You say "public", but what I think you mean is "publicly available". Even publicly available data has copyrights, and unless that copyright is "public domain", you need to follow some rules. Even licenses like Creative Commons, which would be the most permissive, come with caveats which OpenAI doesn't follow [0].
It is unclear if someone breaking someone else's copyright to use A can claim copyright on a work B, derived from A. My point is that OpenAI played loose with the copyright rules to build its various models, so the legality of their claims against DeepSeek might not be so strong.
I am not saying OpenAI did good by using publicly available data. I meant these are separate activities. None is good. But DeepSeek is slightly better by making theirs opensource.
It is unclear if someone breaking someone else's copyright to use A can claim copyright on a work B, derived from A. My point is that OpenAI played loose with the copyright rules to build its various models, so the legality of their claims against DeepSeek might not be so strong.
[0] https://creativecommons.org/share-your-work/cclicenses/