Europe quietly turns its satellites into ai spies with madrid at the helm
While Silicon Valley chases self-driving taxis, a nine-company European squad led by Spanish aerospace firm GMV has just lit the fuse on Myriad, a €5 million, 48-month project that will teach EU defence ministries to see through clouds, darkness and deliberate camouflage in near real time.
The kick-off meeting, held behind closed doors last week, locked in partners from Italy, Germany, France and the Netherlands. Their brief: build machine-learning pipelines that can swallow optical and radar feeds from multiple European satellites, spit out reliable threat cues, and explain to a colonel why the algorithm flagged a convoy at 03:14 a.m. near the Arctic Circle.
Why this is not just another ai grant
Brussels has banked the entire cheque from the European Defence Fund, the Commission’s newest piggy bank for dual-use tech. That matters. Unlike civilian space programmes, Myriad must deliver deployable code to the EU Satellite Centre (SatCen) in Torrejón de Ardoz before the decade is out. No slide decks, no pilot fever—just working code on classified servers.
GMV’s project director, Begoña Rojo, puts it bluntly: “Today women represent roughly 21 % of the defence industry. Myriad’s analytics lab will be 50 %.” The statement sounds like corporate boilerplate until you realise the engineering lead on synthetic-aperture-radar fusion is already a 29-year-old PhD from Seville who joined GMV last year after refusing a cushier offer in Menlo Park.
Four winters of live-fire feedback
Myriad will iterate in four annual cycles. Each winter, defence intel officers from at least six EU member states will fly to Madrid, dump real crisis imagery on the consortium and watch the algorithms squirm. If the false-positive rate on vehicle detection stays above 5 %, the pipeline is yanked and re-trained overnight. The goal is to cut analyst workload per scene from 45 minutes to under four, a figure that would let a single desk officer monitor the entire Baltic littoral instead of one border crossing.
The trick is radiometric calibration at chip level. Optical satellites love sunshine; SAR birds laugh at clouds but hate ambiguity in side-lobe returns. GMV’s engineers will fuse the two streams through a shared vector space that updates weights on the fly, something academic papers promise but few defence contractors ever ship. If they nail it, the same backbone can be re-targeted for disaster relief, illegal fishing or pipeline leaks—Brussels’ favourite hedge for justifying military R&D.
Europe bets on its own eyes
Dependence on American or Chinese analytics engines has become an open sore inside EU councils. Myriad is the antibiotic: every line of Python, every labelled dataset, stays on European silicon. The consortium even plans to publish a subset of the training corpus—minus coordinates and national signatures—through the Copernicus data portal, a move that could seed an entire generation of start-ups while keeping the sensitive layers in a classified enclave.
Expect first demos by late 2025. By then the EU hopes to have its new constellation of governmental imaging birds in orbit, thirsty for exactly this kind of AI. If Myriad delivers, the phrase “made in Europe” will finally apply not just to the satellites overhead but to the invisible logic that decides what soldiers, spies and disaster managers see before anyone else does.