What changes in your analysis if this is an open-weight...

For open-weight releases, the requirement for full data disclosure makes it easier for plaintiffs to identify infringing sources. Furthermore, commercial use in an open context removes the educational fair-use factor. Therefore, the strategy for open releases must shift toward conservative sourcing, such as CC-licensed or public domain data. Closed commercial models, by contrast, can rely on fair-use arguments that do not need to be publicly disclosed.