Yup, either you own it, or you don’t own it and ownership is being redefined as a quasi license.

Nobody "owns" it for the most part. Most LLM output can not be copyrighted, because there is not enough human essence in it. Meaning that it is public domain the moment it is produced.

There might be a few exceptions, if you really guided your model in a certain particular way, but that is rare.

What /could/ be is that you violate their ToS. If that ToS is valid and legally and practically enforceable is a different question and will depend a lot on your jurisdiction.

If something is generated in someone’s private paid account that might not be public domain.

It’s kind of like causing what Google Docs autocomplete sentences helps someone write is still.. theirs.

Apparently, they use "own" in the same misleading way that digital media providers use "buy", but later remove the purchased content from my library.

It's named "yours" at the time of payment, but actually only a license to use under certain conditions.

If it's under someone else's license, then you can't copyright / patent the code generated from LLM in the commercial codebase.

This basically makes their AI useless.

> Subject to your compliance with our Terms, we assign to you all of our right, title, and interest—if any—in Outputs.

Originally it was not that way. They changed it because, as you suggested, their service has questionable utility without it.

The outstanding one is the “meaningful human input” part of copyright. The most recent ruling is that prompting alone does not count. If you write/rewrite sections, those are yours. Everything in between is a somewhat untested.