OpenAI reported that it disrupted a large-scale attempt by users affiliated with the Chinese lab Moonshot AI to extract protected reasoning data from its models. The company stated that the effort involved using fraudulent accounts and various evasion tactics to bypass security measures.
Anthropic has also accused Moonshot AI of sending nearly 300 000 requests through fake accounts for model distillation purposes. Additionally, Michael Kratsios, director of the White House Office of Science and Technology Policy, alleged that the company used stolen data from Anthropic’s Fable model.
Moonshot AI denied the allegations, asserting that its results stem from independent innovation. The company released its Kimi K3 model on July 16, 2026, which features 2.8 trillion parameters and an open-weight architecture, allowing developers to run it locally.
The accusations highlight growing concerns in the US regarding industrial-scale distillation, where companies train systems using outputs from more powerful models. Similar claims have previously been directed at other Chinese labs, including DeepSeek, with government warnings dating back to 2024.
