Microsoft is hit with a lawsuit claiming it trained its Megatron AI on nearly 200,000 pirated books.
The complaint was filed Tuesday in a New York federal court by authors including Kai Bird, Jia Tolentino, and Daniel Okrent. They say Microsoft used unauthorized digital copies of their books to teach Megatron how to respond to human prompts.
The authors want a court order to stop Microsoft’s alleged copyright violation and seek damages up to $150,000 per infringed work. 
Megatron is a generative AI that creates text based on prompts. The lawsuit claims Microsoft’s AI mimics the "syntax, voice, and themes" of copyrighted works it was trained on without permission.
Microsoft has not responded to requests for comment. The authors’ attorney also declined to comment.
The lawsuit comes amid a string of legal battles as authors, publishers, and creatives sue tech giants including Meta, Anthropic, and OpenAI over unauthorized use of copyrighted content for AI training.
Just a day before this case, a California judge ruled Anthropic’s use of copyrighted materials as fair use but left open possible liability for piracy. On the same day Microsoft’s complaint was filed, another California judge ruled in favor of Meta in a similar AI copyright case, dismissing the authors’ claims mostly due to weak arguments.
The copyright fight erupted soon after ChatGPT’s debut, involving lawsuits over news archives, music, photos, and movie characters. For example, Disney and NBC Universal recently sued Midjourney for using film and TV characters in its AI art.
Tech companies argue their AI training is fair use and warn that paying for every copyrighted work could choke innovation. OpenAI CEO Sam Altman said creating ChatGPT would have been “impossible” without copyrighted material.
Related: Meta wins AI copyright lawsuit as US judge rules against authors
Related: ‘Impossible’ to create AI tools like ChatGPT without copyrighted material, OpenAI says
