Military AR Headset Prototype Lets Soldiers Control Drones with Eye Movements and Voice Commands

Defense-tech company Anduril has released new details about its augmented-reality headset prototype developed with Meta, revealing a system that could allow soldiers to order drone strikes using only eye movements and voice commands. The headset, still years from production, represents a major leap in human-machine teaming for military operations.

“My fundamental goal is to optimize the human as a weapons system,” said Quay Barnett, Anduril vice president and former Army Special Operations commander. The system is designed to synchronize drones and soldiers, enabling them to share information and make decisions as a unified entity.

Anduril is working on two separate projects. The first, under a $159 million Army contract called Soldier Born Mission Command (SBMC), involves AR glasses that attach to existing helmets. The second is a self-funded initiative called EagleEye, which includes a custom helmet and headset combo that Anduril hopes the military will eventually adopt.

Background

The efforts come after a previous attempt by Microsoft to build a military AR headset failed. Microsoft had secured a $22 billion production contract for its HoloLens-based system, but the technology did not prove viable and the contract was canceled. Anduril and Meta aim to avoid similar pitfalls by integrating advanced AI and software.

The glasses will overlay data such as maps, drone positions, and AI-recognized targets onto the soldier's field of view. Soldiers will interact using natural language—for example, requesting evacuation routes or planning movements around restricted zones.

Anduril is testing large language models from Google (Gemini), Meta (Llama), and even Anthropic (Claude) to translate speech into commands. The backbone is Anduril’s Lattice software, which aggregates data from various military hardware into a single operational picture. In March, the Army committed $20 billion to integrate Lattice across its infrastructure.

Barnett’s team is designing the headset to execute complex sequences. A soldier might send a drone to surveil an area, instruct it to return upon finding an artillery unit, and then receive recommended actions—such as dispatching another drone to strike—that must be approved through the chain of command.

What This Means

If successful, the headset could dramatically speed up decision-making on the battlefield, reducing the time from target identification to action. However, the reliance on AI for translation and target recognition raises ethical questions about autonomous systems in warfare. Both Anduril and Meta emphasize that human oversight remains mandatory for lethal actions.

Production is not expected until at least 2028, and the Army has not yet committed to a final design. The previous failure of Microsoft’s headset underscores the technical challenges involved.