Security researchers at cloud security company Wiz discovered that DeepSeek left a database completely unencrypted and publicly accessible, exposing a significant volume of chat history, backend data, and sensitive information, according to Wiz's vulnerability report. The researchers said they gained access "within minutes" and found log streams, API secrets, and operational details.
Wiz alerted DeepSeek, and the database was locked down within an hour, Wired reported. However, the security firm said it had difficulty reaching anyone at DeepSeek, resorting to LinkedIn messages and emails to every DeepSeek-related account they could find or guess. Nobody at DeepSeek replied to Wiz's attempts at contact before the database was secured.
Exposure at the Front Door
Wiz vulnerability research head Nir Ohfeld told Wired that this kind of exposure usually appears in neglected services that take hours of scanning to find. With DeepSeek, he said, the security issues were right "at the front door." The researchers also found that DeepSeek's infrastructure mimics OpenAI's almost exactly, according to Wired.
Wiz chief technology officer Ami Luttwak told Wired that while mistakes happen, this was a dramatic one because the effort level to access the data was very low and the access level obtained was very high. He said the service is not mature enough to be used with any sensitive data at all.
Wiz researchers noted that if anyone else had accessed the databases before they did, they could have taken as much internal data as they wanted with minimal effort. The exposure lacked any authentication or defense mechanism to the outside world, according to the report.
DeepSeek immediately took action to secure its databases once alerted, Wiz said. The incident highlights security challenges for rapidly growing AI companies handling large volumes of user data.
Editorial Notes