China’s Cyberspace Administration reportedly summoned representatives from all seven Chinese companies named in Anthropic’s report before narrowing its focus to DeepSeek and Moonshot.
- According to The Information, the CAC is examining whether sensitive Chinese police, military and state-linked corporate data was transferred to U.S. AI models through the companies’ use of Claude.
- The probe follows Anthropic’s identification of millions of Claude exchanges involving DeepSeek and Moonshot, raising data-handling concerns in China.
- The probe comes ahead of President Donald Trump’s meeting with Chinese President Xi Jinping, with both sides discussing a potential AI incident-reporting mechanism.
China’s internet regulator has reportedly opened an investigation into DeepSeek and Moonshot AI after Anthropic alleged that the Chinese AI companies secretly routed sensitive user data and requests to its Claude models.
The probe follows a September threat intelligence report from Amazon (AMZN) and Alphabet (GOOG, GOOGL)-backed Anthropic that detailed what the company described as large-scale illicit use of Claude by seven Chinese AI companies – DeepSeek, Moonshot, Alibaba (BABA), Zhipu, SenseTime, MiniMax and Xiaomi.
According to a report by The Information, China’s Cyberspace Administration (CAC) summoned representatives of all seven companies named in Anthropic’s report before focusing more specifically on DeepSeek and Moonshot.
The regulator is reportedly examining whether sensitive data involving Chinese police, military and state-linked companies was transferred to U.S. AI models.
DeepSeek, Moonshot Accused Of Routing Claude Requests
Anthropic’s investigation found that DeepSeek and Moonshot did more than access Claude for their own development. The company alleged that both firms routed requests from their own users to Claude, meaning customers could have believed they were using a Chinese company’s model while their requests were actually being processed by Anthropic.
In its report, the AI company stated that Moonshot routed more than 23 million exchanges to Claude between May and July. DeepSeek, meanwhile, routed more than 12.1 million exchanges over a 14-day period in July. It added that some of the traffic included sensitive information.
One example involved engineers working on a case-management system for a municipal Public Security Bureau in China. Anthropic said DeepSeek relayed those engineers’ requests to Claude, including data used to compare an individual’s movements with police records and national identification numbers.
Anthropic also said Moonshot routed requests involving surveillance footage from hundreds of cameras in Chengdu, including cameras near facilities associated with the People’s Liberation Army and state-owned enterprises.
Probe Lands Ahead Of Trump-Xi AI Talks
President Donald Trump and Chinese President Xi Jinping are scheduled to meet in Washington on Thursday. Ahead of the summit, U.S. and Chinese officials have discussed establishing an AI dialogue that could include a mechanism to notify each other of significant AI incidents with national security implications.
U.S. officials, including Treasury Secretary Scott Bessent, have emphasized greater transparency between Washington and Beijing, while the two governments continue to compete over advanced AI technology and infrastructure.
Anthropic is also reportedly preparing for an IPO in November, adding another layer of investor attention to the company’s rapidly expanding role in the global AI market.
Nvidia’s Open-Weight Bet Meets China’s AI Push
The allegations also put Nvidia (NVDA) in an interesting position. CEO Jensen Huang has publicly defended Chinese open-weight models, including those from DeepSeek and Moonshot, stating that strong open models should be used and can expand overall AI adoption.
Anthropic is accusing those same companies of using Claude, which is a closed model, to extract capabilities for their own models.
Chinese AI companies have also continued to use Nvidia chips to train and run models. DeepSeek, for example, has used both Nvidia and Huawei hardware, with its R1 model trained on Nvidia’s H800 chips before the company shifted more of its workload toward Huawei processors.
NVDA stock edged 0.5% lower in early morning trade on Tuesday. On Stocktwits, retail sentiment around the AI bellwether remained in ‘bearish’ territory over the past day.
China AI Probe Puts Focus On AMZN, GOOGL, BABA
AMZN stock edged 0.12% higher in early trading, while retail sentiment improved to ‘neutral’ from ‘bearish’ territory over the past day. GOOGL stock was up about 0.3%, while sentiment remained in the ‘bearish’ zone.
BABA stock rose roughly 3% in pre-market trade. CEO Eddie Wu said at the company’s Apsara Conference that Alibaba Cloud plans to expand its global data-center capacity to more than 20 gigawatts by 2032. The company also unveiled its Zhenwu V900 AI chip and called it the ‘most powerful in China today.’
Read also: AMD Stock Hits Record High, Briefly Joins $1 Trillion Club Amid Chip Sector Rally
For updates and corrections, email newsroom[at]stocktwits[dot]com.
Read about our editorial guidelines and ethics policy
