AI Training Data, Copyright Risk And Commercial Liability: What Businesses Need To Know

On 8 October 2026, USA Today Co. and its affiliated newspaper companies, which publish 19 titles including USA Today, the Detroit Free Press and The Arizona Republic, sued seven OpenAI entities in the US District Court for the Southern District of New York. The complaint alleges that OpenAI copied their journalism to train its models, stripped identifying information from it, and now uses those models to give ChatGPT users summaries that substitute for the original reporting. The publishers seek more than USD 250 million in damages, an injunction and an order for the destruction of models and training sets containing their content.

These are allegations, not findings, and none of them has yet been tested in court. However, the case is a useful lens for any business building, buying or integrating AI tools, because it separates neatly into the three distinct ways in which copyrighted material can create legal exposure, i.e., how training data is acquired, what the system produces, and