The Anthropic Ruling: A Roadmap for AI’s Copyright Future

Source: Nate’s Newsletter Date: 2025-06-24 URL: https://natesnewsletter.substack.com/p/the-anthropic-ruling-a-roadmap-for

Summary

Judge William Alsup ruled that training AI models on copyrighted books constitutes fair use — “quintessentially transformative,” analogous to human learning — but acquiring those books through piracy does not. The ruling distinguishes method of acquisition from use of training data: Anthropic’s use of Books3 and Library Genesis pirated content was condemned, though the company’s subsequent shift to legitimate purchases may affect damages. The decision establishes that financial capability removes any excuse for piracy, creating a compliance framework for the industry.

Implications

Capital thread. The ruling creates a new cost structure for frontier model training: legitimate book licensing is now legally required, not optional. For well-capitalized labs this is an operating expense; for smaller labs it’s a meaningful barrier. The publishing industry gains a real revenue lever from AI training data, which may accelerate licensing deal activity.

Enterprise adoption thread. The fair-use-for-training precedent significantly reduces one category of legal risk for enterprise AI deployments — models trained on legitimately acquired data are on firmer legal ground. But the piracy-acquisition distinction means due diligence on training data provenance becomes a vendor evaluation criterion for legally conservative buyers.

Watch: How this ruling propagates to pending cases against OpenAI, Meta, and Stability AI, and whether the piracy-acquisition distinction holds in jurisdictions outside the US — the ruling’s logic is compelling but its scope is currently limited to one court.