Governance review

Internal AI model inventory + source-backed due diligence is the natural use case for the catalogue's transparency surface. Each verified field is paired with a primary-source citation and a deterministic freshness state; data gaps are explicit; the reverification queue lists every record due for manual re-check.

Signals to walk during a governance review:

• Verification status — a record is verified, partial, or unverified. Partial is meaningful: the model exists on the catalogue but some metric fields are null because the vendor does not publish, or because automated retrieval is blocked (the OpenAI 403 case).
• Lifecycle status — an active model is eligible for new workloads; deprecated and retired records are retained for inventory continuity but should not be used for new procurement.
• Source citations — every verified field points at a primary-source URL with a retrievedAt date. Open the citation, confirm the value still matches, and treat the record as confirmed.
• Freshness state — fresh / review_due / stale / blocked / unknown, computed deterministically against siteConfig.buildDate. Stale ≠ false; it means re-verify before reuse.
• Reverification queue — surfaces every source that has aged out, every blocked vendor doc, and every partial-coverage provider with a suggested manual action. Open at /reverification.

Important caveat. A record marked "verified" means the field is backed by a primary-source citation on the date recorded — it is not a certification, a compliance attestation, or a fitness-for-purpose claim against any specific regulatory regime. The catalogue is a source-backed starting point for due diligence, not a substitute for legal, security, or compliance review.

The shortlist below collects models with any source citations recorded (the minimum bar for a governance review). For the full source registry across every provider see /sources; for the verification coverage matrix see /coverage.