French Gendarmerie Used Facial Recognition Illegally During 2023 Riots
A report by French inspection bodies published in October 2024 confirmed that the gendarmerie illegally used facial recognition through the Briefcam software during the 2023 Nanterre riots to identify alleged vandals. The facial recognition function was activated by default in the software and lacked legal safeguards, violating French law that prohibits such use without explicit consent or legal authorization.
What happened
In summer 2023, during the Nanterre riots, the Service Central du Renseignement Criminel of the gendarmerie used the Briefcam facial recognition tool to identify alleged vandals in Fosses. Investigators injected two photographs of suspects from the criminal record file (TAJ) into the Briefcam software, which identified matches using facial recognition analysis.
Briefcam deployed a facial recognition function in November 2018 that was activated by default and could only be disabled by an administrator. The software had been in use since 2017 in the gendarmerie for video analysis purposes, and the facial recognition feature was added without specifically informing users of its presence.
Reported impact
The report states that although the individuals were identified by the facial recognition function, subsequent investigation showed their presence at the vandalism sites was unrelated to any participation in such acts, and they were not charged. The practice violated Article 6 of French data protection law, which prohibits facial recognition without the consent of the individual or legal authorization.
Index analysis
The report notes there were no legal safeguards or oversight mechanisms governing the use of facial recognition by the gendarmerie. Investigators acknowledged the use of the function, and the commission found their conduct was undertaken in good faith but that the activation of the feature could not be independently verified, as system logs are automatically deleted after 365 days.
Governance controls
The report criticizes the absence of user training on the facial recognition functionality and the lack of warning to users when the feature was deployed. The report found that the presence of an active facial recognition function in software historically ordered for non-recognition purposes created a legal vulnerability for users.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.