China investigates DeepSeek and Moonshot for data leak to Claude

China's internet authority CAC has opened an investigation into the two Chinese AI companies DeepSeek and Moonshot. According to the news outlet The Information, the authorities have sent staff out to interview management and employees at the companies to determine how serious the problem is.

The case begins with a report from the American AI company Anthropic, which makes the chatbot Claude. Anthropic claims that DeepSeek and Moonshot not only used Claude in their own development, but also passed their customers' requests on to Claude without saying so. This means users believed they were talking to a Chinese AI system, while their data was in reality being processed by an American company.

According to Anthropic's investigation, Moonshot passed more than 23 million requests on to Claude between May and July. DeepSeek passed on 12.1 million requests in just 14 days in July.

What worries the Chinese authorities most is what type of data may have ended up with the American company. There is suspicion that sensitive information from Chinese police, military and other state systems may have been sent to Claude via this route, including requests involving police systems and national ID numbers.

Neither DeepSeek nor Moonshot has commented publicly on the case, and it has not yet been decided whether the companies will be punished. The news caused share prices of Chinese AI companies to fall.

For ordinary users, the case is about the fact that you don't always know which company is actually processing your data when you use a chatbot. Even when you think you are talking to a local system, your request may in practice be passed on to an entirely different company in another country.