IP law and machine learning, who owns the model?

Does anyone know how IP law works generally with respect to models produced using so called private data?

Specifically, I'm referring to the case where someone uses private data to derive a model.

Can the 'owner' of the data claim ownership of the model, even though the model may represent a domain far more generic than the privately owned data?

(Let me know if this should be posted elsewhere... just thought you folks might have a bead on what the current view of this is.)