Since the training data is basically the “source code” via this reasoning, does that mean there are trade secret claims (source code theft) to be made against GenAI companies? The argument would be that sites like SO, Reddit, etc implicitly grant permission to read the “code”, but not to “run in production”.