The EU’s AI Act aims to ban “unacceptable-risk” AI, yet leaked COREPER papers show France engineered major loopholes. Police and border guards may keep using facial and emotion recognition under a national-security clause. Backed by Italy, Sweden, Finland and others, Paris ensured real-time facial recognition stays legal for 16 offences (including “environmental crime”) and could be widened.
Another article lets authorities source data from private firms or third countries, a move legal experts say clashes with EU court rulings. The ban on biometric categorisation was likewise softened for law-enforcement use. A late self-certification clause allows vendors to decide if their tools are “high risk”, undermining oversight according to the European Parliament’s legal service. Civil-rights groups fear mass surveillance and say the Act prioritises Europe’s AI market over privacy.
↗ Full article at Netzpolitik.org
Methodology:
- As part of this investigation, we evaluated internal EU Council minutes
- We also spoke with individuals who were involved in the negotiations