[Improvement] Suggestion for Section 3.6 (Transparency obligations) of the AI SBOM Guide

[AttorneyTao]

https://github.com/OpenChain-Project/Reference-Material/blob/master/AI-SBOM-Compliance/en/Artificial-Intelligence-System-Bill-of-Materials-Compliance-Management-Guide.md#36-transparency-obligations

Describe the improvement

In Section 3.6 (Transparency obligations), dataset lineage should not be the only consideration. Auxiliary and base models used during development—such as foundation models, reward models for reinforcement learning, and filtering/pre‑processing models—also materially influence the final model’s behavior and may introduce additional transparency or disclosure obligations.

Expected behavior

A modified version for reference:
3.6 Transparency obligations
[The Original Part]
The process shall also consider dataset lineage together with the provenance, licensing, and characteristics of any base or auxiliary models used during development or operation (e.g., foundation models, reward models for reinforcement learning, filtering or preprocessing models). These components can materially influence the behavior of the final model and may themselves introduce transparency or disclosure obligations.