OpenAI said it identified and shut down a coordinated effort to extract protected reasoning from its artificial intelligence models, with the company linking a core cluster of the activity to Chinese AI startup Moonshot AI, which develops Kimi.
The activity began in early July before intensifying sharply, with OpenAI recording 16,000 requests from more than 4,000 users over a two-day period. The company said its investigation ultimately uncovered related activity involving more than 15,000 users and that the campaign had been fully disrupted by July 28.
Growing Concern Over AI Model Distillation
OpenAI described the activity as “adversarial distillation,” a process in which outputs or reasoning from one AI model are used to train or improve another. The company said attempts to extract protected reasoning could enable competitors to replicate advanced capabilities without making comparable investments in developing and securing frontier AI systems.
OpenAI also said such activity could create potential safety and national security risks.
The company said the campaign did not involve a breach of its encryption, databases or users’ stored conversations. Instead, the operators allegedly manipulated interactions with OpenAI models in an effort to reproduce hidden reasoning in a format that could be accessed by the requester.
The disclosure comes weeks after Anthropic accused several Chinese AI developers, including Moonshot AI and Alibaba, of using its Claude model to help train their own AI systems. The allegations highlighted growing concerns among U.S. AI companies over competitors using established models to accelerate the development of rival systems.
OpenAI said it could not determine whether every operator involved in the latest activity was connected to a single organization. However, it attributed a core cluster of the campaign to individuals associated with Moonshot AI.
The company said it has shared its findings with other AI developers through the Frontier Model Forum and with government information-sharing channels as it works to address what it considers a growing challenge for the AI industry.