Overview

• Judge William Alsup ruled on June 23 that Anthropic’s use of copyrighted works to train its Claude models constitutes transformative fair use under U.S. law.
• Judge Vince Chhabria found on June 25 that Meta’s Llama models did not dilute the market for copyrighted texts and urged creators’ compensation from AI profits.
• Both decisions recognize AI’s pattern-learning as transformative use without addressing allegations of sourcing material from shadow libraries like Books3.
• More than 21 related lawsuits are active in U.S. courts and parallel cases and legislative proposals in India continue to challenge unlicensed AI training data.
• The rulings highlight persistent uncertainty over fair-use boundaries for generative AI and intensify debates on data transparency and creator remuneration.