Shield AI and Destinus completed the third and most complete phase of a planned integration on a Spanish test range this week, running an autonomous counter-drone sequence on the Destinus Hornet interceptor: ground-control mission planning, radio testing, autonomous terrain following, in-flight retargeting, and an autonomous terminal maneuver on operator command, according to a joint press release distributed via PRNewswire.
The Segovia, Spain, campaign closes a three-phase plan the companies announced earlier this year and locks in the next step: the same autonomy software moves onto the Destinus Ruta in Ukraine, where Shield AI's V-BAT surveillance drones and multiple Ruta strike drones will run coordinated behaviors against Russian drone threats. That is the only part of the story with a fully independent, observable test bed. The rest is what the two companies say they did, not what outside observers have seen.
The cast, in plain terms. Hivemind is Shield AI's autonomy software, the layer that lets a drone sense, decide, and act without a human pilot in the loop. The Hornet is Destinus's jet-powered counter-drone interceptor, not a fighter jet, designed to chase and kill incoming hostile drones, including drone swarms and loitering munitions, which are drones that circle a target area before striking. V-BAT is Shield AI's vertical-takeoff surveillance drone. The Ruta is Destinus's low-cost turbojet strike drone, built to penetrate in GPS-denied airspace, meaning environments where the satellite navigation signal is jammed or spoofed.
Per the release, the Segovia test "validated Hivemind AI piloting for autonomy-enabled coordination and in-flight adaptation in contested airspace" and "executed a complete operational concept developed for the Destinus Ruta" on the Hornet as a surrogate platform. Phase 1, the companies say, established Hivemind control on the Hornet in under two months. Phase 2 demonstrated V-BAT and Hornet multi-platform teaming in flight. Phase 3 is the Segovia run.
What the release does not disclose is also part of the story. The piece names no surrogate targets: no specific model of incoming drone the Hornet engaged, no countermeasures deployed, no intercept outcome. It does not describe weapons effects, rules of engagement, the sensor-to-shooter timeline, or the latency between detection and terminal maneuver. It does not say which operators were in the loop, what "operator command" means at the terminal step, whether human authorization is required before the Hornet commits to a kill, or how that authorization is conveyed when communications are degraded.
It also does not name a single independent observer. No Ukrainian military readout, no OSINT tracker with imagery of the Segovia range, no humanitarian monitor with line-of-sight, no third-party technical evaluator appears in the release. The verification regime, in other words, is the press release itself.
That matters because the next test is not a Spanish range. The companies frame Ukraine as the live proving ground, where V-BAT and Ruta systems will run coordinated strike behaviors against an active drone threat. A warzone makes independent observation harder, not easier: signals are jammed, airspace is dynamic, and the systems will operate against adversaries who are also adapting. It also makes independent observation more important, because the consequences of an autonomous interceptor misidentifying a target are measured in civilian lives, not in test telemetry.
The hard questions follow the deployment. What targeting authorities will an autonomous Hornet or Ruta carry in Ukrainian airspace, and who grants them. What does "operator command" mean in practice when the operator is remote and the decision window is seconds. What civilian-harm mitigation is built into the autonomy stack, and what does it look like in code. What sensors feed the terminal decision, and what happens when those sensors are spoofed, as Russian electronic warfare has done repeatedly in this conflict. What independent observers, including Ukrainian military, OSINT analysts, and humanitarian monitors, will be positioned to verify in real time.
The Segovia milestone is real. Three planned integration phases, named platforms, a named location, and a confirmed Phase 3 designation are not in dispute. The structural claim is also credible: cheap, autonomy-enabled interceptors are a real answer to the saturation problem that cheap drones have created on modern battlefields, and the companies are putting capital and engineering behind that thesis.
The unresolved claim is whether the autonomy works as billed under conditions the vendors do not control. The press release closes that question by asserting it. Ukraine will not.