Control · C-055 · Governance & Accountability

AI inventory and registration

All AI systems and models, whether built, bought or embedded in other products, are recorded with owner, purpose, tier, data, supplier and status, and the record is reconciled periodically.

Control type: Detective
Family: 02 Governance & Accountability
Risks addressed: 7
Audience: Board · Heads-of · Practitioner
Frameworks: Principle 5 · GOVERN 1 · Principle 1
Responsible AI domain: 02 Governance & Accountability
Lifecycle stage: 1 Ideate & screen
2 Assess & tier
7 Operate & monitor
8 Change or retire
Reviewed: 2026-11-10

What it does

It tells the organisation what AI it has. Every system is registered with enough detail to govern it, and the register is checked against independent sources so that it stays complete. A governance function owns the inventory; each system owner is accountable for their entries. Entries are created at intake, updated at each change and reviewed at least annually.

How to implement it

Define what must be registered, erring towards inclusion.
Record owner, approved use, tier, regulatory class, data, supplier, dependencies and review date.
Record which systems consume the outputs of which others.
Create the entry at intake, so registration is a consequence of approval.
Reconcile against procurement, supplier lists, architecture records and discovery scans.
Make owners attest to their entries on a set cycle.

Evidence it operates

Design: an inventory standard with required fields and a definition of scope. Operation: the inventory itself, with no blank owners or tiers; reconciliation results and the entries added as a result; owner attestations; internal audit testing of completeness.

Where it fits

The base on which tiering, validation, monitoring and regulatory classification rest. It is the primary control for incomplete AI inventory and for model interdependency and aggregation, and is kept complete by shadow AI discovery and AI use-case intake and approval.

Limits

An inventory is accurate on the day it is reconciled and decays from then. The definition of AI is contested, and a narrow one leaves out systems that matter. Registration that depends on teams volunteering will miss whatever they did not think counted. Being listed is not the same as being governed.

Risks this mitigates

Primary mappings

Incomplete AI inventory High: Records every system with owner, purpose, tier and status and reconciles the record. The organisation cannot list the AI systems and models it builds, buys or uses, so risk assessment, oversight and regulatory duties are applied to only part of what it has.
Model interdependency and aggregation Medium: Records which models feed which. Models feed other models, so an error or change in one propagates, and the combined uncertainty across the chain is never assessed.

Supporting mappings

AI not mapped to important business services High: Supplies the list of AI systems to map. AI components that an important business service relies on have not been identified in the resilience mapping, so their failure is not planned for.
Unassessed AI in third-party products High: Brings the feature into the inventory. A supplier switches on AI features inside a product already in use, bypassing the organisation's assessment and approval.
Unclear AI ownership and accountability Medium: Records the owner against each system. No named individual is accountable for a given AI system, or for AI risk overall, so decisions about it are made by default and nobody answers for the result.
Uncontrolled AI retirement Medium: Shows what depends on the system and updates its status. A system is withdrawn without managing dependent processes, retained data, model artefacts and continuing obligations.
Use beyond validated scope High: Records each system's approved use. A model approved for one purpose is reused for another, or for a new population, without reassessment.

Framework and clause references

Framework	Clause	Title
Code of Practice for the Cyber Security of AI	Principle 5	Identify, track and protect your assets Asset identification.
NIST AI Risk Management Framework (AI RMF 1.0)	GOVERN 1	Policies, processes and practices for AI risk are in place and effective Inventory.
SS1/23: Model Risk Management Principles for Banks	Principle 1	Model identification and model risk classification The model inventory.

Related terms

How to cite

Axiom Verity (2026). AI inventory and registration. Reference Library, C-055. https://axiom-verity.com/controls/ai-inventory. Reviewed 2026-11-10.

Found an error or a better source? Report a correction.