SECAI Core · phase 14 of 15 · noun

MITRE Adversarial Threat Landscape for Artificial-Intelligence Systems

A globally accessible framework developed to understand the various techniques and tactics adversaries use to attack AI and machine learning systems

The Explain card

Plain English
MITRE ATLAS is a globally accessible framework that catalogues the tactics and techniques adversaries use against AI and machine learning systems, modelled on the familiar ATT&CK matrix.
Example
A red team planning an assessment of a document-summarising assistant uses ATLAS to structure the engagement, selecting techniques such as prompt injection and model extraction, and reporting findings by technique ID.
Why it matters
A shared vocabulary lets defenders compare coverage, map detections to techniques and communicate AI risk in terms the rest of the security team already understands.
Hook
ATT&CK's sibling, raised on machine learning.

Word knowledge

How the word is built, where it came from, and what it sits beside in memory.

In a sentence

The security team mapped a model-poisoning test against the MITRE Adversarial Threat Landscape for Artificial-Intelligence Systems before running it on the image classifier.

Why these words
  • MITRE English corporate name coined in 1958 identifies the organization responsible for the named resource
  • adversarial Latin adversarius, opponent marks an opposing or hostile relationship
  • threat Old English threat, pressure or menace names the harmful possibility being organized
  • landscape Dutch landschap, region or pictured scene presents the subject as a broad field containing related elements
  • for Old English for, for or on behalf of links the landscape to its technical domain and is omitted from the acronym
  • artificial-intelligence Latin artificialis, made by art, and intelligentia, understanding forms a hyphenated compound modifier specifying the kind of systems
  • systems Greek systēma, organized whole supplies the plural head noun modified by artificial-intelligence
Where it came from
Origin
English institutional coinage by MITRE and its collaborators; ATLAS is formed from Adversarial, Threat, Landscape, Artificial-Intelligence, and Systems
Entered the language
2021
What changed
The project began as the Adversarial ML Threat Matrix in 2020 and received the MITRE ATLAS name with version 2.0 in spring 2021.
How it is spelled
Pattern
The initial letters of the five content words spell ATLAS.
Pattern
Systems takes the regular plural ending -s.
Pattern
The proper name uses title capitalization while MITRE and ATLAS remain fully capitalized.
Breaks the pattern
The function word for does not contribute a letter to ATLAS.
Breaks the pattern
Artificial-Intelligence retains an internal hyphen in the official expanded name, although artificial intelligence is usually open in ordinary prose.
Breaks the pattern
MITRE is written in capitals as a corporate name rather than expanded as part of ATLAS.

Spelled like

  • machine-learning
  • natural-language
  • threat-modeling
Broken into chunks
  • ATLAS
    • ATT&CK
    • D3FEND
  • advers
    • adverse
    • adversary
  • threat
    • threaten
    • threatening
  • scape
    • seascape
    • cityscape
  • artific
    • artifice
    • artificial
  • intellig
    • intelligent
    • intelligible
  • system
    • systemic
    • systematic
What it sits beside

Same subject

  • MITRE ATT&CK
  • adversarial machine learning
  • AI red teaming
  • threat modeling
  • model poisoning

Same shape

  • CAPEC
  • STIX
  • TAXII

MITRE security resources

  • MITRE ATT&CK
  • MITRE D3FEND
  • MITRE CAPEC

AI security assessment vocabulary

  • AI red teaming
  • attack surface
  • threat modeling

Where it sits in the deck

Phase 14: Incident Response, Evaluation, and Knowledge Bases

When detection fires, teams need structured response plans, scoring frameworks, and curated knowledge bases to triage, measure, and learn from AI security events.

Neighbours in this phase