Anthropic's $1.5bn author settlement finalised; OpenAI MDL pending

Summary

Anthropic's settlement with authors in Bartz v. Anthropic reached final approval at a 14 May 2026 hearing, America's largest copyright settlement at ~$1.5bn, about $3,000 per work. It follows Judge William Alsup's June 2025 ruling that training on copyrighted books was fair use but storing pirated copies was not. The model resolves liability for past piracy without settling the broader fair-use question. Openai faces a consolidated MDL in the Southern District of New York bundling 16 copyright suits, lead cases include the New York Times, with no final ruling as of June 2026. The disputes sit alongside the wider governance reckoning over how frontier models are built.

By the numbers

• ~$1.5bn, Bartz v. Anthropic settlement total.
• ~$3,000, estimated payout per work.
• 14 May 2026, final approval hearing.
• 16, copyright suits consolidated against OpenAI in SDNY.
• Jun 2025, Alsup's fair-use-but-not-piracy ruling.

Why it matters

The settlement prices the legal risk of training on pirated data without resolving whether training itself infringes, leaving every lab exposed until courts or Congress draw the line. A loss in the OpenAI MDL could reset the economics of model training across the industry.

What to watch

• Rulings or settlements in the OpenAI SDNY MDL.
• Whether Congress legislates a training-data fair-use standard.
• Copyright claims against image/video and Chinese open-weight models.