DeepSeek and Moonshot accused of secretly relaying user prompts to Claude via fake accounts

burny_tech · x · 2026-09-11

A viral post claims DeepSeek and Moonshot relayed customer prompts to Claude through fraudulent accounts, saving exchanges to train their own models—users thought they were getting Chinese AI but received Claude's answers. It alleges Moonshot forwarded nearly 300,000 requests in ten days, that DeepSeek flagged users running models inside coding tools like Claude Code and OpenRouter's OpenCode to route them selectively, and cites sensitive cases including military-linked queries. Note: this is a third-party, unverified retelling; treat Anthropic's official threat-intelligence report and the companies' responses as authoritative.

Original post →

More from Companies & People

Companies & People channel →