The Information Machine
ConcludedRan 1 dayin the edition since 26 Sep 20267 sourcesupdated 26 Sep 2026

DeepSeek and Moonshot's alleged Claude routing

The gist

Latest: China's CAC Probes DeepSeek and Moonshot Over Routing Data to Claude

The investigation puts two prominent Chinese AI labs under domestic regulatory scrutiny for allegedly exposing sensitive Chinese user data, including state and military information, to a foreign model. Anthropic's role as the source of detailed evidence supporting the allegations is described by one outlet as the first instance a US AI firm provided such evidence to back claims of this kind.

When stories like this move, I write about them in an email each morning.

One free email each morning, around 8 AM US Eastern.

The full picture

China's Cyberspace Administration of China (CAC) is investigating DeepSeek and Moonshot AI following allegations from Anthropic that both companies secretly routed millions of user requests to Claude and used the responses as training data for their own models, a practice Anthropic calls illicit distillation. The CAC initially summoned representatives from seven companies named in the original report, including Alibaba, Zhipu, SenseTime, MiniMax, and Xiaomi alongside DeepSeek and Moonshot, but has since narrowed its focus to DeepSeek and Moonshot specifically. Anthropic alleges Moonshot sent over 23 million requests to Claude between May and July, and DeepSeek sent 12.1 million over a 14-day period in July. The routed prompts allegedly contained names, email addresses, company data, and information from state-affiliated actors, including data from engineers building a case system for a Chinese police bureau. The probe is ongoing and no penalties have been announced.

How it developed
First in the edition · 26 September 2026
25 September 2026

Second report confirms China's internet regulator launched the investigation following Anthropic's allegation

24 September 2026in the 26 Sep edition

Anthropic alleged September 25 that DeepSeek and Moonshot secretly routed millions of user requests to Claude via fraudulent accounts to harvest training data, a practice it calls illicit distillation, and China's Cyberspace Administration (CAC) narrowed a seven-company probe to those two firms.

Anthropic's count puts Moonshot at over 23 million requests between May and July and DeepSeek at 12.1 million over 14 days in July; no penalties have been announced. The probe coincides with DeepSeek's scheduled UN Security Council briefing on AI risks alongside Anthropic CEO Dario Amodei.

Sources
2 more sources

Get the daily email

When stories like this move, I write about them in an email each morning.

One free email each morning, around 8 AM US Eastern.

See the latest edition · Privacy