DeepSeek AI Security Breach Exposes Over a Million Sensitive User Logs

Cybersecurity researchers at Wiz discovered a major security lapse at Chinese AI startup DeepSeek, known for its DeepSeek-R1 AI model. The company had left a ClickHouse database exposed, which contained over one million log entries with chat history, secret keys, and backend details. The exposed data included API secrets, internal logs, and plaintext chat messages. Wiz researchers responsibly disclosed the issue to DeepSeek, who promptly secured the database. This breach comes at a pivotal moment for DeepSeek, as the company has seen rapid success but may face regulatory hurdles if concerns over data security persist.