Find approved information and keep its provenance.

Data owners, investigators and governance leads need to establish what was received and how a search result was produced. Start with acquisition records and access boundaries before comparing search features.

08

Data Ingestion and Provenance for Intelligence - AICLarger two-stage evidence flow from NERVE to CORTEX with approved holdings, evidence references, and human review checkpoints.Illustrative evidence pathwayPLATE 01010203INDEX / MODEEVIDENCENERVECORTEXapproved holdingsevidence referenceshuman reviewDocumented evidence interfaceInterfaces and deployment scope requireconfirmation.

Inventory sources before building an importer

List approved files, messages, device packages, repositories, databases, APIs and feeds. For each, record format, volume, update frequency, legal authority and owner. A connector family does not establish support for every proprietary export or schema.

Keep originals and processing history distinct

NERVE preserves accepted originals and records versioned transformations. Extracted text, translations and other derivatives should retain a route back to their input. A later processing run must not silently change the evidence behind an earlier citation.

Test discovery under a denied-access case

Search results, counts and facets can reveal information even when a document cannot be opened. Evaluate policy before discovery with an intentionally restricted user. Inspect organisation, dataset, purpose and privacy constraints, alongside the exact evidence locator for an authorised result.

Acceptance criteria

Can a reviewer recover the accepted original? Does reprocessing preserve previous locators? Which source metadata is retained? How are retention and deletion evidenced? Where CORTEX consumes a reference, test the complete resolution path rather than assuming the interfaces match.