I think we can all agree that AI has incredible potential. But that doesn't give anyone the right to take someone else's work without asking. Creators put time, money, and heart into what they make. Using their data to train AI without permission or payment isn't innovation — it's exploitation. It's no different than a student copying a book chapter for their own profit without buying it. If we allow this, we're saying creators' rights don't matter as long as the tech is cool. That sets a dangerous precedent. Fair use has limits, and mass commercial training isn't covered by it. We need consent and compensation, period.

Did you know that over 90% of the text used to train major AI models comes from copyrighted sources without permission? That's not a fringe issue. You say training isn't the kind of copying copyright law prevents, but the law explicitly protects reproduction and derivative works. Training literally requires making permanent copies of the data inside the model's weights.

Take the New York Times lawsuit against OpenAI. The Times found that ChatGPT could reproduce their articles nearly verbatim. That's not pattern recognition — that's memorization and reproduction. Your human analogy falls apart because when I learn from a book, I don't store a mathematically compressed version that can spit out the original on command. The model's weights literally encode copyrighted text. You mention Google Books, but that was about indexing and snippets for search, not generating new content in competition with the original. AI training is fundamentally different — it's building a commercial product that can replace the need to visit the creator's site or buy their work. If I read a thousand novels and then sell my own novel, I'm not in court. But if my novel contains passages from those books, I am. That's the