Alex Karp Says France’s Digital Sovereignty Push Risks Hurting AI Competitiveness

France’s decision to move its domestic intelligence service away from Palantir has put a sharp question to Europe’s AI strategy: is digital sovereignty a route to resilience, or a costly form of protectionism?

At the G20 Innovation Ministerial in Chapel Hill, Palantir CEO Alex Karp argued that French restrictions on U.S.-made software and Europe’s wider regulatory direction are undermining competitiveness. France’s government takes the opposite view: for intelligence and other critical public systems, reducing reliance on foreign technology is a strategic security requirement. 18

Karp’s case: sovereignty should not mean weaker technology

Karp’s central claim is that regulation and procurement restrictions can backfire when they keep governments and companies from using capable, established technology. Speaking to AFP, he said French restrictions—including limits on government use of U.S.-made software—were proving counterproductive and that the regulatory environment was not helping either France or Europe. 18

His criticism is especially pointed at the idea that a service becomes genuinely sovereign simply because a French company sits between the customer and the underlying technology. In Karp’s framing, dependence has not disappeared if the systems that matter still rely on AI models or infrastructure developed elsewhere. The practical risk, he argues, is that governments bear the cost and disruption of migration without gaining a comparably independent technological base. 18

This is an argument about capability and speed. Karp’s view is that Europe should prioritize deployment, scale and access to leading tools during an intensifying global AI contest—not restrict suppliers first and hope domestic alternatives catch up later.

Why the DGSI switch became the test case

France’s domestic intelligence agency, the DGSI, selected French company ChapsVision to replace Palantir’s data-analysis tools. The French prime minister’s office described the decision as replacing the U.S. company with a French rival, but Reuters reported that the process is likely to take several years because Palantir’s long-term contract had been renewed at the end of 2025.

The choice is consequential precisely because intelligence systems are difficult to replace. They are embedded in sensitive operations, data flows and staff workflows. That makes the transition a practical test of whether a domestic supplier can meet demanding operational needs while providing France more control over critical technology.

France has presented the policy as part of a broader sovereign-technology push. In announcing the DGSI decision, Prime Minister Sébastien Lecornu also referred to a move of health-data hosting from Microsoft to Scaleway.

France’s answer: dependency itself is the risk

France’s position is not simply that domestic suppliers deserve preference. Its stated concern is strategic dependency: whether a state can retain meaningful control over the technology used for sensitive public functions.

For an intelligence agency, the relevant questions extend beyond product performance. They include who controls the supplier relationship, where sensitive data is handled, how systems are maintained, and whether a foreign provider or its home jurisdiction could create an unacceptable point of dependency. France24 reported that Lecornu described the goal as avoiding new strategic dependencies in the digital sphere.

That logic accepts a difficult trade-off. Building or migrating to domestic capabilities may take longer and may initially impose operational costs. But France’s bet is that those costs can be justified if they reduce exposure in critical systems over the long run.

The EU AI Act does not settle the DGSI question

The disagreement should not be reduced to a simple clash between France’s sovereignty policy and the EU AI Act. The Act takes a risk-based approach for AI within its scope, but it explicitly excludes systems used exclusively for military, defence or national-security purposes. It also does not affect member states’ competences in national security. 13

That means the DGSI procurement decision can be treated primarily as a national-security and strategic-procurement matter rather than as an ordinary application of the EU’s civilian AI rules. The exemption is purpose-based: an AI system is outside the Act only insofar as it is used exclusively for military, defence or national-security purposes. 3

The larger political tension remains, however. Karp sees Europe’s regulatory posture as a drag on adoption. France sees security-sensitive technology as an area where market access and short-term convenience cannot be the only criteria.

The G20 contrast: Washington’s lighter-touch approach

The French debate unfolded as the United States urged G20 countries to avoid new AI-specific oversight bodies and broad new rules that could slow innovation. Reuters reported that U.S. technology adviser Michael Kratsios promoted the proposed “Carolina Principles,” under which countries would reserve new regulation for genuinely novel issues rather than treat each emerging technology as requiring a new rulebook.

That approach aligns with Karp’s emphasis on adoption and competitive speed. It also reflects the United States’ interest in an international environment that does not create major new barriers for the largest AI companies, most of which are American.

The real trade-off

Neither side has demonstrated a definitive answer yet. Karp is making a warning: a sovereignty policy that merely replaces a high-performing foreign supplier with a less capable intermediary may reduce European competitiveness without delivering real independence. France is making a different warning: reliance on foreign technology in intelligence, health and state systems can become a strategic vulnerability even when the product is strong.

The success of France’s approach will depend less on the nationality of a supplier than on whether domestic alternatives can deliver reliable capability, sustained investment and meaningful control over sensitive systems. The DGSI transition will be an important test of that proposition.