[jasomill]

Facts aren't subject to copyright.

As for what constitutes a derivative work, this is a matter of law. In the US,

A "derivative work" is a work based upon one or more preexisting works, such as a translation, musical arrangement, dramatization, fictionalization, motion picture version, sound recording, art reproduction, abridgment, condensation, or any other form in which a work may be recast, transformed, or adapted. A work consisting of editorial revisions, annotations, elaborations, or other modifications which, as a whole, represent an original work of authorship, is a "derivative work".

(17 U.S.C. § 101)

[Lerc]

Good point about facts. It applies similarly to properties that provide functionality. One tends to lose track of the fact that much of software shouldn't be copyrightable in the first place, it's just a pretence that has evolved due to how much people like money.

It's a stretch to say that training a model falls under that definition of derivative work. It's be like saying that building a house after reading a book on how to build a house makes the house a derivative work. I can just imagine cookbooks introducing limited licences on who you can feed with their recipes.