AML.T0050

MITRE ATLAS / Command and Scripting Interpreter

Adversaries abuse command and script interpreters to execute commands, scripts, or binaries. In AI environments this commonly means shells reachable from a notebook, a training job, or a tool an agent is allowed to call.

Think of it likeOnce inside the building, using the intercom system that was already wired to every room.

In plain English

Interpreters are already there and already trusted. A notebook cell, a training entrypoint, or an agent's shell tool gives an attacker ordinary command execution inside your AI environment.