> I'm not sure I understand this argument.

Your analogy is flawed because you are saying the information was free to begin with.

That analogy works for IBM granite models because the information they trained on is free to use.

The major LLMs did not do that.