A U.S. court ruling tied to a $1.5 billion settlement established that training AI models on copyrighted material may qualify as fair use, but storing or reproducing pirated copies of that material does not. The distinction creates a consequential legal boundary: AI developers who can demonstrate their pipelines do not retain unauthorized copies are in a materially different legal posture than those who cannot. For financial institutions assessing AI vendors or AI-exposed companies, this ruling signals that IP liability risk is now a concrete, quantifiable cost line, not merely a theoretical exposure, and that vendors unable to document their training-data provenance carry incremental legal and reputational risk.
226 other tracked developments touching the same claims.