Sora 2: Licensing is all you need
jperla.com
jperla.com
Isn't that fairly close to what actually happened though? "Spotify's Beta Used 'Pirate' MP3 Files, Some from Pirate Bay (2017)" https://news.ycombinator.com/item?id=43169461
One difference though is that for Spotify removing pirated files from their servers was trivial, while for LLM companies like OpenAI and Google they would have to retrain their models from scratch, which would be extremely expensive, so what they do instead is to filter problematic results that are too obvious.