Breaking

China Opens Data-Security Probe Into DeepSeek and Moonshot AI After Anthropic Claude-Routing Claims

September 22, 2026 at 21:48 EDT

Foundation Models
AI Agents
Policy & Regulation

China’s internet regulator is investigating DeepSeek and Moonshot AI after Anthropic alleged that the Chinese AI companies routed sensitive user requests through Claude models without clear user consent. The inquiry, reported on Sept. 22, centers on whether customer data may have been exposed to Anthropic as part of alleged model-distillation campaigns involving Claude.

Sept. 22, 2026 · AI data security

China probes Claude routing claims

DeepSeek and Moonshot AI face scrutiny after allegations that sensitive user requests were silently routed through Anthropic's Claude models.

Chinese companies named by Anthropic

23M+

Moonshot-attributed exchanges

12.1M+

DeepSeek-attributed exchanges

5,380

Fraudulent accounts in one Moonshot window

Alleged exchange volumes

A proportional stacked-block comparison of the two companies now under investigation.

23M+

Moonshot AI

12.1M+

DeepSeek

What regulators are testing

Routing

Whether users knew which model handled their requests.

Data exposure

Whether sensitive prompts crossed company or national boundaries.

Model training

Whether Claude outputs were saved for capability distillation.

Why it matters

The probe may force clearer disclosure of model endpoints, routing partners and retention policies.

Core risk

Users may not know when code, credentials or sensitive records are sent to another provider.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.