Algorithme de notation de la CAF

Algorithme de notation de la CAF is the French-language working campaign register through which the Paris-anchored La Quadrature du Net (LQDN) and a coalition of fifteen — subsequently twenty-five — French and European civil-society organisations have carried their multi-year contestation of the algorithmic risk-scoring system operated since 2010 by the Caisse nationale des Allocations Familiales (CNAF), the French national family-benefits agency. The system assigns "scores de suspicion" (suspicion scores) to households covering more than 32 million people, of whom some 13 million are children, and the resulting scores target administrative checks for fraud. The framing operates as a working register — campaign name, substantive policy argument, and rhetorical anchor at once — rather than as a hashtag-anchored slogan, and is the corpus's first French-language message register and its anchor for Continental European civil-society contestation of welfare-algorithm deployment. Its working public-record artefacts are the November 2023 source-code disclosure, the October 2024 filing before the Conseil d'État by the original fifteen-organisation coalition, and the January 2026 expansion of the coalition to twenty-five organisations including European Digital Rights (EDRi), AlgorithmWatch, and the Panoptykon Foundation.

Origin

The framing's working public life begins on 27 November 2023 with LQDN's publication of two obtained versions of the CNAF algorithm source code — the 2010-2014 version and the 2014-2018 version — secured via the Commission d'accès aux documents administratifs (CADA) after some eighteen months of administrative-law effort. The launch banner, "L'indécence des pratiques de la CAF désormais indéniable" (The indecency of CAF's practices now undeniable), reframed the file from a technical debate about algorithmic variables to a moral-and-political register: the substantive argument is not that the algorithm is broken but that it succeeds precisely in systematically scoring the most vulnerable beneficiaries — low-income, single-parent, disabled, and unemployed — for heightened administrative-control suspicion. The framing's rhetorical anchor in the disclosure post, "algorithme de la honte" (algorithm of shame), travels alongside the canonical campaign register "algorithme de notation des allocataires" (beneficiary scoring algorithm) and the working substantive register that the system constitutes a "double peine" (double penalty) under which the same beneficiary populations are both the targets of welfare provision and of the agency's anti-fraud-control suspicion.

The November 2023 disclosure rests on a sustained legal-effort spine. CADA, the French administrative authority charged with arbitrating access-to-public-documents requests, ordered the CNAF to release the algorithm's source code after LQDN's freedom-of-information requests were repeatedly stonewalled or partially redacted. The CNAF's redaction of variable names in the eventually released code is the working evidence in the LQDN publication that the agency understood the substantive sensitivity of the file's disclosure. LQDN's own working register positions the disclosure as the end of what it names "faux débat technique" (false technical debate): the publication of the source code, including the original training pattern, the variable weightings, and the substantive demographic axes that drive the score, is the working evidentiary base on which the subsequent multi-year coalition litigation rests.

Travel into the Conseil d'État litigation

The framing's most consequential downstream artefact in the French legal record is the 16 October 2024 collective filing before the Conseil d'État, France's highest administrative court. The original fifteen-organisation coalition consists of LQDN as coordinator, AADJAM (a French anti-discrimination civil-society organisation working with Roma beneficiaries), Aequitaz (a French equality-and-justice civil-society organisation), Amnesty International France, ANAS (Association nationale des assistants de service social — the French national social-workers' association), APF France handicap (the French national federation of associations of disabled persons), Collectif Changer de Cap (a French civic-coalition working on economic-justice), Fondation Abbé Pierre (the French national anti-poverty foundation, now Fondation pour le Logement des Défavorisés), Gisti (Groupe d'information et de soutien des immigré·e·s, the French national immigrant-support legal-defence organisation), Le Mouton numérique (a French civic-tech collective on the politics of digital systems), La Ligue des droits de l'Homme (the French national human-rights league, founded 1898), MNCP (Mouvement national des chômeurs et précaires, the French national unemployed-and-precarious-workers' movement), MFRB (Mouvement français pour un revenu de base, the French basic-income movement), CNDH Romeurope (the French national human-rights collective for Roma populations), and the Syndicat des avocats de France (the French trade-union of progressive lawyers). The coalition's working sociological shape — anti-poverty, anti-discrimination, disability-rights, social-work, immigrant-defence, basic-income, civic-tech, and lawyers' organisations — is itself the substantive working argument of the case: the populations the algorithm disproportionately scores are precisely the constituencies the coalition represents, and the coalition's joint filing is the substantive working evidence of the convergence between the affected populations and French civil-society's contestation of the system.

The coalition's headline argument before the Conseil d'État is the canonical campaign-register working claim that the system "assimile précarité à suspicion de fraude" (equates precarity with fraud suspicion). The substantive legal-arguments spine runs along three axes: a French and European anti-discrimination axis under the principle of equal treatment in administrative decision-making; a data-protection axis under GDPR Article 22 on the right not to be subjected to a decision based solely on automated processing where the decision produces legal or significantly similar effects; and a proportionality axis under French administrative law on the necessity-and-proportionality test for state administrative interference with the rights of beneficiaries. The substantive working argument Amnesty International cites — through Secretary General Agnès Callamard's quoted line that the system "operates in direct opposition to human rights standards, violating the right to equality and non-discrimination" — translates the campaign's French-language working register into the Anglosphere human-rights-treaty vocabulary the international coalition uses, and is the framing's clearest public-record international-civil-society translation.

The January 2026 European expansion

The January 2026 expansion of the coalition brought the case from a fifteen-organisation French civil-society coalition to a twenty-five-organisation French-and-European one. The ten new joining organisations are the CGT (Confédération générale du travail — the French national trade-union confederation), Union Syndicale Solidaires (the French alternative trade-union federation), FSU TEIOS (a French public-sector trade-union body), Data for Good (the French civic-tech volunteer-data-scientist association), EDRi (European Digital Rights), AlgorithmWatch, the European Network Against Racism, the Panoptykon Foundation, the Mouvement des mères isolées (the French single-mothers' movement), and Féministes contre le cyberharcèlement (the French feminist-anti-cyberharassment collective). The expansion's working sociological reading is that the French trade-union confederations have joined the coalition — supplying the substantive worker-organising and labour-rights register to the file — and that the European digital-rights coalition the November 2023 disclosure originally sat outside has formally joined the litigation, supplying the cross-European civil-society institutional register and the structural comparison to adjacent Continental European welfare-algorithm cases.

The European-coalition expansion's working register is the cross-Continental civil-society infrastructure that the EDRi network membership, AlgorithmWatch's Berlin-anchored algorithmic-accountability work, and the Panoptykon Foundation's Warsaw-anchored Polish-and-Eastern-European-digital-rights work together carry. LQDN's structural counterparts in EDRi on the algorithmic-accountability axis — AlgorithmWatch on the German side and Panoptykon on the Polish side — are themselves the principal national civil-society anchors in their respective countries on adjacent welfare-algorithm and ADM-system files, and the coalition's expansion is the corpus's clearest single working-record instance of cross-Continental European civil-society convergence on a single national welfare-algorithm case. The framing's January 2026 expansion sits inside the same wider Continental civil-society infrastructure on welfare-algorithm contestation that includes the European-record adjacent precedents the Amnesty International statement names: the Dutch childcare-benefits toeslagenaffaire welfare-algorithm scandal and the Serbian welfare-algorithm case.

Working-register grain and rhetorical variants

The framing operates inside the corpus's small set of welfare-algorithm and ADM-contestation framings with a distinctively French-language register grain. Its working canonical phrase, "algorithme de notation des allocataires", names the technical-and-administrative artefact directly — the scoring algorithm and the beneficiaries it scores — and is the working phrase across LQDN's November 2023, March 2024, October 2024, January 2026, and February 2026 publications. Its rhetorical variant "algorithme de la honte" (algorithm of shame) names the substantive moral-and-political accusation directly and is the working phrase for the campaign's mobilisation and public-discourse register; its substantive-argument variant "score de suspicion" (suspicion score) names the algorithm's working operational output and is the working phrase for the file's data-protection and discrimination-rights argument; its structural-pattern variant "double peine" (double penalty) names the substantive sociological pattern the file documents and is the working phrase for the campaign's anti-poverty and disability-rights register. The four registers operate as a single working campaign vocabulary across the multi-year file rather than as competing alternatives, and the framing's working life is the seam between them.

The framing's distinctive feature relative to the corpus's existing Anglosphere algorithmic-contestation framings — notably the UK Fuck the algorithm chant and the broader #BanTheScan and Ban biometric mass surveillance European-and-Indian biometric-surveillance framings — is the operation in French without translation into an Anglosphere coalition register, and the operation on a Continental European welfare-algorithm-contestation register rather than an exam-grading or biometric-surveillance register. Where "Fuck the algorithm" anchored on a single-day Westminster student protest in August 2020 and crystallised into a portable UK working chant, "Algorithme de notation de la CAF" has been built across a multi-year strategic-litigation arc co-extensive with the duration of the file LQDN brought before the Conseil d'État, and operates principally inside the legal-argument and civil-society-coalition register that the Conseil d'État proceedings have organised the campaign around. The framing's distinctive grain is the welfare-algorithm-contestation seat that the corpus's other framings on algorithmic-decision-making in public administration do not anchor.

Travel into the wider LQDN portfolio and the regional litigation field

The framing has carried into LQDN's wider portfolio of French welfare-control-algorithm litigation as the lead-file working register. LQDN's campaign portfolio on French administrative algorithms includes complaints against the France Travail (the French national unemployment agency) algorithmic risk-scoring, the Caisse nationale d'Assurance Maladie (the French national health-insurance fund) algorithmic anti-fraud-control system, and the French tax administration's risk-scoring; the CAF file anchors the working framing the portfolio is built around. The March 2024 escalation, under which the CNIL approved CNAF access to the DRM (Dispositif de Ressources Mensuelles) monthly-income database for real-time income-surveillance of 12 million beneficiaries, is the framing's working evidence that the November 2023 disclosure had not deterred the agency's continuing operational expansion; LQDN's reading of the expansion as "fuite en avant" (doubling down) translates the framing's substantive accusation into a continuing-pattern register, and the substantive working datum that recovered overpayments amount to only 0.2% of total CAF disbursements is the framing's principal working efficiency-rebuttal to the agency's substantive operational justification.

The framing has also carried into the international civil-society record on welfare-algorithm contestation. The Amnesty International statement on the October 2024 filing places the framing into the Anglosphere human-rights-treaty record, and the campaign sits inside the wider Continental civil-society infrastructure on welfare-algorithm-contestation that the Dutch toeslagenaffaire and the Serbian welfare-algorithm case anchor. The 2025 internal CNAF study acknowledging the algorithm's discriminatory effects and the parallel finding of the Défenseure des droits (the French Ombudsman) supply the institutional-record substantive corroboration of the coalition's substantive working case, and the February 2026 source-code publication under which the CNAF released the algorithm code publicly but omitted essential operational parameters operates as the framing's working register through the pending 2026 Conseil d'État hearing.

Why it has carried

Three features have made the framing durable across the November 2023 to spring 2026 working arc of the campaign.

First, the framing's working register is built on a substantive material disclosure rather than on a rhetorical demand: the November 2023 publication of the algorithm's source code is the working evidentiary spine on which the framing's substantive moral, political, and legal arguments all rest. The disclosure converted the substantive debate from a contested technical question — does the algorithm discriminate? — into a documented administrative-and-political question: how does the agency justify continuing the operation of a documented-discriminatory algorithm? The framing's working register has carried the substantive consequences of that conversion into the Conseil d'État proceedings, the Défenseure des droits' confirmation of the discrimination finding, and the 2025 internal CNAF study's acknowledgement of the discriminatory effects.

Second, the framing operates inside the French civil-society institutional grain rather than as an imported Anglosphere vocabulary. The original fifteen-organisation coalition is composed of French national civil-society organisations — Amnesty International France as the French national section of the international human-rights organisation, La Ligue des droits de l'Homme as the French national human-rights league founded in 1898, Gisti, APF France handicap, the MNCP, and the Syndicat des avocats de France as French civil-society and trade-union bodies — and the working-register phrases ("algorithme de notation", "algorithme de la honte", "score de suspicion", "double peine") operate in the French legal and political vocabulary the coalition's substantive case is built in. The framing's January 2026 expansion to twenty-five organisations through the formal addition of EDRi, AlgorithmWatch, Panoptykon, the European Network Against Racism, and Data for Good extends the French national working-register into a cross-European civil-society register without abandoning the French legal vocabulary the campaign anchors on.

Third, the framing has built itself a sustained working register across a multi-year strategic-litigation arc that the corpus's adjacent single-moment framings have not. Where Fuck the algorithm crystallised on a single Westminster afternoon and where #GobiernoEspía crystallised on a single 19 June 2017 launch day, "Algorithme de notation de la CAF" has been built across the November 2023 source-code disclosure, the March 2024 surveillance-expansion publication, the October 2024 Conseil d'État filing, the January 2026 European-coalition expansion, and the February 2026 source-code publication — each working publication carrying the same canonical campaign register forward through the next operational stage of the file. The substantive working durability is what has converted the framing from a November 2023 disclosure banner into the working civil-society register through which the Continental European contestation of welfare-algorithm deployment is now carried — the corpus's first French-language message anchor and its anchor for the Continental European welfare-algorithm-contestation register that the wider regional civil-society field is built around.