AI Can Train on Books, But Piracy Costs $1.5 Billion

You know that ChatGPT and other chatbots seem to know everything? That's because they were trained on enormous amounts of written work—including books, articles, and essays—often without the authors' permission. Many writers are upset, but a recent court ruling shows the law isn't clear-cut. In one major case, a judge ordered AI company Anthropic to pay $1.5 billion to writers. But here's the twist: the judge said the training itself was legal. The penalty was for getting the books from pirated online libraries.

The judge compared AI training to a writer studying literature. Just as a human reads books to learn and create something new, AI reads text to learn patterns—not to copy. Copyright law protects against copying, not against reading or learning. That's good news for AI companies because it means they can use copyrighted works as learning material, as long as they don't take them illegally.

But not all cases go that way. In another case, a company called Ross Intelligence lost because it used Thomson Reuters' legal content to build a directly competing product. Courts tend to say no when AI is used to rival the original creator. So authors could win if they prove chatbots are replacing them, but so far that argument hasn't succeeded.

The real problem? Copyright law hasn't been updated since 1976—long before the internet. Judges are trying to apply 50-year-old rules to brand-new technology, and they often disagree. Until Congress updates the law, expect more lawsuits and more mixed rulings. For anyone who writes, reads, or works with creative content, this is a story to watch.