OpenAI Accuses Moonshot of Distillation Campaign
As Chinese models become more powerful, OpenAI and Anthropic are amping up their criticism of Chinese labs for alleged distillation.
OpenAI said it recently identified and mitigated a “coordinated model-distillation campaign” involving individuals associated with Moonshot AI, the Chinese developer of Kimi models.
OpenAI said the operators of the campaign tried to extract protected reasoning—the model’s hidden thinking process before it generates an answer—by interacting with models in ways that could make their reasoning visible. This included “copying encrypted reasoning from one conversation and asking a model in another conversation to decrypt and transcribe the hidden reasoning content,” the company said on Thursday.
While it was unclear whether all operators originated from a single actor, OpenAI attributed a core cluster of the activity to individuals linked to Moonshot.
Distillation, a common technique used in AI training, has become a highly controversial topic due to the intensifying tech rivalry between the U.S. and China. OpenAI has previously accused DeepSeek of distilling its models, while Anthropic has accused many Chinese AI labs, including Moonshot and DeepSeek, of engaging in industrial-scale distillation attacks against its Claude models.
OpenAI said low-volume distillation activity began on July 1. The volume jumped on July 24 and July 25 to 16,000 requests from more than 4,000 users. The company eventually identified similar activity from more than 15,000 users, and fully disrupted the activity by July 28.
“Adversarial distillation poses safety and national security risks. Extracted reasoning could be used to train another model without preserving the safeguards applied to the original model’s user-facing outputs,” OpenAI said.