MITRE Corporation

MITRE is a federally funded research and development center (FFRDC) operating multiple FFRDCs sponsored by the U.S. government. In AI security, MITRE is the publisher of MITRE ATLAS (Adversarial Threat Landscape for Artificial-Intelligence Systems), the primary adversarial technique knowledge base for AI/ML systems.

AI Security Role

MITRE ATLAS extends the ATT&CK methodology to AI/ML systems, documenting adversary techniques, sub-techniques, mitigations, and real-world case studies. ATLAS is exclusively adversary-centric — it catalogs attacks but does not provide defensive control specifications.

Q1 2026 Activity

MITRE ATLAS had the most rapid agentic threat coverage expansion of any framework in Q1 2026:

• v5.3.0 (January 2026) — 18+ new techniques; Zenity Labs contributions; SesameOp case study (AML.CS0042)
• v5.4.0 (February 2026) — “Publish Poisoned AI Agent Tool,” “Escape to Host”
• OpenClaw Investigation (February 9, 2026) — 7 new techniques; CVE-2026-25253; case study AML.CS0050

ATLAS now covers 84 techniques, 16 tactics, 56 sub-techniques, 32 mitigations, and 42 case studies.

Key Publications

• MITRE ATLAS — adversarial AI threat knowledge base
• Arsenal CALDERA plugin — automated red team integration (limited 2026 updates)
• OpenClaw Investigation (February 9, 2026) — rapid response threat report