Abstract
A system and method for consolidating machine-learning models trained on separately governed contributions into one sector model, inside a governed data custody system. Each contributor's deposits train a separate, removable component, such as an adapter or expert module, on a shared base; the consolidated model combines the components through recorded composition, merging or routing. Every consolidation is a governed job admitted against the terms of all deposits involved, executed inside custody nodes, and recorded in an append-only ledger. The consolidated model's lineage record is composed from the lineage records of its components, so any part can be traced to its deposits. When a contributor withdraws, the components derived from that contributor's deposits are removed and the consolidated model is rebuilt from the remaining components, with an attestation that binds the withdrawal, the removed components and the hashes of the model before and after. Each contributor's share of the consolidated model's performance is measured and recorded as contribution credit. The method extends the governed custody system disclosed in SI-DP-008, "SI Data Bank™: Governed Custody of Sector Data and Models with Regulated AI Access, Data Reset and Verified Restore Points" (Technical Disclosure Commons, submitted 10 October 2026).
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Wang, Edward D. H., "SI Data Bank™: Consolidating AI Models from Separately Governed Contributions, with Removable Components, Composed Lineage and Attested Withdrawal", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/12156