The adversarial threat landscape for AI systems

MITRE ATLAS, made visible and runnable.

Every technique is badged by how honestly it can be shown — and for the ones that can truly be simulated client-side, we ship deterministic guided visualizations that need no API key.

— techniques runnable in-browser — techniques total

ATLAS™ is a trademark of The MITRE Corporation. This is an independent educational visualization; it is not affiliated with or endorsed by MITRE. Content is generated from the official atlas.mitre.org data.

Technique index

Static pages for 101 techniques under 16 tactics. No JavaScript required; generated from the same data as the matrix.

AML.TA0002 Reconnaissance (8)

AML.TA0003 Resource Development (10)

AML.TA0004 Initial Access (7)

AML.TA0000 AI Model Access (4)

AML.TA0005 Execution (6)

AML.TA0006 Persistence (9)

AML.TA0012 Privilege Escalation (4)

AML.TA0007 Defense Evasion (15)

AML.TA0013 Credential Access (6)

AML.TA0008 Discovery (9)

AML.TA0015 Lateral Movement (2)

AML.TA0009 Collection (4)

AML.TA0001 AI Attack Staging (6)

AML.TA0014 Command and Control (3)

AML.TA0010 Exfiltration (6)

AML.TA0011 Impact (9)

Techniques without a tactic mapping (3)