DeepSeek, a popular Chinese AI startup, accidentally left a database open online, potentially giving hackers access to sensitive information.
Gal Nagli, a security researcher from Wiz, discovered that the exposed ClickHouse database could be controlled by anyone.
The exposed data included a million log entries with chat history, secret keys, backend details, API secrets, and other sensitive information. Wiz notified DeepSeek, and after that, the company fixed the issue.
It is said that the database, which is hosted at oauth2callback.deepseek[.]com:9000 and dev.deepseek[.]com:9000, allowed for the illegal access of a large amount of data.
As per the Wiz, the vulnerability made it possible to escalate privileges and have total database control in the DeepSeek environment without the need for authentication. However, it is still unclear if hackers took advantage of exposure before it was secured.
Nagli warned that AI companies need to focus on security as simple mistakes like this can pose serious risks.
Recently launched DeepSeek has gained attention for its open-source AI models, which are compatible with other top plates like OpenAI. Its chatbot quickly became a top app but faced “large-scale attacks,” forcing it to pause new sign-ups.
