Opinion
Europe's own battlefield AI exists. Our ministries should buy it now.
Helsing's Clarity does what NATO once bought from America, with the decision kept in human hands. Waiting past the Bundeswehr's own 2027 deadline is not prudence, so fund the finalists and let performance pick the winner.
Europe built its own Maven, and this time you should know its name. On 11 November, Helsing unveiled Clarity, an agentic intelligence platform for military intelligence, built in Europe, run on infrastructure the customer owns, with a human keeping the decision over every attack. It reads the flood of data a unit cannot: more than 10 terabytes of satellite data a day for a national military, over a billion radar pulses from one reconnaissance flight, fused in seconds into targets a commander can judge. NATO bought American, for lack of a European alternative. Now one exists. Our ministries should buy it, and buy it fast.
Those who would wait say: battle-proven beats home-built, and Palantir's spokesman says their software has proven itself on the battlefields of Ukraine and as NATO's operating system, and is fully sovereign because no one, not even Palantir itself, can reach a customer's data without that customer's involvement. That argument held only while we had nothing of our own. The Bundeswehr is already testing some 30 solutions from European companies, Helsing's among them. And how control is designed matters more than years on the battlefield: military representatives warn of an attitude they call 'The computer said to do this'. Clarity documents every attack decision with the name of the person responsible. Thomas Daum, head of the Bundeswehr's cyber troops, told Handelsblatt that Maven is not an option for Germany at the moment, because it is unthinkable that employees of the company would have access to military data. The Bundeswehr set its own prototype deadline of mid-2027. Waiting past a date our own procurement wrote is not prudence.
What we build in these five years decides the next twenty.
One of the thirty can be the European winner. The next years of European defence AI will be decided by national orders, not lectures. Spain decided in October 2026 not to renew Palantir's contract for its Armed Forces Intelligence Center and to develop the technology with a Spanish company. The UK announced a similar split at the start of October 2026. France's domestic intelligence service is already replacing Palantir with ChapsVision, and the Netherlands' defence ministry wants out within two years. For you, an engineer or a soldier starting now, that chain is the job market: a European winner with national orders behind it means the targeting software of the next decade is built here, and you are the one building it. No home market, and what is under test stays a test.
Berlin should run an open competitive procurement among the solutions already under test, and fund at least two finalists to the mid-2027 prototype. Then write control into the requirements: a named person authorises every transfer to a weapon system, and every decision is documented. Even the American military warns its officers against doing what the machine decides for them. A system that keeps the decision with a human matters more than the flag on the login page. The pull of the battle-proven incumbent will resist a fast European buy, so keep the contest public, date it, and let performance, not patriotism, pick the winner. If that sounds demanding, it is. You will work under these systems for years. Demand now that they stay accountable. Europe can close the gap today. Now the ministries must sign for it.