Microsoft quietly removed a blog post offering developers a guide to train large language models (LLMs) using the Harry Potter book series. The November 2024 post, authored by senior product manager Pooja Kamath, suggested leveraging the series to demonstrate Microsoft's new Azure feature. It recommended downloading a Kaggle dataset containing all seven books, incorrectly marked as public domain, to create Q&A systems and AI-generated fan fiction.

Microsoft has not publicly explained the deletion, though the move followed significant criticism on Hacker News. Critics argued the guide encouraged copyright infringement, as J.K. Rowling actively protects her works.

The incident highlights ongoing legal risks for AI firms training models on copyrighted material, a practice facing lawsuits. Microsoft's action, likely a damage control measure, underscores the sensitivity around using beloved intellectual property to promote AI tools.