WSJ Tech News Briefing · Thursday, September 10, 2026
Anthropic has reported that Chinese AI companies, including Moonshot AI and DeepSeek, allegedly attempted to clone American AI models. This was reportedly done by routing millions of real user queries through intermediary platforms to Anthropic's Claude, some of which contained sensitive data like passwords. The report also noted increasing evidence of Claude being used for criminal hacking and potentially dangerous biological research, with OpenAI observing similar query types.
“Enthrofic has released a report that alleges Chinese AI companies, including Moonshot AI and DeepSeek, tried cloning American AI models by routing millions of real user queries to Enthrofic's Claude through intermediary platforms.”
“Some of those queries contain sensitive information, like location data and passwords.”
“In its report, Enthrofic said that it has seen increasing evidence that Claude was used for criminal hacking campaigns and potentially dangerous biological research.”