> ## Documentation Index
> Fetch the complete documentation index at: https://docs.guardway.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# ISO/IEC 42001 Mapping
> How Guardway capabilities support an ISO/IEC 42001 AI management system — clause by clause and control by control, with the limits stated.
**What certification means here.** ISO/IEC 42001 certifies *your
organization's AI management system (AIMS)* — its policies, processes and
evidence — not any vendor's product. **Guardway does not hold an ISO/IEC
42001 certification and a tool cannot confer one.** What Guardway provides
is operational evidence and enforcement your AIMS can point to. Where a
clause needs something a gateway cannot supply, this page says so.
## Clause mapping (4–10)
| Clause | What it asks | How Guardway supports it | What it does not cover |
| ------------------------------ | ---------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------- |
| **4 — Context** | Understand your AI footprint and stakeholders | [Agent discovery](/platform/agents) inventories the agents, models, tools and MCP servers actually running in your cloud accounts; the [gateway fleet view](/platform/topology) shows where traffic flows | Organizational context, scoping decisions |
| **5 — Leadership** | Policy, roles, accountability | Role-based access with a protected owner role; ownerless-identity findings flag AI agents nobody is accountable for | Writing your AI policy |
| **6 — Planning** | Risk assessment and treatment | Red-team scans with rate-based severity, OWASP ASI findings per discovered agent, Endpoint AI Risk reports; treatment = guardrail policies in monitor or block mode | Risk acceptance decisions and criteria |
| **7 — Support** | Resources, competence, documented information | This documentation set; per-request logs, audit logs and reports as retained documented information | Training your people |
| **8 — Operation** | Operational planning and control of AI systems | The gateway *is* the operational control point: [policies](/platform/configuration/security) enforce which models, guardrails, budgets and rate limits apply per application (scoped per gateway or org-wide); budgets, quotas and rate limits are additionally enforced per key and team through key and team settings | Processes outside the traffic path |
| **9 — Performance evaluation** | Monitoring, measurement, analysis | [Dashboards](/platform/dashboard/overview) for usage, spend and security; [SIEM export](/platform/settings/integrations) into your monitoring estate; alert rules on the signals you choose | Internal audit and management review |
| **10 — Improvement** | Nonconformity and corrective action | Guardrail events, drift incidents (beta) and red-team regressions give the *inputs* to corrective action, with severity and evidence attached | The corrective-action process itself |
## Annex A controls where Guardway is direct evidence
| Control theme | Guardway evidence |
| -------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **AI system impact assessment** (A.5) | Per-agent OWASP ASI findings with severity, evidence and remediation; endpoint risk reports; red-team results with bypass rates |
| **AI system lifecycle — operation and monitoring** (A.6) | Request-level logging with identifiers, token counts, cost and outcome; MCP server health checks; provider failover; alerting |
| **Data for AI systems** (A.7) | PII detection and masking at the gateway; guardrail event previews masked by the built-in content filter before they are written (model-detected entities outside its patterns may still appear); the [data boundary](/resources/privacy-and-security#the-data-boundary) — prompts and completions stay on your gateway by default (gateway v0.4.41+); local storage under your own retention management |
| **Information for interested parties** (A.8) | Concealed-AI-identity detection: agents instructed to hide their AI nature or impersonate humans are flagged — direct evidence for transparency commitments |
| **Responsible use of AI systems** (A.9) | Enforced guardrail policies (PII, injection, toxicity, keyword, IP), budgets and quotas per key/user/team/org, immediate revocation |
| **Third-party and supplier relationships** (A.10) | Provider and model inventory under central control; MCP server registry with health status; [skill/MCP repository scanning](/discovery/repositories) for supply-chain findings before adoption |
## What to hand your auditor
The practical artifact set Guardway can produce for an AIMS audit:
1. **Inventory** — agents, models, tools, MCP servers, gateways (discovery + fleet views).
2. **Control configuration** — the policy set in force per application, viewable in the dashboard configuration screens.
3. **Operating evidence** — request logs, guardrail events with redacted previews, administrative audit log, SIEM-delivered copies in your own retention system.
4. **Testing evidence** — red-team run results with severities computed from bypass rates.
5. **Incident inputs** — drift incidents (beta), guardrail blocks, unreachable-dependency findings.
## Machine-readable mappings
Agent findings are tagged with OWASP ASI Top 10 categories (`ASI01`–`ASI10`)
in the API and dashboard; the explicit statement of which categories were
assessed, and with what limitations, is returned by the discovery summary API
(dashboard surfacing planned). ISO 42001 clause identifiers as
machine-readable tags are **not shipped today**; this page is the mapping.
See also the [EU AI Act mapping](/resources/compliance/eu-ai-act) and
[Privacy and Security](/resources/privacy-and-security). For your current
attestation package, email
[compliance@guardway.ai](mailto:compliance@guardway.ai).