France's DGSI Drops Palantir for ChapsVision After a US AI-Access Restriction

The DGSI's move off Palantir is the headline, but the reasoning is the point. Lecornu tied it directly to a US order requiring Anthropic to limit access to its most advanced models for all foreign nationals on national-security grounds — a restriction whose practical effect reached well beyond its intended target. “We cannot depend on the goodwill of certain partners who are capable, as we have seen in recent days, of cutting off access,” he said. The package adds €655 million for AI through 2030 and a Mistral-powered assistant for state employees.

Palantir pushed back, saying its DGSI contract remains in force; Germany's domestic service had reportedly made the same Palantir-to-domestic switch weeks earlier. The direction of travel is unmistakable either way: European states moving critical intelligence and AI workloads off American platforms because access to those platforms has become a lever Washington has shown it will pull.

The proprietary read. This is the sovereignty argument winning on evidence rather than principle. France did not drop Palantir over an abstraction; it dropped it because a US AI-access restriction demonstrated, in real time, that dependence on American software is dependence on American policy. The order was aimed at adversaries and caught allies — exactly the case European advocates of sovereign AI have been making, now proven by Washington's own hand. The €655 million and the Mistral assistant are the bet that a controllable European stack beats a more capable American one you can be cut off from. Tracked in Signal No. 83.