France’s domestic spy agency is trying to move off Palantir—but its contract runs to 2028 and full switch may wait until 2029

France’s domestic intelligence service says it is beginning a gradual break from Palantir, the U.S. data-analysis software company whose tools have been used for nearly a decade in certain intelligence activities.

The shift is being framed publicly around “digital sovereignty”—a push to avoid new strategic dependencies in sensitive technology, according to multiple consistent sources cited in the French press. But the transition is proving far more complicated than swapping vendors: reporting indicates Palantir’s contract has been renewed through 2028, while a complete operational replacement is being discussed on a 2029 timeline.

That stretched-out schedule reflects how deeply Palantir’s software has been woven into workflows, data formats, analyst habits, and processing chains. Even as the agency seeks to reduce reliance on non-European technology, it also has to keep investigations running at full speed and manage a migration where any disruption could carry immediate operational costs.

France’s “digital sovereignty” push collides with the reality of entrenched tools

The agency at the center of the story is the DGSI—France’s Direction générale de la sécurité intérieure, roughly comparable to a domestic security and counterintelligence service. Its core argument for moving away from Palantir is straightforward: stop accepting new strategic dependencies in the digital realm.

Multiple publications describe a government-level political line aimed at reducing exposure to non-European companies for functions considered sensitive. In that framing, Palantir represents a structural risk: the platform may be effective, but it is built and maintained by an American company operating under U.S. legal and compliance obligations outside French control.

French reporting also emphasizes that dependency isn’t just about a contract or license. It grows through daily use—proprietary data formats, connectors, automations, dashboards, and enrichment methods. Over time, organizations can become locked into a technology stack, making an exit expensive, lengthy, and risky.

In intelligence work, that lock-in is even more consequential because it touches classified data, compartmented environments, and procedures that may be only partially documented by design.

The geopolitical backdrop adds weight to the argument. European administrations have spent years trying to reduce non-European dependencies in cloud services, cybersecurity, and data analytics. Some sources point to Germany as a parallel case, with Berlin also grappling with reliance on an American solution for internal security uses.

The DGSI’s posture, as described in the article, is not that U.S. technology is inherently less reliable. The emphasis is on control: the ability to audit, negotiate strict terms, manage updates, and ensure continuity even in the event of a commercial or diplomatic rupture. Still, the existing system remains in place in the near term, with a contract reported to run through 2028.

The public debate has also been colored by Palantir’s political associations in the United States, which some observers treat as symbolic. But the DGSI’s stated objective remains pragmatic: maintain a robust analytics chain while limiting blind spots tied to vendor dependence and retaining control over critical security parameters.

https://www.europe-infos.fr/actualites/9960/2026-5-e-19-et-20-septembre-train-des-mouettes-a-9h30-18h-et-depot-ouvert-ce-qui-surprend-aux-journees-du-patrimoine/

https://www.europe-infos.fr/actualites/9951/wall-street-cloture-en-baisse-nasdaq-sous-pression-depenses-massives-dans-lia-remises-en-cause-ce-que-les-profits-doivent-prouver/

Why the timeline stretches: contract through 2028, full operational handoff discussed for 2029

The available reporting points to a two-step calendar. On one hand, sources say Palantir has communicated that its DGSI contract was renewed through 2028 after earlier renewals. On the other, multiple articles describe a handoff to the French company ChapsVision around 2029.

The apparent contradiction comes down to the difference between the end of a contract, the end of day-to-day use, and the end of technical dependence.

In critical systems, migrations typically happen in stages. Teams often run two pipelines in parallel—sometimes called a “double run”—to validate result quality, connector robustness, and performance under load. For a service like the DGSI, the priority is uninterrupted analytical capability, with high requirements for availability and traceability. The switch is not just installing new software; it means rebuilding an entire ecosystem.

Data is the central challenge. A transition can involve moving historical records, indexes, correlation models, and custom-built query logic. Data-analysis systems rely on schemas, metadata, and cleaning rules; changing them can alter outputs, create false positives, or hide weak signals. A poorly managed migration can temporarily reduce capability, which helps explain the cautious timeline.

Another sensitive point is internal expertise. Analysts and engineers trained on one tool develop muscle memory and methods. Moving to a new platform requires training, rewriting procedures, and accepting an initial productivity dip. In an intelligence service, that dip has to be offset—through reinforcements, transitional tooling, or carefully phased migration scopes.

A 2029 horizon suggests a strategy of progressively shrinking Palantir’s footprint rather than a hard cutoff. It also leaves room for audits, security testing, and legal validations. The article’s bottom line: “exit” is measured not by announcements, but by the real ability to operate daily missions without the original vendor.

https://www.europe-infos.fr/actualites/9935/souverainete-des-donnees-au-maroc-voici-pourquoi-les-entreprises-accelerent-leur-migration-vers-le-cloud-hybride/

ChapsVision emerges as the French alternative for data analysis and “fusion”

ChapsVision is repeatedly cited as the expected successor for analysis and data-fusion functions. For the French state, choosing a domestic company is presented as offering tighter control over the vendor relationship, the ability to contract for specific requirements, and reduced exposure to a non-European legal framework.

In intelligence work, the location of teams, the ability to support constrained environments, and the handling of security clearances carry significant weight.

The replacement would not necessarily replicate the existing setup feature-for-feature. Products can differ in user experience, performance on certain graph or search workloads, and the maturity of connectors. The DGSI could prioritize reproducing key functions first, then expand—forcing hard choices about which data flows move first, which missions require strict continuity, and where a transition period is acceptable.

For ChapsVision, the challenge is demonstrating it can handle high volumes, strict security constraints, and demanding use cases. That means assurances on update cycles, long-term maintenance, and resilience. Public-sector buyers often seek precise commitments: patch timelines, compatibility with isolated environments, audit mechanisms, and the ability to operate without uncontrolled external dependencies.

The transition also tests France’s industrial strategy. The state wants domestic players capable of meeting “sovereign” needs, but those firms must remain competitive against global giants. The intelligence market is narrow, highly demanding, and opaque. A successful migration could become a showcase for other agencies; a failure could weaken the credibility of the sovereignty strategy.

Sources cited in the article stress how risky the migration is—not as an automatic vote of no confidence in a French supplier, but because of how integrated the prior tool has become and how critical the use cases are. In this kind of project, evaluation ultimately comes down to measured performance, side-by-side results, security audits, and operational feedback, beyond public messaging.

The operational risks of migrating classified, deeply integrated intelligence systems

The biggest risk is a temporary loss of capability. In an intelligence service, reduced performance can mean longer delays to exploit a signal, cross-reference information, or prioritize an alert. Tools like Palantir are valued for linking heterogeneous data, visualizing networks, and speeding search. Replacing those functions requires comparable results on real use cases—not just vendor demos.

A second risk involves data integrity and continuity of evidentiary chains. Analytics systems generate traces, query histories, access logs, and other elements needed for internal accountability. During a migration, the agency must preserve the ability to reconstruct what was done, by whom, and on what basis—raising compliance and internal-control issues. In classified environments, every transfer step must be secured, documented, and audited.

Cybersecurity is a third pillar. Introducing a new tool—even a domestic one—adds new components, interfaces, and sometimes new software dependencies. Teams must conduct code audits where possible, penetration tests, and validation of update mechanisms to avoid new vulnerabilities or unintended entry points. Vendor access is also sensitive: support, patches, and any remote intervention must be tightly governed.

The human factor matters as much as the technology. Analysts have to regain speed and collaboration patterns. A different interface or query logic can slow experienced teams. The DGSI therefore has to manage training, documentation, and support while keeping operations running—where success often hinges on change management and strong internal champions.

Ultimately, the migration reopens a familiar tradeoff: autonomy versus performance. Digital sovereignty is about control, but it has to coexist with operational effectiveness. The long timeline and the precautions described point to a gradual approach—one that prioritizes not weakening missions the state considers vital.

https://www.europe-infos.fr/actualites/9893/en-2026-albert-einstein-parcimonie-conceptuelle-preuves-claires-pourquoi-lia-generative-laurait-surpris-et-rebute/