OpenAI says it was hit by organized model distillation, points to Moonshot AI's Kimi

OpenAI said on its website that it recently faced an organized distillation campaign. The activity began in the first week of July and involved unauthorized use of one model's outputs or reasoning to help train, copy, or improve another model.

OpenAI said traffic surged on July 24 and 25, when more than 4,000 users made 16,000 requests. A further review found over 15,000 users with similar prompt activity. By July 28, OpenAI had fully blocked the campaign.

OpenAI said the activity did not break its encryption, hack its databases, or directly access user conversations, but it violated its terms of service. The company shared information with industry partners through the Frontier Model Forum and said it does not yet know the attacking entity, though it believes the main group is linked to China's Moonshot AI, which develops the Kimi model.