La Quadrature du Net coalition challenge to the CAF / CNAF risk-scoring algorithm before the Conseil d'État (2024–ongoing)

On 16 October 2024 La Quadrature du Net, the Paris-anchored French digital-rights association, filed a petition before the Conseil d'État — France's highest administrative court — at the head of a fifteen-organisation civil-society coalition seeking the withdrawal of the algorithmic risk-scoring system operated by the Caisse nationale des Allocations Familiales (CNAF), the national administrative body that coordinates the Caisses d'Allocations Familiales (CAF) family-benefits network. The system calculates more than 13 million monthly suspicion scores across approximately 32 million people living in CAF-benefit households and uses those scores to target administrative fraud-control checks. The case is, on LQDN's own framing and as independently corroborated by civil-society coverage, the first time French civil society has challenged a public-administration algorithmic social-scoring system before the Conseil d'État, and is the most substantial continental-European parallel to the corpus's Anglosphere welfare-algorithm strategic-litigation register — distinct in venue (the highest French administrative court rather than the English High Court), in claimant register (a fifteen-organisation civil-society coalition rather than a single membership organisation or representative claimant), and in legal grounding (GDPR data-protection arguments alongside non-discrimination claims).

Precursor: the 2020–2023 source-code-disclosure campaign

The Conseil d'État petition is the apex moment of a multi-year LQDN campaign whose foundational work was the public extraction and analysis of the CAF risk-scoring algorithm's source code. Working with the Stop Contrôles and Changer de Cap collectives, LQDN spent more than a year pursuing the CAF's algorithm through the Commission d'Accès aux Documents Administratifs (CADA) — the French administrative-documents access authority — before securing public disclosure of two historical versions of the system, one in operation between 2010 and 2014 and a second between 2014 and 2018. The CAF refused to disclose the version in current operation, but LQDN's published analysis of the disclosed versions documented approximately forty parameters that weight up a claimant's "suspicion score," including low income, unemployment status, receipt of the Revenu de Solidarité Active (RSA) minimum-income benefit, receipt of the Allocation aux Adultes Handicapés (AAH) disability allowance while employed, residence in disadvantaged neighbourhoods, high rent-to-income ratios, and unstable employment. LQDN's 27 November 2023 analysis — under the title "Scoring of welfare beneficiaries: the indecency of CAF's algorithm now undeniable" — provided the documentary basis on which the October 2024 petition was built. The disclosure-and-analysis register is structurally analogous to Foxglove and GMCDP's UK General Matching Service disclosure pressure, with the substantive difference that LQDN had the source-code itself to analyse rather than the operational explanation Foxglove extracted in the DWP file.

Coalition and tactics

The October 2024 petition was filed jointly by fifteen French civil-society organisations: La Quadrature du Net, AADJAM, Aequitaz, Amnesty International France, the Association Nationale des Assistants de Service Social (ANAS), APF France handicap, the Collectif Changer de Cap, the Fondation Abbé Pierre, the Groupe d'Information et de Soutien des Immigré·e·s (GISTI), Le Mouton numérique, the Ligue des droits de l'Homme, the Mouvement National des Chômeurs et Précaires (MNCP), the Mouvement Français pour un Revenu de base (MRFB), the Conseil National des Droits Humains Romeurope (CNDH Romeurope), and the Syndicat des avocats de France (SAF). The coalition's composition is its substantive feature: it combines digital-rights expertise with anti-poverty advocacy (Fondation Abbé Pierre, MNCP, ATD Quart Monde-adjacent work), disability rights (APF France handicap), legal-profession backing (Syndicat des avocats de France, Ligue des droits de l'Homme), and the affected-constituency register of social workers (ANAS) and migrant-rights groups (GISTI, CNDH Romeurope). Amnesty International France's parallel press release frames the algorithm as a system that "assimilates precarity with suspicion of fraud" and participates in "stigmatisation and institutional mistreatment of the most disadvantaged."

The tactical centre of the campaign is the Conseil d'État petition itself, structured around two principal legal arguments. The first is a data-protection argument under the General Data Protection Regulation (GDPR), challenging the lawfulness of CNAF's processing of personal data on the scale of 32 million people and the use of an algorithmic system to assign suspicion scores without an adequate legal basis. The second is a non-discrimination argument, anchored on LQDN's source-code-disclosure findings that the system weights up risk scores for low-income, single-parent, disabled, and migrant-status households — categories whose differential treatment by a public administration's automated-decision-making system is incompatible with the principle of non-discrimination in French and European law. As Bastien Le Querrec of La Quadrature du Net put it, "This algorithm is the manifestation of a policy of persecution of the poorest people."

The European coalition expansion of January 2026

On 20 January 2026 LQDN announced that ten additional civil-society organisations had joined the case at the close of the written-submission phase. The additions broadened the coalition along two structural axes. On the French labour-and-civil-society axis: the Confédération Générale du Travail (CGT), Union Syndicale Solidaires, and the Fédération Syndicale Unitaire Travail Emploi Insertion Organismes Sociaux (FSU TEIOS) — taking the case past digital-rights and anti-poverty groups into the French trade-union register on welfare-administration automation. On the European digital-rights axis: European Digital Rights (EDRi), AlgorithmWatch, and the Panoptykon Foundation — making the case the first national-court algorithmic-public-administration file in continental Europe to draw direct interventions from the three EDRi members whose national-anchor positions on algorithmic accountability are the corpus's principal continental-European reference points. Additional French civil-society additions included Data for Good (the French technologist-volunteer network), the European Network Against Racism, the Mouvement des mères isolées (a single-mother grassroots advocacy organisation), and Féministes contre le cyberharcèlement (the French feminist anti-cyberharassment collective). The expanded coalition counts twenty-five co-petitioners.

The 20 January 2026 update also confirmed a substantive disclosure outcome: the CNAF had on 15 January 2026 disclosed the source code of its current risk-scoring algorithm — the version LQDN had been unable to obtain through CADA during the 2020-2023 disclosure-pressure phase. The disclosure was a direct response to the litigation pressure and represents the campaign's clearest transparency outcome in advance of any merits ruling. As of the close of the written-submission phase in late January 2026, the case awaits its public hearing in spring 2026.

Significance for the broader AI-good movement

The CNAF challenge is the corpus's first continental-European public-administration algorithmic-litigation campaign, the first French-court strategic-litigation campaign of any kind, and the first to anchor on a coalition of fifteen-then-twenty-five civil-society organisations rather than on a single claimant or a digital-rights or sectoral-charity claimant. Its substantive distinctness from the corpus's existing Anglosphere welfare-algorithm campaigns — Foxglove and GMCDP's DWP General Matching Service file, Foxglove and JCWI's visa-streaming case, and Foxglove and Ofqual on the A-level standardisation algorithm — sits on three axes. First, on legal grounding: the CNAF case mounts GDPR data-protection arguments alongside non-discrimination claims, opening a public-law register that the UK cases (operating under judicial-review and Equality Act 2010 instruments) cannot. Second, on disclosure register: LQDN's CADA-extracted source code provides the underlying technical specification of the system being challenged, where the UK cases have operated on operational descriptions of opaque systems. Third, on coalition register: the fifteen-to-twenty-five-organisation civil-society coalition stretches the anti-AI-discrimination claim across the French anti-poverty, disability-rights, migrant-rights, trade-union, and feminist civil-society fields, broadening the constituency whose engagement with algorithmic-public-administration arguments the case constructs. The campaign also sits in a load-bearing relationship to the EU AI Act fundamental-rights civil-society coordination: the CNAF file supplies the European policy debate with a substantive national-level test case of public-administration algorithmic social scoring of the kind the AI Act categorises as high-risk, and the January 2026 entry of EDRi, AlgorithmWatch, and the Panoptykon Foundation into the case formalises the connection between the corpus's continental-European digital-rights coalition and the French national-court venue.