AI Security Audit and Evidence Checklist
Purpose: Define audit evidence required to verify AI security controls across every family of the control library, from discovery and DLP to agents, models, supply chain and incident response.
Audience: Audit teams, governance teams, control owners, security engineering.
| Control Area | Control IDs | Evidence Required | Evidence Owner |
|---|---|---|---|
| AI Discovery | AI-CTRL-001 | AI inventory, usage dashboard, user/device reports | Security Engineering |
| Prompt Inspection | AI-CTRL-002 | Prompt detection logs, policy actions, sampled test results | AI Security Team |
| File Upload Protection | AI-CTRL-003 | Upload block/warn logs, classification policy, screenshots | DLP Team |
| IDE Governance | AI-CTRL-004 | IDE assistant inventory, source code/secret detection events | DevSecOps |
| Runtime Protection | AI-CTRL-005, AI-CTRL-020, AI-CTRL-021, AI-CTRL-039 | Prompt injection test reports, runtime logs, guardrail evidence, grounding and content safety results | AppSec |
| Agent Governance | AI-CTRL-006, AI-CTRL-022, AI-CTRL-023, AI-CTRL-024, AI-CTRL-025, AI-CTRL-026, AI-CTRL-027 | Agent inventory, tool access logs, approval logs | AI Platform Team |
| Sovereignty | AI-CTRL-007 | Residency, retention, training exclusion and admin access evidence | GRC / Legal |
| SOC Integration | AI-CTRL-009 | SIEM events, parser mapping, alert correlation and runbook | SOC |
| Policy Enforcement | AI-CTRL-010 | Policy configuration, action logs by type, exception register, bypass test results | AI Security Team |
| Auditability | AI-CTRL-008 | Exportable audit records, log schema, retention policy, sample reconstruction | Security Engineering |
| Governance, Risk and Use-Case Registry | AI-CTRL-011, AI-CTRL-012, AI-CTRL-037 | Use-case registry, intake approvals, risk tiering, policy-to-control mapping, risk register, exception register | AI Governance Lead |
| Privacy and Vendor Assurance | AI-CTRL-013, AI-CTRL-014 | Impact assessments, data subject request tests, vendor assessments, contract clauses | Privacy Office / Procurement |
| Identity and Access | AI-CTRL-015, AI-CTRL-016 | Identity registry, sponsor records, permission maps, recertification and drift reports | IAM Team |
| Data and Retrieval | AI-CTRL-017, AI-CTRL-018, AI-CTRL-019 | Masking rules, tenant isolation tests, retrieval permission tests, ingestion provenance | Data Owner / Application Owner |
| Model, Training and Pipeline | AI-CTRL-028, AI-CTRL-029, AI-CTRL-030 | Model access list, dataset provenance, artifact signing, release gate results | ML Engineering / MLOps |
| Supply Chain, Infrastructure and Resilience | AI-CTRL-031, AI-CTRL-032, AI-CTRL-040 | AI-BOM, approved source list, scan reports, segmentation and secrets configuration, fail-mode and failover test results | DevSecOps / Platform Engineering |
| Security Review and Adversarial Testing | AI-CTRL-033, AI-CTRL-034 | Threat models, approvals, red team reports, regression suite results, agreed thresholds | Security Architecture / AI Red Team |
| Control Assurance | AI-CTRL-038 | Control effectiveness results, evidence packs with collection dates, owner attestations, findings log | GRC / Internal Audit |
| Fairness and Explainability | AI-CTRL-041 | Bias test results, threshold approvals, explanation records, human-review log | AI Governance Lead / Model Risk |
| Impact Assessment | AI-CTRL-042 | Completed assessments, data protection officer advice, approvals dated before go-live, review dates | Data Protection Officer / AI Governance Lead |
| Authority Engagement | AI-CTRL-043 | Authority register, filing and registration records, contact log | Legal / Compliance |
| AI Literacy | AI-CTRL-044 | Role and needs matrix, completion records, effectiveness checks | AI Governance Lead / HR |
| Incident Response and Forensics | AI-CTRL-035 | Playbooks, exercise records, preserved-state checklist, provider escalation path | SOC / Incident Response |
| Cost and Abuse | AI-CTRL-036 | Cost attribution reports, budget and quota configuration, anomaly alerts | FinOps / AI Platform Team |
Evidence Quality Criteria
Evidence should be:
- Exportable.
- Timestamped.
- Attributable to user/device/application where applicable.
- Linked to a policy or control ID.
- Retained according to approved retention requirements.
- Sufficient for incident reconstruction and audit sampling.
Exception Register Fields
Exception ID:
Control ID:
AI System / Vendor:
Reason for Exception:
Compensating Control:
Business Owner:
Risk Owner:
Approval Date:
Expiry Date:
Review Status: