Anthropic Settlement Sparks Dispute Over Claims by Publishers and Agents

Anthropic’s $1.5 billion copyright settlement is moving from courtroom approval to the much harder task of distributing money. Some authors say they were surprised to receive notices that a publisher, or another organization, had filed a claim against payments associated with their books. In several cases, the authors believe the claimant no longer owns the relevant rights or is requesting a larger share than the settlement permits.

How the settlement is supposed to work

The case rests on an important distinction. A judge previously found that training AI models on copyrighted material can qualify as fair use in certain circumstances, while obtaining that material through piracy does not receive the same protection. Once the settlement received final approval, payments could proceed for nearly 500,000 titles. Each qualifying pirated work is assigned a payment of $3,000.

The allocation rules are relatively straightforward on paper. If a book remains in print with a traditional publisher, the payment is generally split 50-50 between the author and publisher. If the book was self-published, or if the publisher allowed it to go out of print and the rights reverted to the author, the author should receive the full payment. The practical difficulty is proving when those rights changed hands. Under the settlement, an author seeking a 100% claim based on rights reversion must show that the reversion occurred before August 10, 2022, the settlement’s relevant “download date.”

Why authors are objecting

Mystery and thriller writer April Henry publicly questioned why HarperCollins had claimed one of her books even though its rights had reverted many years earlier. Victoria Strauss of the Writers Beware blog said complaints generally fall into two groups: publishers claiming books for which they no longer have a valid interest, and publishers seeking 100% of a payment when they should receive only half.

Strauss has cautioned against immediately treating every disputed claim as deliberate misconduct. Publishing companies often maintain complicated records spanning multiple imprints, contracts and rights reversions. Some publishers have reportedly acknowledged mistakes and asked Anthropic to correct them. Authors Guild CEO Mary Rasenberger likewise said she did not view the reports as proof of a coordinated attempt by publishers to take authors’ money.

At the same time, Strauss argues that the volume and consistency of the complaints are significant. If authors repeatedly encounter the same errors, the problem may be more than a handful of routine glitches in a large administrative operation. Literary agencies have also drawn criticism after some were reportedly said to be filing claims. Authors object that agents generally represent writers but do not own the underlying book rights.

What the dispute means

The case shows that AI copyright disputes do not end when a judge approves a settlement. The next challenge is administrative: identifying the correct rightsholder, interpreting old contracts and delivering funds through a process that authors can audit and contest.

Authors should preserve contracts, rights-reversion notices and royalty records, then compare those documents with any settlement claim. Publishers and agencies, meanwhile, need a traceable rights chain before filing automated claims. For future AI copyright settlements, transparent eligibility rules, accessible dispute procedures and verifiable allocation records will be as important as the headline settlement amount. A court-approved framework can still lose credibility if the money is assigned using outdated or incomplete rights data.

Source: TechCrunch AI