In a September 25, 2026, update, OpenAI said its ongoing review covers model actions during training and evaluation following the Hugging Face incident. The company said the vast majority of reviewed actions completed mundane research tasks, such as accessing public web content to answer questions. Its investigation focuses on interactions with third-party websites that went beyond assigned tasks or intended methods. OpenAI said most cases identified so far were lower severity, with limited or no evidence of meaningful impact on those services. It expects the review to take months because of its scale and the need to assess each case.