Solution 01
Gate ingestion on machine-readable source provenance
01 — Root cause addressed
This addresses bulk shadow-library material entering a retained central corpus without a lawful justification for acquisition and possession.
02 — Implementation
Require every bulk dataset to carry machine-readable provenance before corpus admission: source identity, acquisition date and method, license or other lawful basis, and the scope of permitted use. A validation stage should reject incomplete records and quarantine unverified material instead of admitting it to the central library. A linked ledger should make the affected works auditable and support removal by source.
03 — Prevention
Block ingestion when provenance fields or lawful-basis evidence are absent, and regularly reconcile stored corpus objects against the provenance ledger.
04 — Trade-offs
Verification slows acquisition, narrows opportunistic corpus growth, and requires tooling and review capacity. It is the wrong choice for a narrow corpus containing only clearly licensed in-house data when the same controls would duplicate an already authoritative inventory without reducing exposure.