We've written quite a bit about the scariest trend in military hardware: autonomous weapons systems. What used to be science fiction has moved from the drawing board to the showroom floor. Defense contractors openly market AI-driven weapons that think, move, and strike with minimal human oversight. The fact they can fire missiles on their own has gone from a nightmare scenario to a selling point.
The U.S. Navy has already deployed autonomous submarines like the 51-foot Orca capable of anti-submarine warfare, while AI-controlled F-16 fighter jets have conducted live dogfights at supersonic speeds, and fully unmanned warships like DARPA's Defiant now patrol the seas with no crew aboard whatsoever.
But at last week's Association of the U.S. Army symposium, the future of warfare was pushed even further into the autonomous realm -- and may have finally crossed a dangerous line.
Of course, there's an obvious advantage if a military craft has no crew inside. If a mission can be accomplished without risking soldiers' lives, that's undeniably a good thing. Autonomous vehicles can venture into combat zones, contaminated areas, dense air defense networks, or positions under constant artillery fire--without putting a single person at risk. If they get destroyed, we lose hardware. We don't lose anyone's son or daughter.
The idea is compelling, and AI technology is making it possible. But you'd hope that somewhere in a bunker, there's a human behind the controls making the decision to actually fire the missile. That's the case with remote-operated Patriot missile launchers in Ukraine. They are in the field right now, allowing operators to defend airspace from the safety of hidden command posts while Russian forces hunt for crews that aren't there. But the launchers don't launch themselves.
That might be about to change.
At the recent AUSA symposium, defense giants Oshkosh Defense and Raytheon unveiled their latest autonomous launchers--and they're far more sophisticated than anything we've seen before. These aren't remotely piloted vehicles requiring constant human control. They're autonomous platforms designed to operate independently across vast distances.
Oshkosh's Family of Multi-Mission Autonomous Vehicles (FMAV) comes in three sizes, from light trucks carrying anti-drone weapons to massive 10×10 platforms hauling Tomahawk cruise missiles--weapons designed to strike targets hundreds of miles away. Raytheon's DeepFires system can pivot between offensive and defensive roles, carrying everything from AIM-9X air defense missiles to Patriot interceptors to containerized Tomahawks.
The operational concept is disturbingly simple: an operator can control up to six vehicles simultaneously using waypoint navigation. The vehicles plot their own routes, navigate terrain autonomously, and position themselves at firing points with minimal oversight. Alternatively, they can follow a single crewed vehicle in a convoy, then disperse to pre-programmed locations on command.
They're small enough to fit inside a C-130 cargo plane, meaning they can be rapidly deployed anywhere in the world. And they're being specifically marketed for "distributed operations" across the vast Pacific theater, where they could be scattered across hundreds of islands, waiting for orders to strike.
Or do they need orders at all?
The systems are explicitly designed to operate with "minimal communication" and can function when "comms are degraded." It's considered a feature, not a bug in modern warfare, because jamming and cyber attacks routinely disrupt communications. So engineers have built autonomy that doesn't depend on constant human oversight.
But when it comes to missiles that can wipe out vast numbers of civilians, humans should probably be in the loop.
That's not necessarily part of the program. The promotional materials emphasize that these vehicles will "self-navigate from point A to point B" and execute pre-programmed firing sequences. They're designed to complete their missions even if they lose contact with human controllers. That's the whole point--resilience in contested environments.
What if the war ends while communications are down? What if diplomats in Geneva shake hands on a ceasefire while autonomous missile trucks in the Pacific are still executing strike packages loaded hours or days earlier? Are we about to deploy weapons that can continue fighting after we've decided to stop fighting?
Defense contractors present these scenarios as edge cases, unlikely combinations of worst-case circumstances. But in warfare, worst-case circumstances happen with disturbing regularity. Communication fails. Orders get confused. Technology breaks down precisely when you need it most.
The executives at AUSA were excited about "magazine depth"--how many missiles their autonomous trucks can carry before needing resupply. They were proud of the "resilience" and "independence" of their systems. Nobody seemed to be asking the more fundamental question: how do you turn them off?
Because these launchers don't need a crew, they can be hidden across Europe or the Indo-Pacific theater, wait patiently for years, until the crises arises. As one executive boasted, these uncrewed vehicles can hide across the globe for whenever they are needed.
The technology is impressive. The intention is good--keeping soldiers safe. But once those systems are deployed, concealed across theaters of operation, programmed with strike packages and designed to function without constant human oversight--how exactly do we find them all? How do we recall weapons that were built specifically to hide and operate independently?
When the missiles start launching, let's hope it's because someone ordered them to -- and not because they weren't able to order them to stop.