SSG coverage claim · MITRE ATLAS-mapped
Browse AI-system risk by MITRE ATLAS
Browse the MITRE ATLAS v2026.08 tactic & technique taxonomy for AI systems, alongside SSG's own derived coverage tier for each technique — see the methodology note below for exactly what that tier does and does not claim.
An adversary may craft malicious prompts as inputs to an LLM that cause the LLM to act in unintended ways. These "prompt injections" are often designed to cause the model to ignore aspects of its original instructions and follow the adversary's instructions instead.
Tactics
Execution
ATT&CK cross-reference
No Enterprise ATT&CK cross-reference.
ATLAS mitigations & SSG coverage
AML.M0019
Control Access to AI Models and Data in Production
Partial
Fail-closed entitlement gating on production value surfaces
AML.M0020
Generative AI Guardrails
Partial
Guards the agent's actions at the tool boundary, not the model's token output
AML.M0021
Generative AI Guidelines
Not addressed
Prompt and system-instruction practice
AML.M0022
Generative AI Model Alignment
Not addressed
Model alignment training
AML.M0024
AI Telemetry Logging
Partial
Detective control: records every governed call; does not block it
AML.M0033
Input and Output Validation for AI Agent Components
Enforced
Content matchers validate tool arguments before execution
AML.M0035
AI Red Team
Partial
SHAKEDOWN benchmark exercises containment; not a full red-team program
Methodology. Catalog: MITRE ATLAS v2026.08 (atlas.mitre.org).
Coverage tiers are derived from MITRE's own published mitigation↔technique mappings — never hand-assigned per technique.
SSG's claim is limited to which ATLAS mitigations it implements as controls; it does not claim to cover ATLAS techniques directly.
A tier records which SSG control exists for a MITRE-published mitigation — it is not a statement about outcome, and it is not a certification, an audit result, or an endorsement by MITRE.