SpaceX, xAI join secretive $100M drone challenge

The US Department of Defense has quietly kicked off a six-month, $100 million competition that sounds like science fiction: build autonomous drone swarms that can take spoken commands and turn them into coordinated battlefield action.

The challenge is part of the Pentagon’s broader AI Acceleration Strategy, an effort to expand artificial intelligence across military planning, logistics, and combat systems. But this particular contest zeroes in on something more specific… and more controversial. The goal is to create software that can translate a commander’s voice into digital instructions that multiple drones can execute together, whether in the air or at sea.

According to Bloomberg, SpaceX and its artificial intelligence subsidiary xAI are among a small group selected to compete. The entry of Musk’s companies into a competition explicitly tied to offensive military applications marks a notable shift. Musk has previously argued against creating “new tools for killing people” and, in 2015, signed an open letter warning about the dangers of autonomous weapons. Yet the Pentagon’s own framing of this contest makes clear the intended stakes: in a January announcement, a defense official said the human-machine interaction “will directly impact the lethality and effectiveness of these systems.”

The competition is being run jointly by the Defense Innovation Unit and the Defense Autonomous Warfare Group, a new entity under the US Special Operations Command. It unfolds in five phases, beginning with software development and culminating in live operational demonstrations. Early stages focus purely on code. Later ones involve real-world platforms, “target-related awareness and sharing,” and eventually what the Pentagon describes as “launch to termination.”

While drone light shows often create the impression that swarming technology is already mature, experts say that’s misleading. Those displays typically rely on pre-programmed routes and centralized control systems. A true military swarm, by contrast, requires each drone to share information, adapt if others are lost, and make distributed decisions without a single point of failure — all while operating in GPS-denied or jammed environments.

That’s where things get complicated.

Battlefields are saturated with electronic interference. Bandwidth is limited. Communications can be disrupted. For a swarm to function under those conditions, each drone must carry significant onboard processing power and operate semi-independently. Translating a spoken order into coordinated action across dozens, or even hundreds, of systems is far more complex than programming a drone to follow a fixed path.

Bloomberg also reports that OpenAI is supporting a bid led by Applied Intuition. In that proposal, OpenAI’s role would be limited to the “mission control” element — converting voice and other instructions into digital commands — not operating the drones or handling weapons integration. A spokesperson said the company would ensure its tools are used in line with its policies.

SpaceX and xAI, by contrast, are expected to work across the full project scope, according to people familiar with the matter.

xAI has been actively recruiting engineers with active “secret” or “top secret” US security clearances, seeking candidates experienced with the Department of Defense and federal contractors. The company previously secured a $200 million Pentagon contract and announced in December that it would integrate its Grok chatbot into government systems.

Still, even within defense circles, the idea of plugging generative AI into weapons platforms is raising eyebrows. Large language models — the technology behind chatbots — are known to produce errors and so-called hallucinations. Several people familiar with the contest told Bloomberg it will be critical to limit generative AI to translation functions and ensure humans remain firmly in the loop.

The Pentagon’s new strategy seeks to “unleash” AI agents across military functions. Whether voice-controlled swarms become a reliable battlefield tool, or remain an ambitious experiment, may hinge on whether engineers can turn spoken words into resilient, coordinated action under the harshest conditions imaginable.

For now, the $100 million prize signals one thing clearly: the race to define AI-driven warfare is accelerating.

More: Texas sues Anzu Robotics over alleged DJI drone deception