Writers Face Growing Legal Battles Over AI Training Data
Federal courts are increasingly scrutinizing how tech companies use copyrighted material to train AI systems, but judges have shown little sympathy for writers' claims so far.
On February 5, authors including Richard Kadrey sued Meta in San Francisco, alleging the company used approximately 190,000 illegally copied books to train its Llama AI model. The case could force Meta to pay substantial damages if the writers prevail.
U.S. District Judge Vince Chhabria, however, has sided with Meta at multiple turns. In March 2026, he allowed plaintiffs to amend their copyright claims but criticized their legal strategy, saying their lawyers had made "many" mistakes and were blaming Meta's counsel instead of taking responsibility for their own missteps.
Dictionary Publishers Face Similar Setbacks
Encyclopaedia Britannica and Merriam-Webster filed a separate lawsuit against OpenAI in March 2026, claiming the company used thousands of copyrighted articles and dictionary entries without permission to train ChatGPT.
The publishers argue that ChatGPT's responses sometimes closely resemble or directly reproduce their original content, which violates copyright law. They also worry the AI will cannibalize their website traffic and revenue.
OpenAI counters that the dictionaries are publicly available to everyone, making their use fair game.
Legal Framework Remains Undefined
The core problem: courts have not settled whether training an AI model on copyrighted material constitutes infringement. A June 2026 ruling found that training an AI model does not automatically violate copyright law, leaving the legal boundaries unclear.
For writers, the stakes are direct. Until courts establish clearer rules, companies can continue using published work to build AI systems without permission or compensation. Learn how AI affects your profession and what protections may emerge as these cases proceed.
Your membership also unlocks:
