Model inventory
This module is your system of record for every AI or LLM your org uses. You can register models, track approvals, record security assessment status, and see what’s allowed, restricted, pending, or blocked.
Why it matters
Model sprawl happens fast. This page keeps a single source of truth so audits are easy, procurement is clear, and developers know what they can use.
Where to find it
Open the left sidebar, click Model Inventory.
What you’ll see
Four counters at the top, Approved, Restricted, Pending, Blocked.
A status filter, All statuses by default.
A table with columns, Provider, Model, Version, Approver, Capabilities, Security assessment, Status, Status date, Actions.
Add a new model
Click Add new model.
Fill the form, then click Save.
Fields
Provider, For example, OpenAI, Google, Anthropic.
Model, For example, GPT-4, Gemini, Claude.
Version, Optional. For example, 4.0, 1.5, 5.3.
Approver, The person accountable for this listing.
Status, Pick Pending, Approved, Restricted, or Blocked.
Status date, When this status took effect.
Capabilities, Select tags like Multimodal, Tools, Code, Video, Summarization, Question answering.
Security assessment, Toggle on when the security review is complete.
Edit or update a model
Use the gear in Actions.
Change status, update version, add capabilities, or flip the security toggle.
Click Save. The Status date helps you keep a clean audit trail.
Status definitions
Pending, Under review, not yet approved for production.
Approved, Cleared for use under your policies.
Restricted, Allowed in limited contexts, for example only non-PII data or sandbox projects.
Blocked, Not permitted, with reasons captured in your internal notes or ticket.
Security assessment
Turn this on when your security team has completed their checklist, for example data handling, storage location, logging, vendor terms, DPIA, and threat model. The table shows Yes or No so auditors can scan fast.
Capabilities tagging
Use short, consistent tags so teams can find the right tool quickly.
Examples, Code, Multimodal, Tools, Video, Summarization, Question answering.
Keep tags singular and reuse them across all entries.
Common workflows
Onboard a new model
Add the entry as Pending.
Leave Security assessment off until review is done.
When security signs off, switch to Approved and set Status date to today.
Restrict a model
Open the row, change Status to Restricted.
Add the allowed use in your ticket or internal notes, for example “Only for public, non-confidential content.”
Consider adding a Capability tag that reflects limits, for example No PII.
Block a model
Change Status to Blocked and set Status date.
Notify owners in Policy Manager or Slack so developers don’t use it accidentally.
Version refresh
Update Version when you migrate.
If the change affects risk, re-run security and update the toggle and status date.
Filtering and paging
Use the All statuses dropdown to narrow the list.
Adjust Rows per page and use arrows to navigate.
Governance tips
Map your Approved and Restricted models to policies in Policy Manager.
Link training about “Approved model list” in Training Registry.
Capture vendor contracts and DPIAs in Evidences so everything is traceable.
Reporting
Use the Reporting module to export an audit-ready list of models with status and assessment flags.
Troubleshooting
Can’t save the form, Check required fields, Provider, Model, Approver, Status, Status date.
A row looks outdated, Open the action menu, update status or version, then save.
Capabilities are missing, Add or standardize tags during the next edit so search works better.
Last updated