Anthropic's $1.5 billion settlement over the use of copyrighted books has run into a new obstacle: authors and publishers cannot agree on how the money should be split. The deal, described as the largest copyright settlement in US history, was meant to resolve claims that the company trained its AI models on books without permission. Now the process has descended into a dispute over who deserves what share of the payout.
What happened
According to reports, the settlement has become bogged down in disagreements between writers and publishing houses. Both sides claim a right to a significant portion of the funds, and the lack of a clear formula for distribution has turned the agreement into a chaotic negotiation. The conflict highlights a broader problem: when a single lump sum is meant to compensate a diverse group of rights holders, deciding who gets paid and how much is far from straightforward.
The settlement itself was intended to close a long-running legal battle over the use of copyrighted works in AI training. Anthropic agreed to pay $1.5 billion, a figure that underscores how serious the copyright question has become for AI developers. But the money has not yet reached the people who say their work was used without consent, and the current infighting risks delaying or even reshaping the final payouts.
Authors argue that their individual works were the foundation of the training data, while publishers point to their role in commissioning, editing and distributing those books. Each side sees itself as the primary victim and the rightful recipient. Without a pre-agreed mechanism, the court or a mediator may have to step in, which could add months to the process and reduce the net amount that ends up in creators' hands.
What it means for business
For companies that build or deploy AI, the case is a reminder that legal risks around training data do not end with a headline settlement. Even a record-breaking payment can be followed by years of distribution disputes, reputational damage and uncertainty about future liabilities. Firms that rely on third-party content should treat data provenance as a core operational issue, not a legal afterthought.
Businesses integrating AI agents into sales, support or analytics should also watch how this unfolds. Clear rules on data ownership and compensation will eventually shape what content can be used and at what cost. Until then, the safest approach is to document data sources, respect licensing terms and build systems that can adapt if the rules change. The Anthropic case shows that the price of ignoring copyright can be enormous, and that paying it is only the first step.
