compliance · ISO 42001
Wytness and ISO 42001
Annex A evidence for the AI agents inside your AI management system.
ISO/IEC 42001:2023 audits your AI Management System: the policies, roles, impact assessments and life-cycle controls around the AI you build and use. Where that system includes AI agents, Wytness evidences the Annex A controls that rest on a record of what the agents did — the event log, operation monitoring, deployment approval, intended use, data provenance and supplier usage. The AIMS itself — AI policy, impact assessments, risk treatment and the Statement of Applicability — stays yours. The other 28 Annex A controls are listed below with the reason the pack does not evidence them.
the artefact
What's in the pack
An ISO 42001 pack contains the agent and tool inventories, monitoring evidence per control, signed event samples, the LLM provider and model mix, and a control-by-control narrative aligned to your Statement of Applicability, and the list of Annex A controls it does not evidence, with the reason for each.
Every claim in the pack is traceable: each control cites the events behind it, and each event carries the signature your own key produced. The pack is sealed, so an auditor can confirm it reached them unaltered.
A real sealed pack is published — nothing here is a mock-up. Download the specimen and check it yourself on the verify page, in your browser, without uploading it. It is a SOC 2 (Trust Services Criteria) pack — the seal and the per-event signatures work identically whichever framework you generate.
// What the pack asserts about itself:
// control_mapping[] — 10 criteria, each carrying
// its evidence and a metric measured over your period:
…// + 6 more, in the table below
// Every control cites events an auditor can re-verify:
the mapping
What we map to what
| Control | Name | What Wytness provides |
|---|---|---|
| A.3.2 | AI roles and responsibilities | Each agent in the Agent Registry carries an assigned operator (a named user of the organisation) and an approval record (approved_by, approved_at); every event records human_operator_id; four-tier role-based access control (viewer / member / admin / owner) is enforced server-side on every route; each register change is appended to the platform audit log with actor, target, and the fields changed. |
| A.4.4 | Tooling resources | Every tool an agent invokes is inventoried from the event stream into the Registry's tool list, the pack reports events by tool, and the first use of a tool by an agent raises a new_tool anomaly. Tools are recorded as used; Wytness does not assess them. |
| A.6.2.4 | AI system verification and validation | Wytness evidences verification of the record, not validation of the model: the hash chain is verified link by link per agent (live and on demand via /verify), per-event Ed25519 signatures are checked against the customer's registered public keys, and a three-way ingest / query-store / archive reconciliation runs on a schedule; the pack states each control's pass / fail / no-data status for the period. Model accuracy, robustness and bias testing remain the customer's and are not performed by Wytness. |
| A.6.2.5 | AI system deployment | The Agent Registry records an approval state for every agent (discovered → intake → development → validation → approval → deployment → retirement) with approved_by and approved_at; agents seen in the event stream are added to the register as auto_discovered so none runs unrecorded, and each approval change is logged separately as agent_approval_changed with its from / to states. Wytness records the approval; it does not block an unapproved agent. |
| A.6.2.6 | AI system operation and monitoring | 7 anomaly-detection rule families run against the live event stream (first-time tool use, capability drift, frequency spikes, error spikes, restricted-data access, off-hours activity, hash-chain breaks) with automated severity-classified alerting; the dashboard surfaces event rate, error rate, anomaly count, and chain-integrity status with an explicit as_of freshness signal, with per-agent and per-session drilldown. |
| A.6.2.8 | AI system recording of event logs | Every tool call, resource access, and inter-agent message is logged and hash-chained (tamper-evident); each event carries a per-event Ed25519 signature; chain breaks surface as anomalies; a separate platform audit log records administrator actions. Events are retained for seven years by a storage lifecycle policy on Wytness-hosted storage, or under the customer's own policy on bring-your-own-storage; alteration is made evident rather than prevented. |
| A.7.5 | Data provenance | Every event records inputs_source and inputs_classification — the provenance and sensitivity class of the data the agent acted on, not the raw payload — and the Agent Registry tags each agent with the data classes it accesses (data_classes_accessed); an event classified restricted raises a restricted_data anomaly. Training and development data sets are outside Wytness's view. |
| A.8.5 | Information for interested parties | Evidence Packs give auditors and regulators a dated account of agent activity for a chosen period: agent and tool inventories, per-control evidence with a measured metric, anomaly history, signed event samples, and the LLM provider and model mix; the printable report is paired with a sealed JSON download an auditor verifies offline without a Wytness account, and the signed agent-register PDF covers the inventory deliverable. |
| A.9.4 | Intended use of the AI system | The Agent Registry documents each agent's intended use and description beside its risk classification (minimal / limited / high / prohibited, with a justification required for high and prohibited), its EU AI Act category, and its approval state; every event carries agent_id, agent_version and agent_instance_id, so the customer can set recorded use against documented intended use. |
| A.10.3 | Suppliers | Wytness records which third-party model providers and models each agent called: the pack's LLM-usage section reports calls, tokens and cost per model and the per-provider call mix for the period. Supplier due diligence, contracts and alignment reviews remain the customer's. |
the boundary
What Wytness does not cover
| Control | Name | Why the pack does not evidence it |
|---|---|---|
| A.2.2 | AI policy | An organisational policy document; Wytness records agent activity, it does not author or hold the customer's AI policy. Shield policies are technical rules in observe mode, not the AI policy. |
| A.2.3 | Alignment with other organisational policies | Policy-to-policy alignment is a governance review Wytness has no view of. |
| A.2.4 | Review of the AI policy | Same as A.2.2. |
| A.3.3 | Reporting of concerns | A whistleblowing / concern-reporting channel; Wytness has none. |
| A.4.2 | Resource documentation | The consolidated resource inventory (data, tooling, compute, people) is the customer's document; Wytness evidences only the tooling slice below. |
| A.4.3 | Data resources | Training and development data sets never pass through Wytness. Operational data classes are evidenced under A.7.5. |
| A.4.5 | System and computing resources | Wytness does not see the customer's compute estate. |
| A.4.6 | Human resources | Competence and HR records are outside the product. |
| A.5.2 | AI system impact assessment process | Impact assessments are recorded per asset in the Registry; this pack does not attach them as evidence. |
| A.5.3 | Documentation of AI system impact assessments | Same as A.5.2. |
| A.5.4 | Assessing AI system impact on individuals or groups | Same as A.5.2. |
| A.5.5 | Assessing societal impacts | Same as A.5.2. |
| A.6.1.2 | Objectives for responsible development | Organisational objectives; no Wytness mechanism. |
| A.6.1.3 | Processes for responsible AI system design and development | Design-process evidence lives in the customer's engineering practice. |
| A.6.2.2 | AI system requirements and specification | Requirements documents are the customer's. |
| A.6.2.3 | Documentation of AI system design and development | Design documentation is the customer's; the register's `description` and `intended_use` are evidenced under A.9.4. |
| A.6.2.7 | AI system technical documentation | Technical documentation of the AI system itself is the customer's; the Evidence Pack is reporting (A.8.5), not the system's documentation. |
| A.7.2 | Data for development and enhancement of AI system | Development data never passes through Wytness. |
| A.7.3 | Acquisition of data | Same as A.7.2. |
| A.7.4 | Quality of data for AI systems | Wytness does not assess data quality. |
| A.7.6 | Data preparation | Preparation pipelines are upstream of the recorder. |
| A.8.2 | System documentation and information for users | End-user documentation of the AI system is the customer's. |
| A.8.3 | External reporting | A channel for interested parties to report adverse impacts; Wytness has none. |
| A.8.4 | Communication of incidents | Connectors push anomalies to the customer's own SIEM / ITSM / chat; communication to the AI system's users is the customer's process. |
| A.9.2 | Processes for responsible use of AI systems | An organisational process; no Wytness mechanism. |
| A.9.3 | Objectives for responsible use | Same as A.9.2. |
| A.10.2 | Allocating responsibilities | Contractual allocation between the customer and its suppliers is outside the product. |
| A.10.4 | Customers | The customer's obligations to its own customers are outside the product. |
- —Certification is your auditor's decision. Wytness is not a certification body, and an Evidence Pack is a control mapping with evidence, not a certificate.
- —The AI policy (A.2), impact assessments (A.5) and the objectives and processes for responsible development and use are management-system deliverables Wytness does not produce. We record what agents did; what that means for your AIMS is your assessment.
- —Model performance, bias testing, and the quality of training and development data (A.7.2 to A.7.4) are outside Wytness's view. We log operational inputs and outputs; we do not evaluate the model.