We track every case. We don't editorialize on active disputes — we explain what courts actually decided, verbatim, and why it matters.
Why Bartz set the floor — not the ceiling — for damages in pirated-book training cases, and what Claude-era defendants should expect from the next wave of filings.
GEMA v. OpenAI rejected the TDM exception as applied to model memorization. That single sentence reshapes European AI training law for a generation.
The U.S. fair-use scoreboard after Bartz, Kadrey, Thomson Reuters v. Ross, and NYT v. OpenAI — with the holdings quoted verbatim.
A walkthrough of the complaint, the motion-to-dismiss ruling, the 20M-log discovery order, and where summary judgment will land.
Every publicly disclosed AI settlement amount as of Q1 2026, with links to the approval orders.
The UK High Court rejected the "model weights as infringing copies" theory — but left territorial training questions open.
Retrieval-augmented generation moves the fair-use debate from training to outputs. Here's what Dow Jones v. Perplexity tells us.