AML.T0103

MITRE ATLAS / Deploy AI Agent

Adversaries launch AI agents inside the victim environment to carry out actions on their behalf, using the environment's own compute, credentials, and network position.

Think of it likeHiring a temp who reports to you, sitting them at a desk inside the office, and letting them do the rounds.

In plain English

Instead of running the attack themselves, the adversary starts an agent in your environment and lets it work. It looks like internal automation and it operates at machine speed.