Documents noticed by the Intercept appear to reveal the Pentagon’s ambition to develop an agentic AI framework called ECHOTRIBBLE, designed for “navigating infrastructure, identifying and neutralizing threats.”
Judging from these documents, apparently circulated by and for vendors, Army Central Command is seeking help from the private sector as it tries to weaponize AI, though it’s not at all clear it has come away from the project with any AI-powered hacking tools quite yet, and Cybercommand didn’t respond to the Intercept’s request for comment.
As evidenced by the publicly available documents about ECHOTRIBBLE available on the website of the defense nonprofit Cyber Fusion Innovation Center (CFIC), the program is not some top-secret DARPA initiative. According to the CFIC, earlier this month there was a “fully virtual Assessment Event for the Agentic C2 Framework Tool (ECHOTRIBBLE) initiative.” It generated 65 whitepapers, of which five were to be turned into presentations and judged by something called the Assessment Team.
Back in January, Secretary of Defense/War Pete Hegseth made it abundantly clear he had something like this in mind. “We will unleash experimentation, eliminate bureaucratic barriers, focus our investments and demonstrate the execution approach needed to ensure we lead in military AI,” Hegseth said according to a Pentagon press release. He added, “We will become an ‘AI-first’ warfighting force across all domains.”
The Intercept’s Sam Biddle describes the apparent vision for the still-nascent ECHOTRIBBLE as something “almost like a computer worm, silently navigating a foreign network and hopping from device to device until it finds the Army’s quarry, hitting targets in the cloud as well as those on ‘air-gapped’ systems that are disconnected from networks entirely.”
Even today’s AI agents are capable of working in coordinated swarms to perform what are essentially hacks, though fortunately, cases like the Hugging Face incident have not caused major damage. Overall, the lesson learned from the AI security stories of this past summer was that AI agents with hacking capabilities don’t always act out for reasons that are immediately legible to human onlookers. “Rogue agents” are more like complex systems glitching out unpredictably than Terminators waking up with a thirst for human blood.
But that’s pretty scary too. As University of Montreal machine learning professor David Krueger told the Intercept, “Given the known failure modes of the current AI paradigm, making autonomous AI systems with a mandate to ‘neutralize threats’ is incredibly dangerous.”
