DeepSeek Leaked Sensitive Information

A serious security flaw at Chinese AI startup DeepSeek exposed over a million chat logs and sensitive data - completely unprotected. How many AI companies actually have similar security holes?

Share
DeepSeek Leaked Sensitive Information

Security researchers at Wiz Research have discovered a serious security flaw at the Chinese AI startup DeepSeek. A publicly accessible ClickHouse database exposed over a million log entries, including chat logs, secret API keys, and internal backend details.

Full Access Without Authentication

Wiz Research identified that the database, accessible via oauth2callback.deepseek.com:9000 and dev.deepseek.com:9000, was completely unprotected and allowed full control over the data content. The flaw meant anyone could:

Gain access to internal chat logs and system data

Read and use secret keys and API access

Modify or delete data in the database

DeepSeek, which recently gained major attention for its DeepSeek-R1 model, an AI system that challenges OpenAI's models in performance and cost-efficiency, took immediate action after Wiz Research reported the security hole.

Lacking Security in the AI Industry

The discovery highlights a larger problem in the AI industry - the lack of robust security measures around sensitive databases. AI startups handle enormous amounts of sensitive information, making them attractive targets for cyberattacks.