Microsoft has filed motions in a New York federal court regarding a class-action copyright lawsuit. The company is requesting the court to recognize the use of books and news materials for training large language models as fair use.
The lawsuits combine claims from authors and news publishers who allege copyright infringement during the AI training process. Microsoft argues that training is transformative and creates new quality rather than copying originals, asserting that AI outputs do not replace protected works.
As part of the evidence, Microsoft provided an expert analysis of 8.2 million dialogues with the Copilot chatbot. The analysis claims that only 24 responses contained fragments of at least 30 words that matched the works of the complaining authors. This figure is currently disputed and awaits court verification.
