Judge delays approval of Anthropic $1.5B author settlement

In short: A federal judge has delayed final approval of Anthropic’s $1.5 billion settlement with authors after objections over lawyer fees, payouts, and future protections.

What happened

A US federal judge, Araceli Martinez-Olguin, declined to give final approval to a $1.5 billion settlement involving AI company Anthropic and a large group of authors. The case centers on claims that Anthropic used pirated books to train AI systems.

Several authors and other members of the group complained that the proposed payouts to writers are too small compared with what the lawyers are asking for. Objectors pointed to a request of more than $320 million in attorney fees, while some authors expect to receive around $3,000 each.

One objector, author Pierce Story, argued that tying lawyer pay to the full settlement fund is unfair if many eligible authors do not file claims and end up unpaid. An attorney for the authors said claims have been filed covering over 92 percent of the more than 480,000 works included in the settlement.

Some objectors also want the deal to include limits on future use of their books. For example, one author asked the court to require Anthropic to destroy all copies of covered works, including scanned physical copies, because authors may not know how their books were obtained.

The judge ordered the authors to respond to objections by May 21. Anthropic must also explain why “late opt outs” should not be allowed.

Why it matters

This dispute affects how money is split in large group lawsuits, and whether people who say they were harmed actually see meaningful payments. It also raises a practical question about AI training data, which is what AI systems learn from, like a student studying from a pile of books, including whether companies must delete disputed materials after a case ends.

Source: Arstechnica