A threat actor has used an open-source AI agent to automate post-exploitation activity inside a government network — marking one of the first publicly documented cases of an AI agent deployed as an au