Expert talks about the safe use of neural networks

Information security expert Yury Drachenin advised users of neural networks not to upload critical data there.

"Everything you send to the AI can be compromised, used against you or even sold," the expert stressed in a conversation with Lenta.Ru on Friday, January 31.

According to him, this applies to trade secrets, business documents, personal data and even ordinary questions, as "the context of your inquiries can also say a lot about your company, project or strategy."

The specialist also noted that AI models should be protected at the level of architecture, but there are always risks of vulnerabilities or just trivial human error. He said that everything new invariably arouses interest, and in such situations "white" hackers become like "black" hackers, wanting to test the system for vulnerabilities. As a result, "it is simply impossible to say how many people have managed to conduct such checks, achieved the same results and downloaded the bases," Drachenin added.

A tip for the spy: what consequences the DeepSeek data leak will cause

Chinese AI model left more than 1 million lines of user information unprotected

On January 30, US cybersecurity company Wiz discovered the user database of Chinese AI model DeepSeek on the Internet. More than one million lines of sensitive information, including the history of user requests, as well as digital program keys, became publicly available.