Coming up with the words and concepts to describe the models is a challenge.

Does the training data require permission from the copyright holder to use? Are the weights really open source or more like compiled assembly?