France Drops Palantir for ChapsVision: What Europe's Sovereignty Reset Means for Your Stack

On 16 June 2026, French Prime Minister Sébastien Lecornu said the DGSI, France's domestic intelligence service, will move its mass-data analysis off Palantir's platform and onto software from ChapsVision, a French firm founded in 2019 that now employs more than 1,000 people. Lecornu framed it plainly, saying France "cannot rely on tools developed by foreign powers" and must avoid new strategic dependencies in the digital sphere.

France is not alone. Germany's domestic security service, the BfV, had already selected ChapsVision over Palantir the previous month. And Paris is backing the shift with money: a €655 million package for sovereign AI, with public-sector staff moving to an assistant powered by the French model-maker Mistral. Two French suppliers, one political message: sensitive analysis on ChapsVision, general AI on Mistral.

The timing is the part worth sitting with. The DGSI has used Palantir since 2016, in the aftermath of the November 2015 Paris attacks, and renewed that contract as recently as late 2025. So what changed?

Days before the announcement, US Department of Commerce export controls led Anthropic to temporarily suspend access to its newest AI models for affected users, before the controls were lifted and access was restored on 1 July 2026. It was a short, compliance-driven suspension rather than a permanent ban. But the lesson landed hard in European capitals: if a foreign government's export policy can interrupt access to a critical tool, then that tool is a dependency, not just a purchase. Lecornu made exactly that argument, warning against relying on the goodwill of partners who can, in effect, turn off the tap. (Anthropic's own account of the suspension and its restoration is here.)

This is the important distinction, and it is not about any single company. The suspension was temporary and quickly reversed. The structural point European governments drew from it is permanent: control of the off-switch sits with the vendor's home jurisdiction, not with you.

For years, "digital sovereignty" was mostly a debate about where data physically sits. This episode reframes it around a sharper question: who can restrict your access, and under whose laws?

Data residency, the location of your data, is not the same as jurisdiction, the legal control over it. A US-headquartered vendor remains subject to US law, including export controls and compelled-disclosure regimes, no matter which region hosts the servers. France's move treats that jurisdictional exposure as an operational risk to be engineered out, not a compliance footnote to be managed.

It is worth being fair about the trade-off, because the French government's critics are too. Swapping a mature platform like Palantir for a younger domestic one is not a like-for-like technical swap, and the real test is whether ChapsVision matches Palantir's capability under load. Sovereignty, in other words, has a cost and a break-in period. France has decided that cost is worth paying. Whether the tooling keeps up is the open question that the next year will answer.

France and Germany are not outliers. In the UK, the NHS's £330 million Palantir data contract is under political review, London's mayor blocked a proposed Metropolitan Police deal, and Germany's military has said it will stop using Palantir products. Across the G7, the temporary AI-model access cut has visibly accelerated a "sovereign alternative" conversation that was already simmering. The direction of travel in European public-sector procurement is now unmistakable: reduce single points of foreign control.

You are not a national intelligence agency, but the underlying exposure is identical in shape. Ask three questions of your own stack:

• Which of our critical tools are controlled from a single foreign jurisdiction? Email, chat, storage, analytics, AI, and your identity provider are the usual suspects.
• What happens to us if access is restricted, even temporarily? If the answer is "we stop operating," that is a dependency worth pricing.
• Is there a credible European or self-hostable alternative, and what would switching actually cost?

The point is not to rip everything out. It is to know, deliberately, where your sovereignty risk sits, so the choice to accept it or reduce it is a decision rather than an accident.

The tools in this story, and the ones your business runs on every day, carry very different levels of sovereignty risk. You can score any tool free with SovereigntyScore, or get a full Full Stack Audit for €99: a complete risk breakdown, scored European alternatives, and a step-by-step migration plan for each tool.

Score a tool free  |  Get your Full Stack Audit for €99

Related reading: The Fable 5 Shutdown: What a 19-Day AI Blackout Taught Europe About Sovereignty · The EU's €30bn AI Gigafactory Plan · What is European Digital Sovereignty? · Sovereignty Risk in SaaS Procurement · OpenAI vs Mistral: Sovereignty Comparison · AWS vs OVHcloud and Scaleway.

Sources: The Guardian, Computing, and Anthropic. This article is independent analysis and does not constitute legal advice.