The $1.5 Billion Line That Redrew AI Copyright

Before Bartz, the AI copyright argument was largely theoretical — labs argued fair use, rights holders argued infringement, and courts had not yet drawn a workable line. Alsup drew it. Training on legally acquired data survives fair use scrutiny; training on pirated material does not. That distinction, embedded in a $1.5 billion settlement approved by a federal court, is not a negotiating position or a think-piece argument. It is the framework inside which every subsequent case will be litigated.