What it is about
MITRE ATLAS is a free, living knowledge base for AI security. It organises attacks on AI systems into tactics, such as reconnaissance, gaining access to a model, and stealing data, and then lists the techniques under each one. Real case studies are linked to the techniques, so you can see how an attack worked in practice. It is updated regularly as new attacks are found.
Why it matters
Before ATLAS, AI threat modelling was mostly guesswork, because network attack maps did not cover model theft, data poisoning, or prompt tricks. ATLAS gives red teams, defenders, and auditors one shared language for AI attacks. Pair it with a governance framework and you have both the attack view and the control view.
Who should read it
AI red teamers, SOC analysts, threat modellers, and students moving into AI security.
Source
Original document by MITRE.
The original document is not hosted here. Our note above is our own summary and opinion, in simple words.
Want to learn this properly?
Go deeper with Cyber Legends training and services.
