According to Wiz, DeepSeek did not secure the database infrastructure of its services, leaving some data and chat histories accessible from the public internet with no password required. Researchers said they discovered the publicly accessible ClickHouse database linked to the Chinese company “within minutes” of beginning their investigation.

Wiz said the publicly accessible database allowed full control over database operations, including the ability to access internal data. 

“This database contained a significant volume of chat history, backend data, and sensitive information, including log streams, API Secrets, and operational details,” the firm said in a Wednesday blog post detailing the data exposure.

Wiz said it had contacted DeepSeek to alert the AI company of the error and it had subsequently secured the data exposure. In the blog post, Wiz warned that the “rapid adoption of AI services without corresponding security is inherently risky.”

“This exposure underscores the fact that the immediate security risks for AI applications stem from the infrastructure and tools supporting them. While much of the attention around AI security is focused on futuristic threats, the real dangers often come from basic risks—like accidental external exposure of databases,” the company said. 

Questions about DeepSeek’s data use

DeepSeek is in the midst of a burst of popularity that sent its AI assistant app rocketing to the top of app stores. However, some countries are already raising concerns about the company’s use of personal data.

Both Italy and Ireland’s data-protection authorities have separately questioned DeepSeek’s use of personal data.

The Italian regulator has asked the company several questions about its data use, including what personal data is collected, on what legal basis, and whether it is stored in China. The app has since become unavailable for download from some app stores in Italy. The Irish data regulator has also requested similar information from DeepSeek on data processing.

Earlier this week, DeepSeek also suffered a cyberattack that forced the company to pause new sign-ups for its AI assistant. On Tuesday, the company said it was the victim of a “large-scale malicious attack” against the platform. The company has since resumed normal operations.

DeepSeek shocks Big Tech

DeepSeek has made headlines and sent shockwaves through the tech world with its low-cost, open-source R1 reasoning model.

The Chinese AI company model outperforms OpenAI’s o1 in some tests but was supposedly built at a fraction of the price with less powerful chips. The U.S. has attempted to limit China’s AI progress by restricting the country’s access to high-powered AI chips through a range of export controls.

The prospect of low-cost AI models trained with less advanced chips sent some tech stocks like Nvidia plummeting earlier this week as investor confidence in Big Tech’s expensive AI strategy faltered.

The success of the Chinese company has lit a fire under some of the U.S.’s leading AI labs, most of which have broadly labeled DeepSeek a good thing for innovation.

OpenAI CEO Sam Altman said he plans to move up some of the company’s releases in response to DeepSeek’s impressive tech, while engineers at Meta are reportedly studying the open-source model in four “war rooms.”

Rumors have been circulating that DeepSeek may have been helped by OpenAI’s existing reasoning model, o1.

Microsoft and OpenAI are reportedly probing whether DeepSeek trained its model with o1 via a technique called distillation. In machine learning, distillation is a process where outputs from a large pretrained model are used to train another, usually smaller model to exhibit similar capabilities. If true, DeepSeek’s actions would amount to a violation of OpenAI’s terms of service.