commission, the European Parliament, and member states were set to reconvene to discuss the obligations placed on AI technologies treated as high risk, including facial recognition. While Parliament pressed for stronger protections, governments of the 27 member states managed to scale back some demands.
Civil groups, including Reclaim Your Face, criticized the apparent weakening of safeguards, arguing that biometric mass surveillance is prone to error and inherently risky in a democratic society.
Last September, more than a hundred European organizations urged a cautious approach to using AI for immigration control, warning that such measures disproportionately affect already marginalized communities and undermine legal and procedural rights through mass surveillance.
EU governments are expected to receive the final text of the AI regulation next Wednesday, January 24, with the goal of adoption on February 2. Even then, the deal will likely need approval from Parliament, which maintains a final opportunity to propose amendments before a formal vote.