Experts hacked DeepSeek in just a few minutes.
The startup DeepSeek has already transformed the landscape of artificial intelligence since the announcement of its latest and most powerful open-source model, R1. However, this doesn't imply that Chinese developers are invincible.
2025-02-03 10:01:16https://naked-science.ru/community/1027748
Experts from Wiz, a cybersecurity company, investigated the internal components of a groundbreaking AI model's databases. Within minutes, they effortlessly gained access to a vast array of completely unencrypted internal data.
“This database contained a significant volume of chat history, internal data, and confidential user information,” representatives from Wiz explained in their report on system vulnerabilities.
After accessing the DeepSeek databases, Wiz researchers learned a great deal about how the Chinese startup's model operates, including that its infrastructure closely mirrors that of the OpenAI chatbot.