Thousands of authors tell AI makers to stop stealing their books

If you ask GPT-4 to write a scene in the style of Margaret Atwood or George Saunders, it will do an uncanny job. The reason is simple: it likely ingested their life's work during training and turned their craft into its own feature. But the people who actually wrote those books have had enough.

More than 8,500 fiction, non-fiction, and poetry authors have signed an open letter taking OpenAI, Alphabet, Meta, and other AI makers to task. The letter demands that these companies stop using copyrighted writings without consent, credit, or compensation.

The endless free buffet

The authors, organized under The Authors Guild, did not mince words. Millions of copyrighted books, poems, and essays have provided the food for modern AI, an endless meal for which tech companies have never paid a dime.

The provenance of training data remains notoriously murky. Did developers scrape snippets from online reviews, digitize whole libraries, or simply download pirated shadow libraries like Libgen? One thing is clear: they never asked the authors or bought commercial licenses.

The letter points to the recent US Supreme Court decision in Warhol v. Goldsmith, arguing that the strictly commercial nature of these AI products invalidates fair use claims. To make matters worse, low-grade AI books have already begun climbing Amazon best-seller lists, and sci-fi publishers are actively drowning in machine-generated submissions.

Instead of launching an immediate lawsuit, the group laid out three clear demands. Companies must obtain permission before using copyrighted works, provide fair compensation for past and ongoing training, and pay writers when their work informs AI outputs, regardless of current copyright technicalities. Authors Guild CEO Mary Rasenberger told NPR that lawsuits take tremendous money and years to resolve, while authors are losing ground right now.

Why it matters

We think this letter puts a spotlight on the original sin of generative AI: the uncredited harvesting of human culture.

Visual artists have already pushed back against image generators, but text has always been trickier to defend. Large language models rarely spit out whole copyrighted paragraphs; instead, they digest style, syntax, and narrative techniques. Skipping costly litigation in favor of public pressure is a smart opening move by the Authors Guild, but tech giants currently have almost zero financial incentive to volunteer settlements.

If courts or regulators eventually force AI labs to license their training corpuses, the economics of frontier models will shift overnight. Smaller startups might get priced out entirely, leaving only deep-pocketed tech giants capable of paying licensing tolls. For writers, the fight is less about stopping technology and more about ensuring that the creators who provide the raw material do not get replaced by the very machines they fed.