ChatGPT copied to train Kimi AI? OpenAI makes serious allegations against China’s Moonshot AI

OpenAI has alleged that China-based artificial intelligence (AI) company Moonshot AI has led a coordinated effort to distill its AI model's reasoning capabilities for the Kimi model. In a blog post, OpenAI revealed that people repeatedly drafted prompts to make its AI models reveal how they arrived at answers, then potentially used those outputs to train another AI model.

“Protected reasoning is the model’s internal record for working through a task; extracting it can reveal information withheld from the final answer and help others reproduce the model’s capabilities,” OpenAI wrote. 

Must read: OpenAI faces lawsuit over Hugging Face system breach incident: What you need to know

How Moonshot AI distilled OpenAI’s ChatGPT

The company revealed that it detected and stopped a large-scale attempt to extract hidden reasoning from its AI models, which allegedly began on July 1 and intensified later that month. OpenAI described the campaign as “adversarial distillation,” which uses one AI model’s outputs or reasoning to train or improve another AI model.

However, OpenAI confirmed that Moonshot AI did not hack its systems, break encryption, access stored chats, or breach a database. Instead, they allegedly manipulated interactions with OpenAI’s models using carefully designed prompts to make normally hidden reasoning visible. 

Must read: OpenAI cancels new AI model launch last-minute over safety concerns: All details

The company identified a large-scale effort to extract hidden reasoning from its AI models. On July 24 and 25, OpenAI saw up to 16,000 requests using a particular prompting technique, coming from more than 4,000 users. After further investigation, OpenAI says it found that more than 15,000 users were using similar prompts. By July 28, the campaign was disabled completely.

Because of this, OpenAI says defending against AI model distillation requires multiple layers of security measures that can adapt as the attackers change their techniques. Although the company is unsure whether all people involved were part of the same group, but it believes a core group of people was associated with Moonshot AI, the Chinese company behind Kimi. If true, the allegations raise the possibility that Moonshot AI used OpenAI’s models to help develop and train its latest model, Kimi K3, which was released on July 16, 2026.