What does the MITRE Security Mapping Tool do?

The MITRE Security Mapping Tool™ maps your existing security controls to the MITRE ATLAS AI attack techniques they actually mitigate. Select your AI systems, select your current controls, and the tool produces: a risk-weighted coverage score, a gap analysis showing which ATLAS techniques are uncovered, identification of redundant controls (where two controls cover the same threat), compliance mappings to SOC 2, ISO 27001, NIST AI RMF, and EU AI Act, vendor product recommendations for gap coverage, a visual coverage matrix, and CSV export for further analysis.

What compliance frameworks does the Security Mapping Tool cover?

The tool maps your security controls to four major frameworks: SOC 2 Trust Services Criteria (security, availability, confidentiality) — showing which controls satisfy which criteria and which criteria remain uncovered. ISO 27001:2022 — mapping to Annex A controls as they apply to AI systems. NIST AI RMF — mapping to the Govern, Map, Measure, and Manage functions. EU AI Act — mapping to technical and governance requirements for high-risk AI systems. For each framework, the tool shows a coverage percentage and specifically identifies which requirements your current controls don't address.

What are the most common security control gaps in AI environments?

Five control gaps appear most frequently in mapping assessments: (1) Prompt injection validation — most organisations have general input validation but not LLM-specific prompt injection detection. (2) Agent permission governance — tool access controls for autonomous agents are rarely formalised. (3) RAG retrieval access control — data access controls exist but are rarely applied at the chunk level in retrieval systems. (4) AI-specific runtime monitoring — general SIEM and logging exists but doesn't capture AI-specific signals (prompt anomalies, output anomalies, agent behaviour). (5) AI vendor security assessment — third-party risk processes exist but rarely evaluate AI-specific criteria like training data terms, model behaviour, and inference security.

What is the overlap detection feature?

Overlap detection identifies when two or more of your selected controls mitigate the same MITRE ATLAS technique. This is valuable for budget optimisation: if both Control A and Control B cover the same threat, and you're considering buying Control C which also covers that threat, the overlap panel flags this redundancy before you invest. Many organisations discover they're over-investing in prompt filtering (where three tools cover the same technique) while having zero coverage on agent tool governance or RAG access control — the same investment redirected would dramatically improve overall coverage.

MITRE Security Mapping Tool™ — Map Security Controls to AI Threats (2026)

️ MITRE Security Mapping Tool™ — Map Your Controls Now

Select your AI systems and security controls. Get 5 result views: coverage overview, gap analysis, compliance mapping, vendor recommendations, and coverage matrix. Free CSV export.

Open Mapping Tool → Explore ATLAS Techniques First

Why Controls Mapping Matters

The standard approach to AI security is additive: implement controls, accumulate them over time, and assume coverage improves. The problem is that AI security controls aren't uniformly distributed across the threat landscape. Most organisations end up with multiple overlapping controls for prompt injection (the most visible threat) and zero controls for agent tool governance, RAG access control, and AI supply chain security (the fastest-growing threat categories).

Without explicit mapping from controls to threats, this imbalance is invisible. An organisation might have 15 security controls that collectively cover 6 MITRE ATLAS techniques — while 30+ other techniques remain completely uncovered. The Security Mapping Tool makes this visible, so investment decisions become rational rather than reactive.

The 5 Result Views

Overview

Risk-weighted coverage score, coverage by ATLAS threat category, and executive summary narrative

Gap Analysis

Uncovered techniques with risk severity, blind spot identification, "what-if" investment simulation, and prioritised recommendations

Compliance

Coverage mapped to SOC 2, ISO 27001, NIST AI RMF, and EU AI Act — with specific gap identification per framework

Vendors

Product recommendations mapped to your specific gaps — tools most likely to improve your coverage score

Coverage Matrix

Visual grid of controls vs ATLAS techniques — heat-mapped to show where coverage is concentrated and where it's absent

Compliance Framework Mappings

A key differentiator of the Security Mapping Tool is its compliance output. Most AI security tools tell you your coverage against threats. This tool also tells you what that coverage means for your compliance obligations.

SOC 2

Maps controls to Trust Services Criteria: Security (CC6-CC9), Availability (A1), and Confidentiality (C1).

CC6.1 — Logical and physical access controls
CC6.6 — Logical access to assets
CC7.2 — System monitoring controls
CC8.1 — Change management

ISO 27001:2022

Maps to Annex A controls as they apply to AI systems and LLM deployments.

A.8.20 — Network security controls
A.8.23 — Web filtering (prompt filtering)
A.8.28 — Secure coding (AI supply chain)
A.5.23 — Information security for cloud services

️ NIST AI RMF

Maps to Govern, Map, Measure, and Manage functions for AI risk management.

GOVERN 1.1 — AI risk policies
MAP 1.1 — Risk context identification
MEASURE 2.2 — AI risk evaluation
MANAGE 1.3 — Response and recovery

🇪🇺 EU AI Act

Maps to technical and governance requirements for high-risk AI systems under the EU AI Act.

Art. 9 — Risk management system
Art. 13 — Transparency requirements
Art. 14 — Human oversight controls
Art. 15 — Accuracy and robustness

The Most Common Control Gaps the Tool Finds

No LLM-specific prompt injection detection

General input validation exists but wasn't designed for semantic prompt attacks. Standard WAFs and input sanitisation detect SQL injection and XSS but are blind to "ignore previous instructions" variants. Gap: AML.T0051, AML.T0054 partially uncovered despite existing controls.

No agent tool permission governance

IAM controls manage human access to systems. They don't govern which tools AI agents can call, with what permissions, or under what conditions. Most organisations have zero formal control over agent tool scope. Gap: AML.T0080, AML.T0090 completely uncovered.

Data access controls not applied at RAG retrieval layer

Data access controls govern who can read files and databases. They don't automatically apply to which chunks get retrieved from a vector database. A user authorised to read a document category may still receive retrieval results from restricted documents. Gap: cross-tenant leakage techniques uncovered despite strong underlying data controls.

General monitoring doesn't capture AI-specific signals

SIEM, logging, and anomaly detection capture network events, authentication events, and file access events. They don't capture prompt anomalies, output anomalies, agent behaviour deviations, or retrieval scope violations. Gap: runtime monitoring ATLAS techniques uncovered despite substantial general monitoring investment.

Third-party risk process not extended to AI vendors

Vendor risk management assesses financial stability, SLAs, and general security posture. It rarely evaluates AI-specific criteria: training data terms, model behaviour, inference security, or model version change management. Gap: supply chain ATLAS techniques largely uncovered despite mature vendor risk programme.

Who Should Use the Mapping Tool?

️

Security Architects

Validate that your AI security architecture actually covers the ATLAS techniques you think it does — and identify the architectural gaps before they become incidents

Compliance Teams

Map existing controls to SOC 2, ISO 27001, NIST AI RMF, and EU AI Act in minutes — identify specific gaps per framework without manual cross-reference work

CISOs & Security Leaders

Use the "what-if" investment simulation to model the coverage impact of specific control investments before budget requests — justify spend with data, not intuition

Red Teams & Pen Testers

Use blind spot analysis to identify the ATLAS techniques most likely to succeed against your current control stack — focus your testing where defences are weakest

Coverage Matrix — See Everything at Once

The visual coverage matrix shows all your controls against all MITRE ATLAS techniques in a single heat-mapped grid. Instantly see where coverage clusters and where the gaps are.

Open Security Mapping Tool → Benchmark Your Score

Frequently Asked Questions

What does "overlap detection" mean?

Overlap detection flags when two or more of your selected controls mitigate the same MITRE ATLAS technique. This matters for budget optimisation — if you're considering a new control that covers threats already addressed by two existing controls, the overlap panel shows this redundancy so you can redirect that investment to actual gaps. Many organisations discover they're triple-covered on prompt injection while having zero coverage on agent governance.

Can I export my mapping results?

Yes — the tool includes CSV export covering your control-to-technique mappings, coverage percentages, gap analysis, and prioritised recommendations. The CSV is structured for import into spreadsheets, risk registers, or GRC tools. For PDF output, use the browser print function (Ctrl+P) which renders the styled results view.

What does the "what-if" simulation do?

The gap analysis tab includes an investment simulation: select any uncovered control and the tool immediately shows the coverage impact — which ATLAS techniques would be addressed, the percentage coverage gain, and where that would place you in the risk weighting. This makes control investment decisions data-driven: "adding AI-specific runtime monitoring would close 8 ATLAS techniques and improve our coverage by 14%."

How does this differ from the Coverage Calculator?

The Coverage Calculator measures your current coverage level across the full ATLAS framework in 3 steps — it answers "how much of ATLAS are we covered against?" The Security Mapping Tool maps specific controls to specific techniques — it answers "which controls cover which threats, and where exactly are the gaps?" They're complementary: run the Calculator first to get your overall score, then use the Mapping Tool to understand exactly which controls are responsible for which coverage and where to focus next.

MITRE Security Mapping Tool™: Map Your Security Controls to Real AI Threats (2026)