A lawyer who used to represent Meta in its copyright battle against publishers is now admitting that the company's AI appears to have ingested pirated books — and if he's right, CEO Mark Zuckerberg could owe upwards of $1 billion in damages. In an interview with New Scientist, Stanford tech law expert Mark Lemley said that Meta's Llama large language model (LLM) was able to spit out the exact contents of copyrighted books because, as he and his team of researchers discovered, it was trained on a dataset called "Books3" that almost 200,000 copyrighted publications, most of which were pirated. When […]