Microsoft AI Model Under Scrutiny: Authors Allege “Stolen” Training Data

Microsoft’s Megatron AI model is under intense scrutiny following a lawsuit from a group of high-profile authors who allege the company utilized nearly 200,000 pirated books as training data. This legal action exacerbates the ongoing tension between intellectual property rights and the rapid advancements in artificial intelligence. The authors explicitly state that the AI was designed to emulate their unique literary styles and themes.
The complaint, lodged in New York federal court, demands a court order to halt Microsoft’s alleged copyright violations and seeks significant financial compensation, with potential statutory damages reaching $150,000 for each alleged instance of misuse. The authors emphasize the critical role that massive datasets play in enabling generative AI to produce realistic and diverse content. They assert that the pirated collection was central to the AI’s mimetic capabilities.
Microsoft has not yet provided an official response to the lawsuit, and the authors’ legal counsel has not offered a comment. This case joins a growing list of high-stakes legal battles involving AI and copyright, including recent rulings in California concerning Anthropic and Meta, illustrating the evolving legal challenges facing the AI industry.
The legal fight over AI and copyright is expanding across various media. The New York Times and Dow Jones have sued AI firms over their archival content, while major record labels and visual artists are also pursuing legal action. Tech companies generally argue for fair use, maintaining that their AI creates transformative new content and that overly strict copyright enforcement could impede the progress of the AI sector.