France Travail tests AI system to profile and prioritize surveillance of job seekers
France Travail, the French employment agency, is testing an AI algorithm that automatically identifies and profiles job seekers for targeted surveillance and compliance checks based on 26 variables including employment history, training level, and declared activities. The system has raised concerns about potential discrimination, bias, opacity, and mass surveillance of over 6 million job seekers, particularly following the mandatory registration of RSA benefit recipients.
What happened
France Travail presented to its AI ethics committee on 10 December 2025 a document describing an algorithm called CRE (Contrôle de la Recherche d'Emploi) designed to automatically identify job seekers for targeted surveillance. The algorithm was trained on 60,000 previous control cases and uses 26 variables including registration tenure, education level, history of non-compliance, and declared activities to sort profiles as either 'suspect' or 'non-suspect'.
The algorithm is currently being tested on profiles of people in conventional job termination situations, but France Travail intends to expand its use to other populations, including the more than 6 million job seekers now obligated to register following mandatory RSA benefit recipient registration as of 1 January 2026.
Reported impact
Despite France Travail's claims that the algorithm removes bias through supposedly objective criteria, concerns have been raised about potential discrimination, opacity, and mass surveillance of vulnerable populations. France Travail itself acknowledged that when the model was applied to current job seekers, it exhibited biases, with the algorithm achieving 79% accuracy in identifying cases requiring in-depth review and 64% accuracy in predicting sanctions or 'redynamization' measures.
Index analysis
The algorithm operates as a decision tree analyzing 26 variables to categorize job seekers for priority monitoring. The system's training data consisted of 60,000 historical control cases, which may perpetuate existing patterns of surveillance and enforcement.
Governance controls
The algorithm was presented to France Travail's AI ethics committee, indicating some internal oversight structure. However, the system remained undisclosed publicly until revealed by La Quadrature du Net in collaboration with Radio France investigation.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.