VERITRAIL Control Library — 64 AI governance controls

The VERITRAIL Control Library v1.0.0 contains 64 controls across five pillars. Each control has an objective, an accountable owner role, examples of the evidence that demonstrates it, and indicative mappings to NIST AI RMF, ISO/IEC 42001 and the EU AI Act.

AO — Accountability & Operating Model

Who owns AI risk, who decides, and how governance is resourced.

Control	Objective	Priority	Applies to
AO-01 AI governance policy	Maintain a documented, approved and versioned AI governance policy that sets scope, principles, risk appetite linkage and accountabilities.	critical	All AI systems
AO-02 Executive accountability and AI governance forum	Establish a cross-functional AI governance forum with documented authority to approve, restrict or stop AI use cases.	critical	All AI systems
AO-03 AI roles, responsibilities and RACI	Define roles and responsibilities for business, technology, risk, legal, privacy and security across the AI lifecycle.	high	All AI systems
AO-04 AI risk appetite and tolerance	Define AI risk appetite, tolerance thresholds and risk-tier criteria approved by leadership.	high	All AI systems
AO-05 AI literacy and role-based training	Provide role-based AI literacy and responsible-use training for staff, operators and decision-makers.	high	All AI systems
AO-06 Escalation paths and decision rights	Define how AI issues, concerns and incidents are escalated, to whom, and within what timeframe.	high	All AI systems
AO-07 Regulatory and contractual obligations register	Maintain a register of AI-relevant legal, regulatory, contractual and customer obligations by jurisdiction and system.	high	All AI systems
AO-08 Responsible AI principles and objectives	Set measurable responsible-AI objectives aligned to organizational values and risk appetite.	medium	All AI systems
AO-09 Third-party AI accountability	Assign accountability for vendor and third-party AI, including allocation of responsibilities between provider, deployer and customer.	medium	Third-party / vendor AI
AO-10 Prohibited and restricted AI use policy	Define prohibited, restricted and conditionally permitted AI uses and screen every use case against them.	critical	All AI systems
AO-11 Stakeholder engagement and feedback	Collect and act on feedback from affected individuals, users and internal stakeholders about AI systems.	medium	High-risk systems
AO-12 Governance resourcing	Allocate budget, people and tooling sufficient to operate AI governance at the organization's scale.	medium	All AI systems

EI — Enterprise AI Inventory

A complete, classified, owned record of every AI system, model, agent and vendor AI product.

Control	Objective	Priority	Applies to
EI-01 Enterprise AI inventory (system of record)	Maintain a complete, current inventory of AI systems, models, agents and vendor AI products in a single system of record.	critical	All AI systems
EI-02 AI risk classification and tiering	Classify every AI system by business criticality, data sensitivity, decision impact, autonomy and regulatory relevance and assign a risk tier.	critical	All AI systems
EI-03 Accountable owner per AI system	Assign a named accountable business owner and technical owner to every AI system.	critical	All AI systems
EI-04 Intended purpose and context of use	Document intended purpose, users, affected parties, context of use and foreseeable misuse for each AI system.	high	All AI systems
EI-05 Model, prompt and component registry	Record models, versions, prompts, system instructions, fine-tunes and configuration for each AI system.	high	All AI systems
EI-06 Data source and lineage registry	Record data sources, classifications, provenance and data flows used to train, ground or operate each AI system.	high	All AI systems
EI-07 Agent and tool registry	Register every AI agent, the tools, APIs, plugins and data stores it can access, and its permitted actions.	critical	Agents and agentic workflows
EI-08 Third-party and vendor AI inventory	Inventory vendor AI products, embedded AI features and foundation-model providers with contract and data-processing details.	high	Third-party / vendor AI
EI-09 Shadow AI discovery	Periodically discover unregistered AI use (SaaS AI features, API keys, browser tools) and bring it into the inventory.	medium	All AI systems
EI-10 Dependency and integration mapping	Map upstream and downstream dependencies and integrations for each material AI system.	medium	High-risk systems
EI-11 Inventory attestation and recertification	Require owners to periodically attest that inventory records are complete and accurate.	medium	All AI systems
EI-12 Use-case intake register	Capture every proposed AI use case through a standard intake before build or procurement.	high	All AI systems

GL — Governance & Lifecycle Control

Risk-proportionate gates from intake through approval, change, monitoring and retirement.

Control	Objective	Priority	Applies to
GL-01 Intake triage workflow	Triage each intake against risk triggers to route it to the right assessment depth and approvers.	high	All AI systems
GL-02 Pre-deployment risk and impact assessment	Complete a documented risk and impact assessment, proportionate to risk tier, before production deployment.	critical	All AI systems
GL-03 Risk-tiered approval gates	Require approvals by authorities proportionate to risk tier before deployment, with recorded conditions.	critical	All AI systems
GL-04 Change management and reassessment triggers	Subject model, prompt, data, vendor, tool and agent changes to change control and trigger reassessment when material.	high	All AI systems
GL-05 Pre-release validation and testing	Validate performance, robustness and safety against acceptance criteria before release.	high	All AI systems
GL-06 Technical documentation	Maintain technical documentation describing design, data, performance, limitations and controls for each material AI system.	high	High-risk systems
GL-07 Deployment conditions and go-live readiness	Verify that approval conditions, monitoring, oversight and rollback are in place before go-live.	high	All AI systems
GL-08 Post-deployment monitoring	Monitor production AI systems for performance, drift, misuse, incidents and unexpected behaviour with defined thresholds.	high	All AI systems
GL-09 Retirement and decommissioning	Define how AI systems are restricted, remediated, retired or decommissioned safely, including data and model disposal.	medium	All AI systems
GL-10 Exceptions and risk acceptance	Record exceptions and residual-risk acceptances with an accountable approver, expiry date and compensating controls.	high	All AI systems
GL-11 Vendor AI due diligence and contract controls	Perform AI-specific due diligence on vendors and include AI terms (data use, model changes, incident notice, audit rights) in contracts.	high	Third-party / vendor AI
GL-12 Periodic review and recertification	Re-review AI systems at a cadence set by risk tier and recertify continued use.	medium	All AI systems

IS — Integrity, Safety & Security

Data, fairness, safety, security and runtime guardrails for AI systems and agents.

Control	Objective	Priority	Applies to
IS-01 AI threat modelling	Threat-model AI systems, including prompt injection, data poisoning, model extraction and unsafe tool use.	high	All AI systems
IS-02 Data governance and quality	Govern the quality, representativeness and fitness of data used to train, fine-tune or ground AI systems.	high	All AI systems
IS-03 Privacy assessment and data minimisation	Assess privacy, consent, retention, cross-border transfer and minimisation requirements for AI processing of personal data.	critical	All AI systems
IS-04 Sensitive data redaction and masking	Redact, mask or exclude sensitive data from prompts, responses, logs and external model calls.	critical	Generative AI systems
IS-05 Bias and fairness testing	Test AI systems for bias and discriminatory outcomes across relevant groups before release and periodically.	high	High-risk systems
IS-06 Safety and harm evaluation	Evaluate AI systems for safety risks and unintended harms to individuals, groups and society.	high	High-risk systems
IS-07 Accuracy and robustness testing	Measure accuracy and robustness, including degraded and adversarial conditions, against declared levels.	high	High-risk systems
IS-08 Transparency and AI disclosures	Inform users when they interact with AI or when AI materially influences a decision, and provide explanations where appropriate.	high	All AI systems
IS-09 Human oversight design	Design human review, override, escalation and appeal mechanisms proportionate to decision impact.	critical	High-risk systems
IS-10 Secure model and prompt supply chain	Secure models, weights, prompts, plugins and dependencies against tampering and untrusted sources.	high	Generative AI systems
IS-11 Least-privilege tool access	Grant AI agents only the tools, scopes and data access required for their declared purpose; deny by default.	critical	Agents and agentic workflows
IS-12 Runtime guardrails and policy enforcement	Evaluate prompts, responses and tool calls against versioned runtime policy before execution.	high	Generative AI systems
IS-13 Human approval for high-impact actions	Require designated human approval before AI agents execute high-impact, irreversible or external actions.	critical	Agents and agentic workflows
IS-14 Prompt-injection and data-leakage testing	Test LLM applications and agents for prompt injection, jailbreak, data leakage and unsafe tool use before release and after material change.	critical	Generative AI systems
IS-15 Kill switch and rollback	Be able to disable or roll back an AI system, model, agent, tool or tenant quickly after a security or safety incident.	critical	All AI systems
IS-16 Rate, budget and resource limits	Enforce rate, token, cost and tool-call limits to contain runaway agents, abuse and denial-of-wallet.	medium	Generative AI systems

SE — Stewardship, Evidence & Assurance

Logging, evidence, control testing, incident handling and continuous improvement.

Control	Objective	Priority	Applies to
SE-01 Control testing programme	Test AI controls periodically for design and operating effectiveness and retain results.	high	All AI systems
SE-02 Evidence repository and retention	Store governance evidence in a controlled repository with retention aligned to obligations.	high	All AI systems
SE-03 Governance metrics and leadership reporting	Report AI governance KPIs to leadership and use them to drive corrective action.	medium	All AI systems
SE-04 AI incident management and reporting	Detect, triage, contain, report and learn from AI incidents, including regulatory notification where required.	high	All AI systems
SE-05 Audit-ready evidence packs	Produce an audit-ready evidence pack for any material AI system on request.	high	High-risk systems
SE-06 Corrective action and continuous improvement	Track nonconformities and corrective actions to closure and feed lessons into policies and controls.	medium	All AI systems
SE-07 Management review	Conduct periodic management review of the AI governance system's suitability and effectiveness.	medium	All AI systems
SE-08 Immutable action logging	Log AI decisions, tool calls, policy decisions and approvals in tamper-evident, retained logs.	critical	All AI systems
SE-09 Post-market monitoring plan	Maintain a post-market / post-deployment monitoring plan for high-risk AI systems.	high	High-risk systems
SE-10 Independent review and internal audit	Subject AI governance to independent review or internal audit on a risk-based cycle.	medium	All AI systems
SE-11 Stakeholder transparency reporting	Provide deployers, customers and affected parties with appropriate information on AI system capabilities, limits and governance.	medium	High-risk systems
SE-12 Model performance and drift monitoring	Monitor model performance and data/concept drift against thresholds and trigger review when breached.	high	All AI systems

Last reviewed: 21 October 2026