AI firms are bulk-buying hundreds of thousands of old Polish books for training data

AI development companies have begun bulk-purchasing books published in Poland before 2002 to use as training data for language models. They specifically target pre-LLM-era titles because the industry considers texts produced after large language models appeared to be "contaminated" by AI. The practice is raising concern among Polish publishers and authors.

Similar stories

💻 Technology AI companies buy millions of books, scan them for training data, then discard originals Decrypt · 8d ago 💻 Technology GPT-6 tipped for early September launch as OpenAI teases "Astra" model capabilities AntyWeb · 1d ago 💻 Technology Reddit's new AI tool judges whether posts break subreddit rules Android Authority · 1d ago