US Army Cyber Command develops autonomous AI cyber weapons

Internal documents obtained by The Intercept reveal that the US Army Cyber Command (ARCYBER) is developing a platform codenamed ‘ECHOTRIBBLE’ to deploy autonomous AI agents as cyber weapons. These agents, powered by machine learning and large language models, are designed to navigate foreign networks independently to identify and ‘neutralize’ targets.

Neutralization could involve temporarily or permanently disabling a system or deleting data. The software is intended to function similarly to a computer worm, moving silently between devices across both cloud environments and ‘air-gapped’ systems without network connections. Once a mission is complete, the agents are designed to self-destruct to avoid detection.

While the system will use predefined triggers, it is intended to operate with minimal human supervision and ‘self-directed behavior.’ Experts, including machine learning professor David Krueger, have warned that granting autonomous systems the mandate to neutralize threats is extremely dangerous due to known failure modes in current AI paradigms.