In AI training data, chain of custody is the documented, verifiable trail of a dataset’s handling - every hand it passed through, every transformation, every access - from source ingestion to model-ready export. It is the artifact that lets you prove, rather than assert, that training data was governed. LabelFort exports it with every dataset.
Chain of custody comes from evidence law: an exhibit is admissible only if every transfer and handler is documented. Training data now faces the same test - regulators, auditors and courts increasingly ask AI operators to show the provenance of the data behind a decision. A dataset without custody records is an assertion; with them, it is evidence.
Ingestion (source, license/consent basis, hash); classification and residency assignment; access grants and every exercise of them; annotation and review actions with actor identity; transformations and exports with manifests; and destruction or retention events. Break one link - an undocumented contractor, an untracked export - and the chain’s value collapses.
Anonymous, rotating crowd workforces cannot produce custody records: you cannot name the handlers, evidence their agreements, or reconstruct their access. That is the structural reason crowd-labeled data fails regulated diligence regardless of its measured accuracy - and why LabelFort uses named, trained teams under role-based access.
Agreement between two annotators, corrected for the agreement expected by chance. κ = 1 is perfect; κ = 0 is no better than chance. Regulated annotation QA contracts above 0.90.
Read More
How consistently independent annotators label the same data. High IAA evidences that labels reflect a shared, documented standard rather than individual opinion.
Read More
Reliability for any number of annotators, tolerant of missing data, across nominal, ordinal, interval and ratio labels - the general case where kappa handles exactly two.
Read More
The documented trail of a training dataset’s handling - sources, handlers, transformations, accesses, exports - verifiable end-to-end. It underpins EU AI Act technical documentation and sectoral audit requirements.
Hash-verified ingestion, named actors under role-based access, immutable action logs, versioned guidelines, and per-dataset Evidence Exports that package the full custody record for your auditors.
IAA scored per cohort, audit trails on every action, evidence exports mapped to EU AI Act Articles 10 & 12. See it on your own data in an evidence-grade PoC.
Certifications & readiness
