OpenAI Ties 16,000-Request Reasoning Extraction Campaign to Moonshot AI
Updated
Updated · CNBC · Oct 1
OpenAI Ties 16,000-Request Reasoning Extraction Campaign to Moonshot AI
3 articles · Updated · CNBC · Oct 1
Summary
OpenAI said operators tried to pull protected reasoning from its models through manipulated prompts, not by breaching encryption, databases or stored user chats.
The activity began in early July, then spiked to 16,000 requests from more than 4,000 users over two days; OpenAI said it later traced related activity across over 15,000 users and fully disrupted it by July 28.
OpenAI attributed a core cluster to individuals associated with Moonshot AI, the Chinese startup behind Kimi, while saying it remains unclear whether every operator was tied to a single actor.
The company described the tactic as adversarial distillation that could let rivals replicate advanced capabilities faster and more cheaply, and said it has shared the findings with other AI developers and government channels.