A major security breach has been discovered at Chinese AI startup DeepSeek, exposing sensitive user data and internal system information through an unsecured database. Cloud security firm Wiz Research identified the vulnerability, which has since been patched.
The exposed ClickHouse database contained over one million log entries, including:
- User chat histories
- API secret keys
- Backend system details
- Operational metadata
The database was publicly accessible without any authentication required, allowing potential attackers full control over database operations. It was hosted on two DeepSeek domains that had open ports typically used for database access.
"This is a dramatic mistake, because the effort level is very low and the access level that we got is very high," said Ami Luttwak, CTO of Wiz. The security researchers noted the database was discovered almost immediately with minimal scanning.
After Wiz Research contacted DeepSeek through multiple channels, the company secured the database within 30 minutes. However, it remains unclear if any malicious actors accessed or downloaded the exposed data before it was locked down.
The incident comes amid DeepSeek's rapid rise in popularity, with its AI chatbot reaching top positions in app stores globally. The company has recently drawn attention for its AI models that claim to rival industry leaders like OpenAI.
The breach highlights growing concerns about security practices in fast-growing AI companies. As these firms handle increasingly sensitive user data, basic security measures like database protection become critical for protecting user privacy.
DeepSeek has not responded to requests for comment about the exposure. The company is already facing scrutiny from regulators in multiple countries regarding its data handling practices and potential security risks.