AIBOM drift is the change in an AI system’s bill of materials over time — a new model, a bumped version, a swapped dataset, a changed license. Catching drift is how you notice a non-commercial or un-provenanced component entering a product before it becomes an audit finding.

How does it compare the two BOMs?

It indexes components by name (and version) in each document and reports what was added, removed and version-changed, plus any newly-added component that lacks a license. It works on standard CycloneDX JSON, including model and dataset components.