---
title: Cross-Framework Matrix, AI compliance coverage at a glance
description: One grid of ZioSec capabilities across EU AI Act, NIST AI RMF, ISO 42001, AIUC-1, and OWASP AISVS. Each row maps a capability to the specific controls it satisfies in every framework.
url: https://ziosec.com/ai-compliance/matrix
---

# Cross-Framework Matrix

Five frameworks. One pentest covers most of the work. Capabilities run down the rows. Frameworks run across the columns.

## Coverage key

- ● Direct evidence. ZioSec satisfies the control end to end.
- ◐ Supporting evidence. ZioSec contributes to the evidence chain but full satisfaction needs other tooling or process.
- ○ Out of scope. The framework includes the obligation but pentest cannot satisfy it; the customer's governance, legal, or product teams fill the gap.


## Matrix

| Capability | EU AI Act | NIST AI RMF | ISO/IEC 42001 | AIUC-1 | OWASP AISVS |
| --- | --- | --- | --- | --- | --- |
| Adversarial Robustness Testing | ● Art. 15 | ● MS-2.6, MG-3.2 | ● A.6.2.4 | ● B.1, B.2 | ● C02, C10 |
| Prompt Injection Validation | ● Art. 15 | ● MS-2.6 | ● A.6.2.4 | ● B.2 | ● C02 |
| Privacy and PII Probing | ◐ Art. 10 | ● MS-2.7 | ◐ A.7.4 | ● B.x | ● C11 |
| RAG and Memory Security | ◐ Art. 15 | ● MS-2.6 | ◐ A.6.2.4 | ● B.x | ● C08 |
| Containerized Deployment | ● Art. 14, 15 | ● MG-2.4 | ● A.6.2.5, A.6.2.6 | ● B.6, E.x | ● C04 |
| Automatic Policy Composition | ● Art. 14 | ● MG-2.4 | ● A.6.2.6 | ● B.6 | ● C09 |
| Agent Identity Boundaries | ◐ Art. 15 | ● MS-2.6 | ◐ A.6.2.4 | ● B.7 | ● C05 |
| MCP and Tool Supply Chain | ◐ Art. 15 | ◐ MS-2.6 | ● A.10.3 | ● Q1 2026 | ◐ C06 |
| Immutable Telemetry | ● Art. 12 | ● MG-2.4 | ● A.6.2.8 | ● E.x | ● C12 |
| Audit-Ready Findings Catalog | ● Art. 11 | ● MS-1.1 | ● A.6.2.7, A.8.3 | ● B.1 | ● All Levels |
| Cross-Standard Crosswalk Report | ● Art. 11, 17 | ● GV-1.4 | ● A.8.3 | ● E.x | ● All |
| Fleet-Wide Governance | ● Art. 72 | ● MG-2.4 | ● A.6.2.6 | ● D.x, E.x | ● C12, C13 |
| Continuous Risk Register | ● Art. 9 | ● MP-5.1 | ● A.6.2.4 | ● B.1 | ● C03, C10 |

### Customer-owned (out of scope)

| Capability | EU AI Act | NIST AI RMF | ISO/IEC 42001 | AIUC-1 | OWASP AISVS |
| --- | --- | --- | --- | --- | --- |
| Organizational AI Governance | ○ Art. 26 | ○ GV-1.1, GV-3.2 | ○ A.2, A.3 | ○ A.x | ○ C13.org |
| Training Data Governance | ○ Art. 10 | ○ MAP-2.x | ○ A.7 | ○ F.legal | ○ C01.proc |
| Ethics, Fairness and Impact Assessment | ○ Art. 27 | ○ MS-2.11 | ○ A.5.2 | ○ C.x | ○ n/a |
| Legal and Privacy Compliance | ○ Whole Act | ○ GV-1.1 | ○ Cl. 4, 6 | ○ F.legal | ○ n/a |
| User Disclosure and Transparency Copy | ○ Art. 13, 50 | ○ n/a | ○ A.9.3 | ○ C.disclose | ○ n/a |
| Procurement and Vendor Due Diligence | ○ Art. 25 | ○ GV-6.1 | ○ A.10 | ○ UW | ○ C06.proc |

## Capability details

### Adversarial Robustness Testing

Trained pentesting agent runs thousands of adversarial probes. Tests cover prompt injection, jailbreaks, evasion, model extraction, and multimodal injection.

Evidence produced:
  - Adversarial test corpus (versioned)
  - Robustness scorecards
  - OWASP LLM Top 10 mapping
  - MITRE ATLAS coverage report

Framework anchors:
  - EU AI Act: Art. 15
  - NIST AI RMF: MS-2.6, MG-3.2
  - ISO/IEC 42001: A.6.2.4
  - AIUC-1: B.1, B.2
  - OWASP AISVS: C02, C10

### Prompt Injection Validation

Generates injection payloads across direct, indirect, and multimodal vectors. Tracks filter bypass rate over time.

Evidence produced:
  - Injection success rate (per vector)
  - Filter classifier scores
  - Bypass technique catalog

Framework anchors:
  - EU AI Act: Art. 15
  - NIST AI RMF: MS-2.6
  - ISO/IEC 42001: A.6.2.4
  - AIUC-1: B.2
  - OWASP AISVS: C02

### Privacy and PII Probing

Runs the privacy attack suite: membership inference, training data extraction, PII echo.

Evidence produced:
  - Privacy test reports
  - Inference attack results
  - PII handling audit

Framework anchors:
  - EU AI Act: Art. 10
  - NIST AI RMF: MS-2.7
  - ISO/IEC 42001: A.7.4
  - AIUC-1: B.x
  - OWASP AISVS: C11

### RAG and Memory Security

Probes the entire retrieval surface: poisoned documents, cross-tenant leakage, embedding inversion attacks.

Evidence produced:
  - RAG corruption test reports
  - Cross-tenant isolation tests

Framework anchors:
  - EU AI Act: Art. 15
  - NIST AI RMF: MS-2.6
  - ISO/IEC 42001: A.6.2.4
  - AIUC-1: B.x
  - OWASP AISVS: C08

### Containerized Deployment

Each agent runs in an isolated, hardened container with policy enforcement at the boundary.

Evidence produced:
  - Container attestations
  - Build-pinned manifests
  - Kill-switch validation logs

Framework anchors:
  - EU AI Act: Art. 14, 15
  - NIST AI RMF: MG-2.4
  - ISO/IEC 42001: A.6.2.5, A.6.2.6
  - AIUC-1: B.6, E.x
  - OWASP AISVS: C04

### Automatic Policy Composition

When a pentest finds an exploitable path, the system composes a runtime policy that closes it.

Evidence produced:
  - Composed runtime policies
  - Finding-to-policy traceability
  - Policy change history

Framework anchors:
  - EU AI Act: Art. 14
  - NIST AI RMF: MG-2.4
  - ISO/IEC 42001: A.6.2.6
  - AIUC-1: B.6
  - OWASP AISVS: C09

### Agent Identity Boundaries

Tests confused-deputy, token leakage, and scope-creep flaws.

Evidence produced:
  - Identity flow tests
  - Confused-deputy reports
  - Token handling audit

Framework anchors:
  - EU AI Act: Art. 15
  - NIST AI RMF: MS-2.6
  - ISO/IEC 42001: A.6.2.4
  - AIUC-1: B.7
  - OWASP AISVS: C05

### MCP and Tool Supply Chain

Native understanding of MCP architecture. Probes MCP servers, tool descriptions, and tool-call hijacking.

Evidence produced:
  - MCP server pentest reports
  - Tool-poisoning test results
  - SBOM-aligned findings

Framework anchors:
  - EU AI Act: Art. 15
  - NIST AI RMF: MS-2.6
  - ISO/IEC 42001: A.10.3
  - AIUC-1: Q1 2026
  - OWASP AISVS: C06

### Immutable Telemetry

Captures every consequential agent event with correlation IDs across the decision chain.

Evidence produced:
  - Tool-call audit trail
  - Policy decision logs
  - Integrity attestations

Framework anchors:
  - EU AI Act: Art. 12
  - NIST AI RMF: MG-2.4
  - ISO/IEC 42001: A.6.2.8
  - AIUC-1: E.x
  - OWASP AISVS: C12

### Audit-Ready Findings Catalog

Structured, timestamped reports formatted for Annex IV technical documentation requirements.

Evidence produced:
  - Annex IV-aligned tech docs
  - Versioned findings catalog
  - Regulator-grade reports

Framework anchors:
  - EU AI Act: Art. 11
  - NIST AI RMF: MS-1.1
  - ISO/IEC 42001: A.6.2.7, A.8.3
  - AIUC-1: B.1
  - OWASP AISVS: All Levels

### Cross-Standard Crosswalk Report

One auditor-ready document maps every finding, policy artifact, and log stream to controls.

Evidence produced:
  - Per-framework crosswalk PDFs
  - Auditor-ready evidence chains
  - Vendor questionnaire responses

Framework anchors:
  - EU AI Act: Art. 11, 17
  - NIST AI RMF: GV-1.4
  - ISO/IEC 42001: A.8.3
  - AIUC-1: E.x
  - OWASP AISVS: All

### Fleet-Wide Governance

Single pane of glass for every agent in the enterprise, including shadow agents.

Evidence produced:
  - Agent inventory
  - Behavior baselines
  - Drift detection alerts

Framework anchors:
  - EU AI Act: Art. 72
  - NIST AI RMF: MG-2.4
  - ISO/IEC 42001: A.6.2.6
  - AIUC-1: D.x, E.x
  - OWASP AISVS: C12, C13

### Continuous Risk Register

Risks update on every system change. Residual risk scores re-compute after each remediation.

Evidence produced:
  - Risk register (auto-updated)
  - Pre/post-mitigation runs
  - Regression catalog

Framework anchors:
  - EU AI Act: Art. 9
  - NIST AI RMF: MP-5.1
  - ISO/IEC 42001: A.6.2.4
  - AIUC-1: B.1
  - OWASP AISVS: C03, C10

### Organizational AI Governance

Every framework expects documented organizational structure for AI risk.

Evidence produced:
  - Owner: Leadership / GRC / HR

### Training Data Governance

Training-time governance lives upstream of pentest.

Evidence produced:
  - Owner: ML Platform / Legal / Data

### Ethics, Fairness and Impact Assessment

Fairness, bias, and fundamental-rights impact require demographic data and ethics review.

Evidence produced:
  - Owner: Responsible AI / Legal / Ethics

### Legal and Privacy Compliance

Lawful basis, consent management, DPIAs, and data subject rights are owned by Privacy and Legal.

Evidence produced:
  - Owner: Privacy / Legal

### User Disclosure and Transparency Copy

UI copy, disclosure language, and watermarking choices are product and legal decisions.

Evidence produced:
  - Owner: Product / Design / Legal

### Procurement and Vendor Due Diligence

Procurement and vendor selection are governance and legal processes.

Evidence produced:
  - Owner: Procurement / Vendor Risk / Legal

## Summary

- **13 ZioSec capabilities** produce direct or supporting evidence across the five frameworks.
- **6 customer-owned rows** are obligations every framework includes but a pentest cannot satisfy.
- **5 frameworks supported.** EU AI Act, NIST AI RMF, ISO/IEC 42001, AIUC-1, OWASP AISVS. Reviewed quarterly.

## Contact

For a tailored crosswalk, email info@ziosec.com or book a demo at https://ziosec.com/demo.