IE 11 Not Supported

For optimal browsing, we recommend Chrome, Firefox or Safari browsers.

How long did it take to hack a million lines of DeepSeek chat history?

Answer: A few minutes.

Hands typing on a laptop keyboard with lines of code on the screen.
As AI chatbots rise in popularity, so must their security. Cybersecurity researchers proved so this week when they hacked one of the industry’s newest offerings in just minutes. The DeepSeek chatbot created by a Chinese startup took the world by storm this week when it emerged as a rival to the favored ChatGPT from OpenAI. So a team from cloud security company Wiz decided to put DeepSeek’s cybersecurity to the test.

After only a few minutes of basic searches, the team discovered a ClickHouse database owned by DeepSeek through which they easily accessed a set of available data sets, including more than a million lines of chat history that included sensitive information. They were also able to control the database’s operations, which would have given them the ability to escalate their access privileges had they been nefarious actors. Wiz notified DeepSeek of the vulnerability and it has since been secured, but it’s possible that the data was already exposed.
Sign up for GovTech Today

Delivered daily to your inbox to stay on top of the latest state & local government technology trends.