Deepfakes have become a serious threat. The technology is increasingly being used in fraudulent schemes both in Russia and abroad. Attackers use deepfakes for social engineering. Forged voice or video helps fraudsters gain trust in order to steal money. This was stated by Sber Deputy Chairman of the Board Stanislav Kuznetsov at the SPIEF.
Moreover, in the near future, AI technologies will erase language barriers for attackers. A fraudster will speak or write in one language, and his victim will hear and read another online — native to himself. In tandem with deepfakes, this is the most powerful cyber weapon of fraudsters.
A deepfake in a fraudulent scheme works the same way as in any other: it deceives people, but not anti-fraud systems. As Kuznetsov noted, although Sber successfully identifies and blocks operations using deepfakes in real time, not all people and organizations can recognize a fake and protect themselves from deception.
Sber has patented an AI-based technology that effectively detects deepfakes. While the internal working title of the service is "Aletheia", its efficiency currently reaches 98%. We plan to open public access to it.
According to him, it is important that technologies for protection against deepfakes are initially embedded in all digital products and services where they can be used. This is especially true for media, social networks, telecommunications companies and other platforms that distribute video and audio on the Internet.
It is also necessary to introduce responsibility for the use of deepfakes for criminal purposes. People should be protected from deepfakes, and the task of business and the state is to ensure this protection.
Deepfake calls have become a dangerous tool for fraudsters. In 2024, according to VisionLabs, tens of thousands of such attacks were recorded. In 2025, their number may increase significantly.
Read materials on the topic:
Robotized voice and strange requests: how to recognize a deepfake and save your money
Criminal liability may be introduced in Russia for deepfakes
MTS AI system: fighting deepfakes with 97.5% accuracy
