It was the company that was on everyone’s lips this week: Deepseek made it speak for having been the main cause of unscrewing on the Nvidia stock exchange. The Chinese start-up has indeed shown with its homemade language model that it was possible to make artificial intelligence at a lower cost, and using fewer Nvidia chips provided for the occasion. In parallel, Openai retaliated by indicating that Deepseek used chatgpt to create his own model.
Deepseek victim of a major security breach
In short, Deepseek is talked about, but also controversial. And the events of this Thursday, January 30, will certainly not help improve confidence with the new Chinese actor in artificial intelligence. Indeed, cybersecurity researchers from Wiz Recently discovered a major security flaw at Deepseek: more than a million newspaper entries, including conversation history, secret keys and backend details, were accessible without protection in the exposed database. And that’s not all: compromise data included API keys, internal newspapers and even text messages. Wiz researchers reported this Deepseek flaw, which quickly secured the database.
But the simple fact that such a breach is possible asks real questions. Admittedly, Deepseek mainly holds its success because its R1 model offers good quality responses quickly, at a cost much lower than that of competition. However, it is this same light and inexpensive infrastructure that contributes to its vulnerabilities in terms of security. It remains to be determined whether the Chinese actor can fill these shortcomings quickly.
It was the company that was on everyone’s lips this week: Deepseek made it speak for having been the main cause of unscrewing on the Nvidia stock exchange. The Chinese start-up has indeed shown with its homemade language model that it was possible to make artificial intelligence at a lower cost, and using fewer Nvidia chips provided for the occasion. In parallel, Openai retaliated by indicating that Deepseek used chatgpt to create his own model.
Deepseek victim of a major security breach
In short, Deepseek is talked about, but also controversial. And the events of this Thursday, January 30, will certainly not help improve confidence with the new Chinese actor in artificial intelligence. Indeed, cybersecurity researchers from Wiz Recently discovered a major security flaw at Deepseek: more than a million newspaper entries, including conversation history, secret keys and backend details, were accessible without protection in the exposed database. And that’s not all: compromise data included API keys, internal newspapers and even text messages. Wiz researchers reported this Deepseek flaw, which quickly secured the database.
But the simple fact that such a breach is possible asks real questions. Admittedly, Deepseek mainly holds its success because its R1 model offers good quality responses quickly, at a cost much lower than that of competition. However, it is this same light and inexpensive infrastructure that contributes to its vulnerabilities in terms of security. It remains to be determined whether the Chinese actor can fill these shortcomings quickly.
