France Travail's 'Ciblage du Contrôle' — an AI Tool Sorting Job Seekers into 'Suspect' and 'Non-Suspect' for Control Priority

France's national employment agency, France Travail (formerly Pole emploi), developed an algorithmic profiling tool named 'Ciblage du Controle de la Recherche d'Emploi' (Job-Search Control Targeting) to help select which job seekers' files receive an in-depth compliance review.

Prioritise job-seeker files for human compliance review by sorting them into 'suspect' or 'non-suspect' categories, as part of a broader overhaul of the control process ('CRE renove').

The tool uses a decision-tree model trained on roughly 60,000 past control cases, scoring each job seeker's file on 26 variables, including declared job-search activity, whether a CV is visible online, targeted occupations, education level, registration tenure and prior non-compliance. It was piloted from July 2024 in eight regions, generalised nationwide by June 2025, and presented to France Travail's internal AI ethics committee on 10 December 2025.

Algorithmically-targeted controls now account for roughly 40% of all controls, about 500,000 of the total in 2025, against a national policy backdrop of controls rising from 200,000 in 2017 toward a stated target of 1.5 million by 2027.

La Quadrature du Net reports that France Travail has refused to disclose the tool's source code, which the organisation says makes independent bias auditing impossible; France Travail's own internal documentation reportedly acknowledged a risk that the model could initially overrepresent job seekers receiving the lowest benefit amounts. This is included as a cautionary case: the operational scale is well documented, but there is no independent, published accuracy or fairness evaluation of the algorithm.