Skip to content

Compliance & governance / Give evidence a context

From a finding
to a framework.

Connect security findings to controls, articles, and techniques. Use the mappings and signed reports to support an assessment with inspectable evidence.

A framework mapping is not certification. A signed report records its assessment; it does not establish universal safety or legal compliance.

The framework atlas

Different lenses. Traceable findings.

The 184 detection rules supply findings that can be mapped across these 9 frameworks. The sections below show selected controls for orientation, rather than claiming complete framework coverage.

01 / Framework lens

OWASP MCP Top 10

A security risk taxonomy for MCP servers, including prompt injection, tool poisoning, command injection, data exfiltration, privilege escalation, and supply-chain risks.

MCP01Prompt Injection
MCP02Tool Poisoning
MCP03Command Injection
MCP04Data Exfiltration
MCP05Privilege Escalation
MCP06Excessive Permissions
MCP07Insecure Configuration
MCP08Dependency Vulnerabilities
MCP09Logging & Monitoring
MCP10Supply Chain

02 / Framework lens

OWASP Agentic Top 10

Security risks specific to agentic AI applications. Mappings connect findings to areas such as goal hijacking, tool misuse, identity abuse, and context poisoning.

ASI01Agent Goal Hijack
ASI02Tool Misuse
ASI03Identity & Privilege Abuse
ASI04Agentic Supply Chain
ASI05Unexpected Code Execution
ASI06Memory & Context Poisoning
ASI07Insecure Inter-Agent Communication

03 / Framework lens

MITRE ATLAS

Adversarial Threat Landscape for AI Systems. ATLAS techniques mapped to MCP-specific detection patterns covering LLM prompt injection, data leakage, context poisoning, and agent manipulation.

AML.T0054LLM Prompt Injection
AML.T0057LLM Data Leakage
AML.T0058AI Agent Context Poisoning
AML.T0059Memory Manipulation
AML.T0060Modify AI Agent Configuration
AML.T0061Thread Injection

04 / Framework lens

NIST AI RMF

The NIST AI Risk Management Framework organizes AI risk management. Relevant findings map to GOVERN and MEASURE functions through audit evidence and human oversight checks.

GOVERN 1.7Human Override Mechanisms
MEASURE 2.6Audit Evidence & Logging

05 / Framework lens

ISO 27001

An information security management standard. Relevant findings map to selected Annex A controls for logging, access control, cryptography, supplier relationships, and system security.

A.5.14Information Transfer
A.5.15Access Control
A.5.17Authentication Information
A.5.20Addressing Security in Supplier Agreements
A.5.21Managing ICT Supply Chain
A.8.15Logging
A.8.22Segregation of Networks
A.8.24Use of Cryptography

06 / Framework lens

ISO 42001

An AI management system standard. Selected control mappings address human oversight and AI transparency; they do not assess the whole management system.

A.8.1AI System Transparency
A.9.1Human Control of AI Systems
A.9.2Human Override

07 / Framework lens

EU AI Act

European regulation on artificial intelligence. Selected findings map to record-keeping, human oversight, robustness, and cybersecurity provisions. Legal applicability requires a separate assessment.

Art. 12Record-keeping (Logging)
Art. 14Human Oversight
Art. 15Accuracy, Robustness, Cybersecurity

08 / Framework lens

CoSAI MCP Security

The Coalition for Secure AI threat model for MCP. Relevant findings connect to authentication, authorization, tool safety, trust boundaries, and supply-chain integrity.

MCP-T1/T2Authentication & Authorization
MCP-T4Tool Output Safety
MCP-T5Cross-Trust-Boundary Data Flow
MCP-T6/T11Supply Chain Integrity
MCP-T8Runtime Sandbox
MCP-T9Multi-Agent Collusion
MCP-T10Availability & Resilience
MCP-T12Audit & Monitoring

09 / Framework lens

MAESTRO

Multi-Agent Evaluation and Security Testing for Robust Operations. Layered security model for AI agent systems covering trust, isolation, observability, and governance.

L3Agent Layer - Integrity
L4Deployment Layer - Isolation
L5Observability Layer - Logging
L7Ecosystem Layer - Trust

What the record establishes

Carry the evidence. Keep its boundaries.

Mappings explain relevance

A finding's framework control explains its security context. It does not prove that an organization has implemented every requirement of that framework.

Inspect the rule taxonomy →

Reports retain assessment scope

Server pages link to framework reports in HTML, JSON, and PDF. Review the findings alongside their assessment method and coverage; failed or unobserved checks remain unresolved.

Browse server reports →

Signatures protect the record

A verified signature checks the integrity and authenticity of the report under its signing scheme. It does not make the findings infallible or the assessed server universally safe.

Read the verification guide →

02 / Trace the mapping

A control points back
to an actual finding.

Start at the finding, inspect its evidence, and then read the control mapping. Keep the assessment method and coverage beside the framework context.

A report can preserve that record. Its integrity and the adequacy of its evidence are separate questions.

Mapped controls support review. They do not assess every framework requirement or certify compliance.

Examine how evidence is recorded

Evidence before trust

Start with the report. Ask what was actually checked.

Inspect a server's findings, recorded coverage, and available framework reports. Use the evidence to support your review, with the unresolved questions still in view.

Explore the registry