Particle.news

Download on the App Store

Anthropic Settles Authors’ Class Action Over AI Training Data

The mediated deal averts a public trial following court findings that long‑term retention of pirated texts can infringe.

Anthropic logo is seen in this illustration taken May 20, 2024. REUTERS/Dado Ruvic/Illustration/File Photo
Image

Overview

  • Lawyers for both sides notified a federal court on Tuesday that a settlement was reached earlier this month through mediation.
  • Terms were not disclosed, with plaintiffs’ counsel saying details will be announced in the coming weeks and that the agreement will benefit class members.
  • In June, Judge William Alsup ruled that training on lawfully purchased books can qualify as fair use but allowed claims over a library of roughly seven million allegedly pirated copies to proceed.
  • The settlement removes the prospect of a jury trial that could have exposed Anthropic to substantial statutory damages of up to $150,000 per work for willful infringement.
  • Separate rulings, including Judge Vince Chhabria’s description of Meta’s training as “highly transformative,” underscore a patchwork legal landscape for AI developers.