Runaway Military Drone Crosses Three States After Losing Contact

Compiled by the editorial desk with reference to official statements, public filings, and industry reports.

In late January 2017, a routine training exercise in southern Arizona turned into an unplanned cross-country journey when a US Army RQ-7B Shadow drone lost all communication with its ground control station shortly after launch. The unmanned aircraft, designed for surveillance and with a normal operational radius of about 77 miles (120 km), continued flying north for more than five hours, eventually crossing the Rocky Mountains and crashing near Denver, Colorado—roughly 632 miles from its starting point.

The drone, part of the Shadow RQ-7Bv2 series, was expected to either circle a designated point, return to base, or land automatically if it lost its link to operators. Instead, a malfunction prevented those emergency protocols from activating. According to military personnel who tracked the aircraft by radar, the drone climbed to an altitude of approximately 12,000 feet (4,000 meters) to clear the Rockies—conditions it was not built for—before fuel exhaustion likely brought it down.

Why the drone failed to follow its contingency procedures remains unclear. Some speculate that a memory error may have stored the coordinates of a previous home base in Washington state, prompting the aircraft to attempt a return flight. Others suggest the emergency systems simply never engaged, leaving the drone to fly on autopilot until it ran out of fuel. No official explanation has been released.

Autonomy and the Risk of Unpredictable Behavior

The incident highlights a growing challenge for militaries integrating autonomous systems: even well-tested drones can behave in unforeseen ways when their programming encounters unexpected situations. The Shadow drone, which costs about $1.5 million, is unarmed and used primarily for reconnaissance. But its armed counterparts, such as the MQ-1 Predator, carry Hellfire and Griffin missiles and are used in combat strikes. Had an armed drone experienced a similar malfunction, the consequences could have been far more severe.

Autonomous robots, like insects, operate based on pre-programmed responses to specific stimuli. When those responses are triggered incorrectly—or fail to trigger at all—the results can be unpredictable. The Shadow drone's flight over three states and a major mountain range with no human intervention illustrates the potential for such systems to act in ways their designers did not anticipate.

This is not a hypothetical concern. In 2015, the Future of Life Institute released an open letter calling for a ban on autonomous weapons, signed by more than a thousand AI researchers. The letter argues that allowing machines to make life-and-death decisions without human oversight poses a grave risk. Yet governments have shown little willingness to halt development of such systems, citing strategic advantages and the difficulty of verifying compliance with any ban.

For now, the Shadow drone's journey serves as a cautionary tale. It demonstrates that even a relatively simple bug in a complex system can have outsized consequences when that system is deployed in the real world. As militaries continue to rely on increasingly autonomous platforms, the question of how to ensure they behave as intended—especially when armed—remains unresolved.