[REVISION HISTORY]

Allegations of AI data redirection by Chinese firms

16 recorded revisions of this situation.

What changed

2026-10-23 07:43 UTC → 2026-10-24 16:49 UTC · added ~~removed~~

Moonshot AI has come under scrutiny following allegations from Anthropic regarding its Kimi K3 artificial intelligence model. Anthropic claims the company used thousands of accounts to route approximately 300,000 user queries to its own Claude models, presenting the responses as Moonshot’s own work. Following these allegations, the Cyberspace Administration of China (CAC) launched an investigation into Moonshot AI and DeepSeek. A September 2026 threat intelligence report from Anthropic expanded these claims, alleging that several Chinese firms engaged in “illicit distillation.” Specifically, Anthropic alleged that DeepSeek rerouted over 12.1 million queries to Claude within a 14-day period in July to capture the model’s “thinking signature” for cheaper training. Recent details from the Anthropic report suggest the scale of Moonshot AI’s activity may be significantly higher, alleging the company routed approximately 23 million user prompts to Claude between May and July 2026. The report further claims these interactions included sensitive information, such as data from a Russian government agency and a Chinese municipal police system. In response to the report, the CAC began investigating whether sensitive Chinese military, police, and corporate data was inadvertently transferred to U.S. servers, potentially violating China’s data security laws. In early October 2026, OpenAI joined the scrutiny, announcing it had blocked a coordinated campaign aimed at extracting the protected internal reasoning processes of its GPT models, which it described as “adversarial distillation.” OpenAI traced the actors to individuals associated with Moonshot AI, noting a surge of 16,000 extraction requests in late July. On October 18, 2026, the U.S. Department of Homeland Security (DHS) ordered the removal of DeepSeek products and any third-party services utilizing its models from all U.S. civil agency IT systems. On October 24, 2026, the United States expanded these restrictions, implementing new sanctions and bans targeting Chinese AI entities.

Versions

Select a version to see what changed compared to the one before it.