A second, distinct thread concerns the regulator's posture toward artificial intelligence oversight: secondary reporting indicates the CNIL is preparing to take on a role as market-surveillance authority for a subset of high-risk artificial intelligence systems under the European Union's AI Act, a designation that is understood to depend on the French implementing law and is reported with probable rather than confirmed confidence given that it has not been independently verified against the regulator's own programme text.
Other Developments
Two injunction closures confirm, rather than escalate, enforcement outcomes. The closure of the FRANCE TRAVAIL injunction and the closure of the injunction against the Google entities both represent the regulator confirming that the controllers concerned had remedied the underlying issues identified in the original orders. In neither case did the regulator proceed to liquidate the penalty payment associated with the injunction, which is the clearest signal available that compliance was found to have been achieved. The Google-related closure sits within a broader context connecting to automated-decision-making proceedings pursued by the Dutch data protection authority against Uber, in cooperation with the French regulator, indicating cross-border coordination on a matter involving the same category of concern.
CNIL positions itself for an artificial intelligence oversight role. Reporting on the regulator's 2026 programme of controls describes preparation for a designation as market-surveillance authority over a subset of high-risk artificial intelligence systems under the EU AI Act. This designation, if and when it takes effect, is contingent on the French law implementing the AI Act domestically, and the claim is carried at probable rather than confirmed confidence because it rests on secondary commentary rather than the regulator's own primary programme document.
Cross-Monitor Connections
The automated-decision-making dimension of the Google-related closure links to parallel enforcement activity pursued by the Dutch data protection authority against Uber, illustrating cross-border regulatory cooperation on automated decision-making that extends beyond France's own jurisdiction; readers tracking artificial-intelligence governance developments more broadly should note the CNIL's anticipated role as an AI Act market-surveillance authority as a structural point of contact between the data protection and artificial-intelligence oversight regimes once the French implementing law is in place.
Outlook
The closure of both injunctions, without penalty liquidation in either case, suggests the regulator's compliance-confirmation process functioned as intended for these two matters; whether further injunctions working through the pipeline will close on a similar timeline is not yet established. The CNIL's anticipated artificial intelligence market-surveillance role remains contingent on the French implementing law and should be treated as a probable rather than settled development until independently confirmed against the regulator's own published programme.
2 earlier updates not shown here.
Standing brief · as of 29 July 2026
Written before the update above. Where they differ, the update is the more recent position.
Lead Signal
The CNIL closed out a wave of high-value sanctions across 2025 and into 2026. The CNIL fined Free Mobile and Free a combined €42 million, treating the two entities as separate controllers each responsible for its own subscriber data despite their corporate affiliation. The CNIL separately sanctioned Free Mobile and Free for incomplete Article 34(2) GDPR breach notifications to affected subscribers. The CNIL also found that Free failed to sort and delete former-subscriber data once it was no longer needed for accounting purposes. France Travail, the public employment agency, is understood to have been fined €5 million, partly because data protection impact assessments had already identified necessary security measures that were never implemented. The CNIL fined the health-data analytics firm IQVIA €5 million, holding that re-identifiable health-data-warehouse data was pseudonymous rather than anonymous. NEXPUBLICA drew a €1.7 million penalty for structural Article 32 security weaknesses in its processing of disability data. A reported 2025 aggregate sanctions figure of €486,839,500 is understood to be concentrated overwhelmingly, on the order of 97-98 percent, in two large multinational decisions issued the same day, separate from the France-specific 2026 sanction wave described above. The EU AI Act's transparency-risk obligations for chatbots and generative content become applicable on 2 August 2026, beginning to layer AI-specific duties onto the GDPR transparency regime the CNIL already enforces.
Other Developments
The CNIL consolidated its cookie and tracker guidelines and recommendation in January 2026, supplemented by 2026 recommendations on multi-device consent. The authority continues to require that consent to trackers be a clear positive act, freely revocable, with refusal exactly as easy as acceptance. The CNIL is understood to identify cookie walls, meaning the conditioning of site access on tracker acceptance, as a distinct compliance-risk pattern under GDPR consent-validity standards. The CNIL is understood to treat GDPR-governed advertising-data processing and Article 82 tracker-deposit rules as separate legal and jurisdictional regimes, avoiding double sanction for the same conduct. The CNIL processed 539 health-data-warehouse authorisation applications in 2025. The CNIL operates a sector-specific authorisation gate for health-data warehouses, such as IQVIA's LRX and EMR warehouses, rather than a blanket data-localisation mandate. The technical référentiel that Arcom must adopt for pornography-site age verification under the SREN law remains at consultation stage; the CNIL opined on the draft on 26 September 2024, favouring privacy-preserving proof-of-majority methods over facial recognition. The G7 group of data protection authorities, meeting in Paris in June 2026, is understood to have adopted a declaration on privacy-preserving age verification and a set of children's-protection principles. The EDPB is understood to have adopted a common breach-notification template together with generative-AI anonymisation, web-scraping and blockchain guidance. The CNIL has sanctioned the scraping-based creation of a facial-recognition database as unlawful Article 9 GDPR biometric processing, and treats biometric-identification cameras as excluded from its general smart-camera guidance. The CNIL and its EU peer authorities are reported to have called for prohibiting AI systems that categorise individuals by inferred protected characteristics such as ethnicity, sex, and political or sexual orientation, though this appears to be an advocacy position rather than an adopted rule.
Cross-Monitor Connections
The CNIL's treatment of banks and insurers as paradigm cases triggering mandatory DPO designation through large-scale client-monitoring activity intersects with financial-crime and data-sharing obligations tracked by the financial-integrity monitor. The Free Mobile and Free telecoms sanctions may also bear on payments-specific subscriber data flows tracked by the world-payments monitor. The EU AI Act's 2 August 2026 transparency-obligation date and its interface with GDPR data protection impact assessments are primarily AI-Act-first developments; readers seeking the regulation-specific analysis should consult the artificial-intelligence monitor, while this brief retains the data-protection angle.
Outlook
The trajectory in France points toward continued high-materiality enforcement of accountability, security and breach-notification duties, alongside a widening interface between GDPR and the incoming EU AI Act. The health-data pseudonymisation question tested in the IQVIA decision, and the unresolved status of France's collective-redress mechanism for data protection claims, both remain open threads for future cycles.