The French Caisse nationale des allocations familiales (CAF) has deployed a new algorithm to verify the eligibility of its benefit recipients [1, 2].
This shift toward automated scoring represents a significant change in how the state monitors social welfare. The system effectively rates individuals to detect potential fraud and determine how aid is allocated, raising questions about the transparency of the criteria used to flag citizens for investigation.
The CAF introduced the system in October 2026 [2]. According to the agency, the tool is designed to fight fraud and ensure that more money reaches the families who need it most [2]. By automating the identification of discrepancies in files, the agency aims to streamline the verification process for benefits such as the RSA and APL [1].
A spokesperson for the CAF said, « Nous versons plus d'argent aux familles qui en ont le plus besoin et auprès desquelles le risque d'erreur est amplifié par des situations de vie » [2].
However, the implementation of the algorithm has met with resistance. Some associations have taken the matter to the Conseil d'État to seek the suppression of the anti-fraud algorithm [2]. These critics argue that the automated process may disproportionately target the most precarious beneficiaries, a criticism that has been leveled at previous versions of the agency's screening tools [1].
The agency maintains that the system reduces the risk of error in aid distribution. While the CAF focuses on the efficiency of the rollout, opponents suggest that the scoring system creates an opaque barrier between the state and its most vulnerable citizens [1, 2].
“The CAF has deployed a new algorithm to verify the eligibility of its benefit recipients.”
The transition to algorithmic scoring for social benefits reflects a broader global trend of 'digital welfare states.' By replacing human discretion with automated risk profiling, the French government can process vast amounts of data to curb fraud, but it risks alienating marginalized populations who may be flagged by the system due to complex life circumstances rather than intentional deceit.
