DeepSeek left a database of chat logs and API keys open to the internet
What happened
Wiz found two publicly accessible DeepSeek ClickHouse databases that allowed anyone to run queries without logging in. They held over a million log lines including users' chat history in plain text, API keys and backend details. DeepSeek secured them quickly after being told.
Shortly after DeepSeek's R1 model drew global attention, Wiz Research scanned the company's internet-facing systems and within minutes found two ClickHouse database servers open to anyone, with no authentication. Using ClickHouse's built-in web interface, researchers could run queries directly. The databases held more than a million lines of logs, including users' chat history in plain text, API secrets and keys, and details of DeepSeek's back-end systems and operations. The level of access would also have let an attacker take control of the databases and potentially move deeper into DeepSeek's environment.
How it happened
Leaky AI product: Exposed datastore. A database, storage bucket or server behind an AI product was left open on the internet without a password.
The AI involved was from DeepSeek. The product was DeepSeek chat, a type of chatbot.
Impact
Data was exposed. Data involved: chat logs and credentials. The impact was confined to one organization and its users or customers.
Severity is rated High, calculated from how far the incident reached and how sensitive the data was.
Was the AI at fault?
No. The failure was in the systems, settings or people around the AI, not in the model's behaviour.
Status and timeline
Confirmed by the affected organization, the AI vendor, a regulator or a named security research firm.
It became public on 29 Jan 2025.
DeepSeek secured the databases promptly after Wiz's disclosure. It did not comment publicly on whether anyone else had accessed them.