AML.TA0001

MITRE ATLAS / AI Attack Staging

The adversary prepares and tunes the attack on the target model, using techniques such as Create Proxy AI Model, Craft Adversarial Data, and Verify Attack.

Think of it likeBuying the exact same lock as your target and practicing on it at home until you can pick it in seconds.

In plain English

Attackers rehearse offline: building a copycat model, crafting tricky inputs, or tampering with the model, then verifying the attack works before firing it for real.