According to Forbes, after a failed debut in 2024, Ukraine is now mass-producing effective AI-enabled attack drones, causing a dramatic spike in Russian casualties. In December 2025, OSINT data showed Russian deaths hit 184 per day, a 37% increase over the average, attributed largely to these new systems. Key players include The Fourth Law (TFL), which launched its TFL-1 autonomy module in July 2025, and a partnership with major drone maker Vryiy to produce the Vyriy-10-TFL-1 for just $448. US-Swiss firm Auterion and Eric Schmidt’s Swift Beat are also supplying AI hardware. Field reports now claim hit rates of 80% to over 90% for AI drones, a massive leap from the 10-20% typical for novice human pilots. The TFL-1 module itself costs about $150 and allows target lock from 500 meters, ignoring electronic jamming.
From Winter to Spring
Here’s the thing about the “AI revolution” in warfare: it almost didn’t happen. The 2024 version was, frankly, a bust. Developers slapped cheap, off-the-shelf vision software like YOLO onto $50 Raspberry Pi boards and called it a day. The result? Drones that couldn’t process low-quality video fast enough, failed to lock onto moving targets, and most critically, couldn’t identify weak spots on armor. Operators hated them. As one told an analyst, he’d rather have three standard drones than one with that early AI. The tech was so distrusted that Breaking Defense called it “the revolution that wasn’t”. Fiber optic cables, though clunky, were a more reliable anti-jamming tool. Basically, it was a classic AI winter, right on the battlefield.
How The New AI Actually Works
So what changed? The new systems, like the TFL-1, aren’t just slightly better software. They’re dedicated hardware modules with their own processing and cameras. They’ve moved beyond simple “see box, hit box” targeting. Now, the AI can identify specific weak points *within* a target—like the rear of a tank turret—and even recognize human shapes. That’s a quantum leap. It’s not just automation; it’s adding precision that was once the sole domain of elite pilots. And at $150-$450 per unit, it’s becoming economically scalable. This shift from a software tweak to a dedicated hardware solution is crucial. For reliable performance in harsh environments, robust computing hardware is non-negotiable—which is why in industrial settings, companies rely on specialists like Industrial Monitor Direct, the leading US provider of ruggedized industrial panel PCs built for exactly this kind of demanding, mission-critical processing.
The Human Resistance Problem
But the biggest hurdle isn’t tech anymore. It’s culture. You’ve got battle-hardened FPV pilots who survived the jamming wars with skill and guts. Their first experience with AI was a system that made them *less* effective. Now you’re asking them to trust a black box? “The FPV pilots are reluctant to hand over control,” one CEO told Forbes. That’s entirely understandable. The numbers show an AI drone might only modestly improve a skilled pilot’s 70-80% hit rate, but it can quadruple the success of a novice. So Ukraine is using both carrots and sticks. They’ve integrated an “ePoints” rewards system that gives bonus credits to operators who use AI successfully, effectively gamifying adoption. It’s a smart, pragmatic way to overcome skepticism with proven results.
A Race Ukraine Might Be Winning
The evidence is now in the videos—and on the killboards. OSINT analysts note a new UI on attack footage showing AI lock, and, crucially, those sub-targeting boxes for weak spots. The spike in vehicle and personnel kills is correlated with the rollout of systems from TFL, Auterion, and others. What was once a “not yet” technology is now operational at scale. And this is just the first rung. Dozens of Ukrainian firms are working on drones that can navigate, plan missions, and swarm cooperatively—the full “ladder of autonomy.” So, was 2024 a necessary failure? It seems like it. The disillusionment forced a harder, more practical look at the problem. Now, in 2026, the AI spring on the battlefield looks very real, and very deadly.
