*or any digitized proprietary works, just as long as they can be parsed correctly. don't worry, the means of how to optain these works doesn't seem to matter[0]

[0]: https://www.arl.org/blog/training-generative-ai-models-on-co...