AML.TA0005

MITRE ATLAS / Execution

The adversary tries to run malicious code or instructions on the system, using techniques such as LLM Prompt Injection, User Execution, and AI Agent Tool Invocation.

Think of it likeTricking someone inside the house into carrying out your orders, like slipping a fake instruction into the family's chore list.

In plain English

Attackers get the AI system to run their code or hidden instructions, for example by smuggling commands into a prompt that an LLM or agent then obeys.