Source: Wired

Meta has scored a major victory in a US federal court after a judge dismissed most claims brought by writers who alleged that the company’s use of their copyrighted books to train AI models infringed their rights. The lawsuit centered on whether using books as training data for large language models constitutes copyright infringement, a hotly debated issue as generative AI proliferates.

The judge ruled that the plaintiffs did not adequately demonstrate how Meta’s use of their works for AI training directly harmed their market or violated copyright law under current precedents. The decision found the process to be sufficiently transformative, noting that the AI model does not simply reproduce entire books but learns from patterns in the text.

This ruling is likely to have significant implications for future AI copyright disputes in the United States, signaling judicial support for the fair use defense when it comes to training generative AI. Authors and publishers may now face greater hurdles in similar cases, and experts predict this will shape the broader legal landscape around AI, intellectual property, and innovation.