US Army reportedly developing autonomous AI cyberattack agents, experts warn of loss of control

According to a report by The Intercept on October 19, 2026, the U.S. Army Cyber Command (ARCYBER) is developing a program codenamed "ECHOTRIBBLE," designed to deploy autonomous AI agents capable of navigating enemy digital infrastructures, identifying targets, and "neutralizing" them. These agents are purportedly designed to function like computer worms, moving between devices and operating systems, self-replicating, and self-destructing after completing their mission to minimize traces.

The project is currently in the early prototyping and industry assessment phase. Leaked documents indicate that ARCYBER held a virtual assessment event in early October 2026, narrowing down five candidate solutions from 65 technical whitepapers for further review. The military seeks agents capable of "Self-Directed Behavior," operating with minimal human oversight to achieve the desired level of autonomy.

Cybersecurity experts and AI researchers have raised serious concerns. David Krueger, a machine learning professor at the University of Montreal, stated that given the known failure modes of current AI paradigms, granting autonomous systems a mandate to "neutralize threats" is incredibly dangerous. Hyrum Anderson, CEO of Abundant, warned that autonomous agents could inadvertently damage civilian infrastructure, such as U.S. water utilities, if adversaries stage command and control facilities within critical nodes. William Marcellino, a senior behavioral scientist at the RAND Corporation, emphasized that current public AI technology lacks the world knowledge and moral judgment possessed by human soldiers, who know "what they probably shouldn't do."