Top AI Firms Tell Copyright Office Model Training is Fair Use

AI companies facing a flurry of copyright infringement lawsuits told the US Copyright Office that scraping enormous volumes of text and images to train large language models without permission is legal under copyright’s fair use exception.

The office released the comments from OpenAI Inc., Meta Platforms Inc., Microsoft Corp., Stability AI Inc., Anthropic PBC, and Google LLC as part of its inquiry examining how copyright law applies to generative AI.

The office has received more than 9,700 comments since the inquiry was announced in late August. Reply comments are due on Nov. 29, and the office has ...