Shares of Chinese AI developers fell following reports of an investigation into DeepSeek and Moonshot
China's internet regulator summoned all seven companies mentioned in the Anthropic report, but focused on two of them
Moonshot AI, a startup preparing for a $3 billion IPO, may have inadvertently leaked data from Chinese surveillance cameras to the U.S.-based company Anthropic, according to media reports / Photo: Tada Images/Shutterstock.com
Shares of Chinese artificial intelligence developers plummeted following reports of an investigation into DeepSeek and Moonshot AI due to data security concerns, according to Bloomberg. Investors are concerned that Beijing will tighten its control over the industry, which would slow the development of Chinese AI models.
The market was spooked by the inspections
Shares of the AI startup Z.AI (Zhipu) plummeted 12% during trading on September 23 in Hong Kong. Shares of rival company MiniMax fell 4.1%. Tech giant Alibaba, which is developing the Qwen AI model, saw its stock price drop by nearly 4.4%. Shares of other tech companies with their own AI models also fell: Xiaomi lost 3.8% during the session, and Tencent lost 2.4%.
The sale followed a report by The Information: According to the publication, China’s internet regulator launched an investigation after the U.S.-based company Anthropic accused both startups of running “sensitive” data through its models, Bloomberg reports.
Investors “fear” that Beijing will tighten oversight of the industry, Bloomberg reports, citing Leonid Mironov, a portfolio manager at Gavekal Capital. According to the expert, market sentiment has deteriorated due to uncertainty regarding the extent of regulation, which could slow down the development of Chinese AI models.
Anthropic List
On September 10, the American company Anthropic stated that seven Chinese AI developers had sent customer queries to its Claude neural network in order to collect its responses. Anthropic refers to this practice as “illegal distillation.” In addition to DeepSeek and Moonshot AI, Anthropic accused Alibaba, Z.AI, MiniMax, Xiaomi, and SenseTime of this practice.
According to The Next Web, citing The Information, the Cyberspace Administration of China (CAC) interviewed representatives from all seven companies mentioned in the Anthropic report. However, the regulator subsequently focused its investigation into a possible data leak on DeepSeek and Moonshot.
According to Anthropic, Alibaba accounted for more than 151 million of the approximately 190 million message exchanges with Claude. However, it did not become the subject of the investigation: according to The Information, the scope of the review “narrowed due to the specific examples detailed by Anthropic, rather than the overall numbers.”
One of the incidents, as The Information discovered, involved a user whom Anthropic believed might be affiliated with the Chinese military. He asked Moonshot’s Kimi model to analyze a video that tracked a person’s movements through the city. The footage was captured by hundreds of police cameras, including those near military facilities and defense-related institutions. Moonshot allegedly forwarded the request and the video to Claude without notifying the user, according to The Next Web.
Delayed inspection
DeepSeek is scheduled to brief the UN Security Council this week on the risks AI poses to international security, according to Reuters. The news agency reports that Moonshot has also been invited to participate in the event.
Moonshot faces another problem: the company is preparing for a $3 billion initial public offering in Hong Kong. The ongoing regulatory investigation is “something that needs to be disclosed in the IPO prospectus,” notes The Next Web.
The Hong Kong newspaper *South China Morning Post* also notes that the heads of DeepSeek and Moonshot are not listed in any of the published versions of the business delegation roster that may accompany Chinese President Xi Jinping during his visit to the United States.
This article was AI-translated and verified by a human editor
