In January 2020, the New York Times revealed the existence of Clearview AI, a startup that had built a database of over 3 billion face photos — collected from Facebook, Instagram, YouTube and other sites — and made them available to police forces for facial recognition. None of the photographed individuals had given consent. The company had no legal obligation to inform them. The database had been sold to hundreds of police agencies in the United States and other countries.
Facial recognition is perhaps the most invasive surveillance technology ever created, because it transforms something we cannot hide — our face — into a universal identifier. Unlike an identity document, which we can choose not to show, or a phone, which we can leave at home, our face is always with us. A camera with facial recognition in a square can identify every person who passes through it, creating a detailed register of who was where and when.
The problem is amplified by the differential inaccuracy of these systems. Joy Buolamwini and Timnit Gebru's "Gender Shades" study demonstrated that Amazon, Microsoft and IBM's facial recognition systems had error rates of 34.7% for dark-skinned women, versus 0.8% for light-skinned men. In a policing context, this means Black people have an enormously higher probability of being erroneously identified as suspects.
Real consequences are already visible. In 2020, Robert Williams, a Black man from Detroit, was arrested based on an erroneous facial recognition identification. He was held for 30 hours before the error was discovered. This is not an isolated case: at least six other cases of wrongful arrests based on facial recognition have been documented in the United States, all involving Black people.
The global response has been fragmented. Some cities (San Francisco, Boston, Portland) have banned police use of facial recognition. The European Union, in the 2024 AI Act, imposed significant restrictions but not an absolute ban. In many countries, no specific regulation exists. Meanwhile, the technology continues to improve and spread — in shopping centres, schools, stadiums, train stations.
Biometrics goes beyond the face: gait recognition, voice analysis, iris scanning, emotion recognition (a pseudoscience that claims to deduce emotional states from facial expressions). Each of these technologies presents the same fundamental risks: surveillance without consent, discriminatory errors and concentration of power in those who control the data.