AI Agents Expand Permission, Supply-Chain, and Offensive-Operations Risk
An FBI warning reports ARTEX agents used in intrusions spanning reconnaissance to data exfiltration, with more than 1,600 exposed consoles and persistent forks raising exposure concerns; the tool itself is not inherently malicious. A separate GAO snippet says multiple AI agents tested malicious code against security products, but scale and impact are unclear. Google reports patch-to-exploit automation and an exposed exploit_agent server with stolen victim data, while CERT-In and Georgia reporting underscore unsafe access to live systems; strong external authorization, least privilege, and egress controls remain essential.
Sources (23)
Updated Nov 21, 2026