OpenAI Disrupts Model Distillation Campaign Linked to Moonshot AI
In July 2026, OpenAI stopped a coordinated attack attempting to extract protected reasoning data from its models. The company linked part of the activity to Moonshot AI, the developer of Kimi.
OpenAI announced that it has detected and disrupted a coordinated 'adversarial distillation' campaign that began in the first week of July 2026, aimed at unauthorized acquisition of its models' protected reasoning processes.
Attackers attempted to make normally hidden reasoning data visible by manipulating model interactions. This activity, which peaked between July 24 and 25 with 16,000 requests sent via more than 4,000 users, was conducted through a network covering over 15,000 users in total. OpenAI announced that it had completely halted the entire process as of July 28.
The company attributed a core set of the activity to individuals associated with Moonshot AI, the developer of the Kimi model. As a security measure, OpenAI closed fake accounts, strengthened infrastructure controls, and blocked pathways that allowed for the replay of encrypted reasoning data. The findings were shared with the Frontier Model Forum and relevant government agencies to protect the ecosystem against similar risks.
Sources
- openai.com · Link