A work of art: deepfakes for attacks on businesses have fallen in price by 100 times
The creation of deepfakes for attacks on companies has become about 100 times cheaper over the past two years, Izvestia found out. If it used to cost several hundred dollars to make a convincing fake video per minute, now it can be generated for $5, and sometimes cheaper. Therefore, this type of fraud is becoming an increasingly serious threat to businesses and banks. In the first nine months of 2026, 900 unique imitations were identified, although there were a third fewer of them last year. How to protect yourself from attacks is in the Izvestia article.
Why creating deepfakes has become cheaper
Attacks using deepfakes — fakes of videos and voices of real people — are becoming an increasingly tangible threat to Russian businesses and banks. The main reason is a sharp reduction in the cost of their generation due to the widespread development of neural networks. If earlier it could cost more than $500 to create a high-quality fake, depending on the complexity, now a separate deepfake can be generated for just $5 or even cheaper, said Grigory Osipov, an independent expert on cryptocurrency investigations, the author of the Digital Cipollino channel.
The trend was also confirmed by the CEO of Phishman, a cybersecurity expert at the Institute of Growth Economics. Stolypin Alexey Gorelkin. According to him, the entry threshold has decreased — now neural networks are taking over the technical part of the work, and the fraudster has to choose a sample and a script.
According to Marina Pavlova, founder of the legal aid agency Juno, the production of deepfakes has fallen in price more than 100 times in two years. In addition, tens of thousands of ready-made media templates are available in the shadow segment of the Internet. The cost of using them may be even lower.
Previously, video creation required skills working with editing and rendering programs, as well as powerful computers, Grigory Osipov added. Now, several legal services — for voice cloning, digital avatar creation, and image processing — can be combined into one chain. Ready-made digital identities have even appeared on shady sites, which are rented out for fraudulent schemes, the expert added.
In some cases, attackers don't even need to achieve perfect quality. A short voice message or video call may be enough to convince a person to fulfill an urgent request, especially if the scammer knows the context of the conversation. Sometimes a ready-made fake using a template can be made for about 50 rubles, said Igor Bederov, Chairman of the Council for Combating Technological Offenses of the Constitutional Court of the National Security Service of Russia.
In 2025, VisionLabs (which develops solutions based on computer vision, machine learning, and artificial intelligence) discovered 627 unique deepfakes on the Russian Internet, and in January-September 2026, it already had about 900. At the same time, attackers can use the same material multiple times, so the number of unique fakes does not reflect the total number of attacks, explained Dmitry Markov, CEO of the company. According to his estimates, we can talk about tens of thousands of fraudulent episodes per year.
According to Sberbank, over the past year and a half, the number of incidents involving deepfakes has increased 26-fold, and financial losses from AI schemes have exceeded 2 billion rubles since the beginning of this year.
The Interior Ministry has other figures. In just five months of this year, about 4 thousand victims of such schemes were identified, and the damage reached 21 billion rubles. Izvestia sent a request to the ministry to assess the number of attacks and damage caused by them for November 2026.
How deepfakes are used against businesses
Fakes are a handy social engineering tool. It is not necessary for attackers to hack into a banking system or gain direct access to an account: sometimes it is enough to convince an employee who already has the authority to transfer money on their own.
In autumn, the media reported that scammers in Italy, using fake voices and videos of the heads of three banks, stole a total of 120 million euros. At the same time, standard insurance in this country does not cover damage from deepfakes, they are designed only for hacking. In Russia, the situation is different — most policies against cyber attacks already include protection against social engineering, and it does not matter whether a fake image was used to deceive an employee, the All-Russian Union of Insurers told Izvestia.
Schemes can be directed against bank customers. Fraudsters pose as representatives of a financial organization, managers or business partners in order to encourage a person to take the necessary action, said the Associate professor of Finance and Sustainable Development at the Russian University of Economics. Plekhanova Maria Ermilova.
At the same time, not only video and audio are used, but also fake employee accounts, for example, in shared chats with secretaries and accountants, said Vladimir Zykov, director of the Association of Professional Users of Social Networks and Messengers, director of projects at ANO Digital Platforms.
According to Svyatoslav Makarov, an analyst at the Speakatel Cyber Threat Monitoring Center, deepfake attacks account for 3% of incidents in financial and telecommunications companies. In addition, users are one and a half to three times more likely to click on links in phishing emails if they are created using AI. Neural networks allow you to make messages more convincing and personalize them for a specific person.
New AI models accelerate cyber attacks
New models of artificial intelligence help criminals to look for vulnerabilities in software and prepare tools for their exploitation.
In particular, Anthropic reported on the capabilities of the Claude Mythos Preview model to find vulnerabilities in software and build chains of their exploitation. OpenAI also described the capabilities of the Astra model to detect previously unknown errors and develop ways to use them. The development of such tools increases the risks for organizations that are slow to update systems, warned Freedom Global analyst Vladimir Chernov.
According to DLBI founder Ashot Oganesyan, banks and companies that use services accessible from an external network and do not control employee accounts are at risk.
At the same time, AI is also used to protect against attacks. According to Anton Redko, head of the Secure Development group at Krosstech, in 2026 the number of attempts to analyze vulnerabilities using AI agents increased three to five times. Banks are increasingly using automated scanners and testing their infrastructure, but reaction speed remains a critical factor.
How can a business protect itself from voice and video tampering
The main problem is that deepfake is not necessarily accompanied by technical hacking. If the employee confirms the payment himself or transmits confidential information, standard security systems may not detect the violation, experts warned.
Technical solutions for detecting fakes are already appearing on the market, said Kontur, an information security expert.Aegis" and Staffcop by Yuri Drachenin. However, the detection of synthetic images or sound should complement, and not replace, the verification of credentials and procedures for approving operations.
First of all, you should not use voice, video, email, or messenger as the only basis for conducting mission-critical operations. If the manager urgently requests to transfer a large amount or change banking details, the request must be confirmed through an independent channel — for example, to call back at a pre-determined number, said Evgeny Romanov, a leading analyst at Expert RA. For large payments, additional coordination and separation of powers between employees are also required.
The principle of zero trust is also important: even a familiar account or recognizable voice should not automatically confirm the identity of the interlocutor. Deepfake detection systems can become an additional filter, but you can't rely on them alone — fake generators are constantly being improved, said Vladimir Ulyanov, head of the Zecurion analytical center.
The problem is particularly acute for small and medium-sized businesses: such organizations often do not have their own specialists and resources to constantly monitor the infrastructure. Therefore, they need access to professional cybersecurity services and clear internal rules for working with payments and data, Denis Frenchov, Vice President of the Chamber of Commerce and Industry, added.
In the coming year, the number of attacks using deepfakes is likely to continue to grow, experts warn. According to the Rodin founder.Alexey Rodin's Capital, their number may increase by 10-15%.
Переведено сервисом «Яндекс Переводчик»
