A new class-action lawsuit filed in California this week alleges that Meta Platforms Inc. unlawfully utilized copyrighted literary works to train its Llama large language model, marking the latest escalation in the ongoing conflict between generative AI developers and content creators. The legal complaint asserts that Meta’s AI system systematically ingested protected material, enabling the software to produce summaries and, in documented instances, verbatim reproductions of original copyrighted books and articles without authorization.
The Growing Conflict Over AI Data Sourcing
The core of the dispute centers on the methodology Meta employed to build its Llama models, which currently serve as the foundation for various AI-powered products. Plaintiffs argue that the company scraped massive datasets from the internet, including digital libraries and author-hosted content, to teach the model linguistic patterns and factual synthesis. By doing so, the lawsuit claims Meta bypassed licensing requirements, effectively stripping authors of their intellectual property rights.
This legal action joins a growing list of grievances against major tech companies, including OpenAI and Google, who face similar scrutiny regarding their data acquisition practices. The central question remains whether training generative models on copyrighted data constitutes
