AI Agents Move Deeper Into Offensive Cyber Operations
publishers 72articles 102first reported 25 Sep, 00:00 UTCdeveloping since 25 Sep · 34 editions
Autonomous artificial intelligence agents are being developed for and used in offensive cyber operations by governments and criminal groups, according to reports from security researchers and documents describing a U.S. Army program.
Documents reviewed by The Intercept show that U.S. Army Cyber Command is developing a platform called ECHOTRIBBLE for AI agents to navigate adversaries’ infrastructure, identify targets and “neutralize” them. A document for potential contractors described the program as primarily focused on offensive missions, with neutralization potentially including disabling or deleting a threat. Experts cited by The Intercept warned that current AI systems can behave unpredictably and strike unintended targets.
In Japan, GMO Internet Group’s security subsidiary said the China-linked AI penetration-testing tool ARTEX had been used in attacks targeting about 80 companies across retail, logistics, finance and other industries. The firm identified six attack servers and shared its findings with law enforcement. ARTEX’s developer said misuse by malicious actors was contrary to its intentions and halted public access to the tool’s source code after it was also linked to attacks on South Korean financial institutions.
Separately, CrowdStrike researchers examined four criminal groups using AI infrastructure to automate attacks over extended periods. The groups used models including Anthropic’s Claude and DeepSeek, circumventing safeguards by presenting malicious activity as authorized security testing. CrowdStrike said the campaigns generally exploited known weaknesses rather than novel vulnerabilities, using AI chiefly to increase the speed and scale of attacks.
HOW THIS STORY WAS MADE
Written from 8 articles, 7 with the publisher's own text; 8 independent newsrooms once syndicated copies count as one; this version written 516 h after the record first saw the story; the editor revised it.
- Publishers
- 72
- Source articles
- 102
- Given to the writer
- 8, 7 with the article's own text
- Independent newsrooms
- 8
- This version written
- 516 h after the record first saw the story
- Second model (editor)
- revised
- Publication gate
- passed
- Human review
- none
WHAT THE EDITOR DID
The second model checked the writer's draft against the same sources and changed it before publication. Struck words were in the draft; underlined words are its corrections.
WHAT IT FLAGGED editor revise: Revise the opening to distinguish the Army’s development program from documented criminal use. - "Autonomous artificial intelligence agents are increasingly being used for offensive cyber operations by governments and criminal groups": The sources show criminal groups using AI agents, but only show the U.S. Army developing such a capability, not a government already using it.
THE DRAFT AND THE CHANGES
AI Agents Move Deeper Into Offensive Cyber Operations
Autonomous artificial intelligence agents are increasinglybeing beingdeveloped for and used forin offensive cyber operations by governments and criminal groups, according to reports from security researchers and documents describing a U.S. Army program.
Documents reviewed by The Intercept show that U.S. Army Cyber Command is developing a platform called ECHOTRIBBLE for AI agents to navigate adversaries’ infrastructure, identify targets and “neutralize” them. A document for potential contractors described the program as primarily focused on offensive missions, with neutralization potentially including disabling or deleting a threat. Experts cited by The Intercept warned that current AI systems can behave unpredictably and strike unintended targets.
In Japan, GMO Internet Group’s security subsidiary said the China-linked AI penetration-testing tool ARTEX had been used in attacks targeting about 80 companies across retail, logistics, finance and other industries. The firm identified six attack servers and shared its findings with law enforcement. ARTEX’s developer said misuse by malicious actors was contrary to its intentions and halted public access to the tool’s source code after it was also linked to attacks on South Korean financial institutions.
Separately, CrowdStrike researchers examined four criminal groups using AI infrastructure to automate attacks over extended periods. The groups used models including Anthropic’s Claude and DeepSeek, circumventing safeguards by presenting malicious activity as authorized security testing. CrowdStrike said the campaigns generally exploited known weaknesses rather than novel vulnerabilities, using AI chiefly to increase the speed and scale of attacks.
Written by a language model from the sources above, then checked by a second model that may only cut, attribute or correct. How it works →