The head of the research and development group of machine learning “Kaspersky Laboratory” Vladislav Tushkanov recommended using security applications to protect his devices from malicious programs.
In an interview with an anadol correspondent, Tushkanov said that the use of artificial intelligence assistants (AI) on mobile devices creates some security threats.
According to him, AI helpers are usually two types: “on the device” and “in access over the Internet.”
“Online assistants work similarly to versions on basic devices, and in this regard the same safety rules are applied. According to these rules, you should not transfer confidential information to assistants and should not trust their tips in critical situations, since they can give false information, ”said Tushkin.
He noted that the information transmitted to the chat bots can be disclosed in case of hacking the account.
“Therefore, you should protect your accounts using reliable passwords and two -factor authentication,” he advised.
Tushkanov emphasized that the assistants on the device largely solve the issues of confidentiality, since the large language model works directly on the smartphone and does not require access to the Internet.
However, if these assistants use tools that send requests to third -party services (for example, a web post), the threats of confidentiality become the same as with built -in assistants from large brands of smartphones, he said.
According to Tushkanov, the applications supported by AI have the same safety risks as other mobile applications.
“Applications suppliers can use chat data to develop their own models or for other purposes, so it is important to carefully read the application policies of the application,” he said.
Tushkanov also advised to be careful, especially when communicating with virtual friends or chat bots who offer emotional support.
“Our smartphones store the most confidential data. If the application provides access to data such as the calendar, contacts, notifications, messages and photos, then the malicious application can lead to a leakage of these data. This does not depend on whether the artificial application supports the application Intelligence or not. also use safety solutions against potentially malicious programs, ”Tushkin said.