How artificial intelligence is reshaping counter-drone warfare — and why Ukraine's battle-tested interceptors are the most wanted weapons in the Middle East right now.
Iran has been launching waves of Shahed drones across the U.S. Central Command area of operations since the start of Operation Epic Fury — including a March 1 strike on Kuwait that killed six U.S. Army soldiers. The attacks have turned Ukraine's low-cost interceptor drones into the most sought-after weapons on earth.
Ukrainian drone makers have spent four years perfecting one of the most cost-effective kill chains in modern warfare — interceptor drones that down Iranian-made Shahed UAVs for a fraction of the cost of any surface-to-air missile. Now, as those same Shaheds rain down on Gulf states, oil facilities, and U.S. military installations across the Middle East, the phones of Ukrainian defense firms are ringing off the hook.
There is just one problem. Ukrainian law currently prohibits these companies from exporting their interceptors. Wild Hornets, maker of the Sting interceptor, confirmed the ban directly: "Our priority is Ukraine's defense. Exports of drones are not permitted at this time." The company added that the Ukrainian government is engaged in bilateral discussions with partner countries, and that Wild Hornets stands ready to fulfill any export mandate if the law changes.
A second major producer, SkyFall, told Reuters its manufacturing capacity has already outgrown Ukraine's domestic demand and it is ready to sell abroad. President Zelensky confirmed that Ukrainian military experts are already in the Middle East sharing operational experience, and that eleven nations have formally expressed interest in obtaining these systems. A change in the law appears to be a matter of when, not if.
Not all Ukrainian interceptors are created equal when it comes to artificial intelligence, and understanding the difference matters enormously for grasping why these weapons are so disruptive.
The Sting interceptor relies entirely on a human pilot from launch to intercept. No AI guidance of any kind. The pilot tracks and closes on the target through the drone's camera feed, making every navigational and targeting decision in real time. Speed and pilot skill are the primary advantages here.
Several of Ukraine's other interceptor designs combine thermal imaging with radar tracking and AI-assisted guidance for the approach phase — navigating toward the target autonomously — but hand control back to a human operator for the final seconds of the intercept. The human makes the lethal decision. The AI handles the computationally intensive tracking work that gets the drone close enough to matter.
The American Merops system, of which the U.S. has reportedly sent 10,000 units to the Middle East, takes AI further still. It uses artificial intelligence specifically to navigate when satellite and radio communications are jammed by electronic warfare — meaning AI is not merely a convenience here, it is the system's core resilience feature. Without it, the drone goes blind and deaf in exactly the environments where it is needed most.
The AI running inside these interceptors bears almost no resemblance to the large language models behind consumer AI tools. ChatGPT, Claude, and their cousins require enormous computing infrastructure and, critically, a live connection to remote servers. In an electronically contested battlespace — exactly where these drones operate — that connection would be severed the moment jamming begins. A cloud-dependent AI would simply stop working at the worst possible moment.
What these drones carry instead is small, specialized, on-board intelligence: neural networks trained for a single narrow purpose. The specific tasks are object detection to distinguish a Shahed from a bird or civilian aircraft, target tracking to maintain a lock on a fast-moving threat, autonomous pathfinding toward the intercept point, and jamming-resilient navigation that continues working when GPS and radio links are cut.
The hardware involved is likely an edge AI chip comparable in size to what you'd find in a modern smartphone — compact, low-power, and fast enough to process sensor data in milliseconds. The models themselves are almost certainly highly compressed versions of computer vision architectures, optimized to run entirely on-board without any external connection.
The name comes from network architecture. In computing, the cloud refers to centralized servers in massive data centers doing heavy processing far away. The edge refers to the far end of the network — the device itself, right where data is being generated and action needs to be taken. Think of a spider web: the cloud is the center, and every phone, camera, sensor, or drone at the outer reaches is the edge.
An edge AI chip runs its intelligence entirely on the device — no sending data back to a server, no waiting for instructions to return. For a drone operating in jammed, contested airspace, this is not a convenience — it is a survival requirement. A cloud-dependent system goes blind and deaf the moment the radio link is cut. An edge AI system keeps thinking, tracking, and flying regardless. The computation never leaves the vehicle. The drone sees, decides, and acts entirely on its own.
The entire AI system on one of these interceptors probably runs on a processor smaller than your thumb. It knows one job exceptionally well. It does not reason, converse, or generalize. It flies toward the threat and, in some designs, hands the final decision to a human.
There is something worth pausing on in the design of these dual-track systems. A fully autonomous interceptor — one that handles the entire engagement from detection to kill without human input — would be simpler to build, cheaper to operate, and far easier to scale into large swarms. The companies building these weapons almost certainly know that. They chose differently.
The international norm around what defense analysts call "meaningful human control" over lethal force represents a genuine ethical commitment that Western-aligned defense firms have strong incentives to honor. Ukraine, dependent on Western support and seeking legitimacy on the world stage, has every reason to be seen adhering to those norms. The human-in-the-loop design is not a technical limitation. It is a deliberate architectural choice that absorbs real cost — in trained pilots, in communications infrastructure, in operational complexity — in order to preserve human accountability for the lethal decision.
The uncomfortable question on the horizon is whether that commitment can survive the arithmetic of modern drone swarms. When hundreds of Shaheds arrive simultaneously, faster than any human crew can process, the pressure to automate the final decision will intensify. For now, the finger on the trigger is still human. Whether it stays that way is one of the defining questions of this new era of warfare.