Microsoft’s artificial intelligence training practices have become a legal quagmire, as a group of high-profile authors has filed a lawsuit alleging the company used nearly 200,000 pirated books to train its Megatron AI. This legal challenge spotlights the complex and contentious issue of data sourcing for advanced AI models. The authors claim that their copyrighted works were used without permission to teach the AI to generate responses that mimic their original writings.
Filed in New York federal court, the complaint seeks both injunctive relief to prevent future infringement and substantial statutory damages, potentially up to $150,000 per alleged misuse. The plaintiffs contend that generative AI models, which produce a variety of media, are critically dependent on large datasets for their learning process. They specifically pointed to the pirated dataset as key to the AI’s ability to imitate human creative output.
As of now, Microsoft spokespeople have not issued a statement in response to the lawsuit, and the authors’ attorney has declined to comment. This legal action comes on the heels of other landmark copyright decisions in the AI space, notably those involving Anthropic and Meta in California.
The legal landscape surrounding AI and copyright is becoming increasingly complex and litigious. Beyond books, lawsuits have been filed by major news organizations, record labels, and photography companies, all asserting their rights against AI companies. Tech companies consistently argue for fair use, claiming their AI produces transformative content and that stringent copyright restrictions could stifle innovation within the burgeoning AI industry.
