---
title: "AI Has Learned to Steal Money: How Fraudsters Clone Voices and Bypass Biometrics in 2026"
description: "In 2026, fraudsters in Ukraine are increasingly using AI to clone voices and create deepfakes, tailoring deception to a specific individual. We break down the schemes and the signs of danger."
date: 2026-09-01T09:02:03.000Z
lang: en
url: https://xab.info/en/posts/ai-learned-to-steal-money-fraud-schemes-2026
tags: [ai-fraud, voice-cloning, deepfake, cybercrime, ukraine, privatbank]
publisher: "XAB.info"
---

# AI Has Learned to Steal Money: How Fraudsters Clone Voices and Bypass Biometrics in 2026

![Schematic illustration of AI fraud: a hooded hacker at a laptop, a digital neural-network head, a cloned voice call on a smartphone, and stolen cash with credit cards](https://xab.info/media/2026/09/01/ii-navchivsya-vorovat-dengi-shemy-moshennichestva-2026/ii-navchivsya-vorovat-dengi-shemy-moshennichestva-2026-1.webp)

## 🎯 Key Points

- PrivatBank is recording AI-generated voice messages in the name of loved ones asking to transfer money, as well as attempts to bypass biometrics through deepfakes and 3D models.
- In 2025, the number of illegal card transactions decreased by 5% to 256 thousand, but losses grew by 24% to 1.4 billion hryvnia, and the average ticket — by 30% to 5,536 hryvnia.
- The Cyber Police emphasize: the target of the attack is human behavior; AI merely accelerates the creation of deception and its personalization.
- The scale of voice cloning is assessed differently: the Cyber Police speak of isolated cases, while the bank reports regular inquiries.

In 2026, financial fraud in Ukraine has reached a new technological level: criminals are actively using artificial intelligence tools to clone voices, create deepfakes, and automate deception. A familiar voice in a messenger voice message, a convincing "bank employee" on the other end of the line, or a video of a well-known person supposedly advising you to invest may all turn out to be part of a single scheme tailored to a specific individual.

### A Loved One's Voice as a Tool of Deception

This scenario is no longer fictional. Natalia Figol, head of the fraud management department at PrivatBank, notes that the bank is receiving reports from clients who received AI-generated voice messages in messengers from friends, relatives, or colleagues asking to lend or transfer funds. According to her, the scheme begins with collecting material to reproduce the voice: criminals hack an account, process publicly available audio and video recordings with neural networks, and create a message that almost identically reproduces the owner's timbre, intonation, and manner of speech. The fraudsters then gain access to the account and send such messages to contacts, creating artificial urgency.

### Deepfakes and Bypassing Biometrics

PrivatBank has also recorded unsuccessful attempts to bypass biometric checks using AI-generated 3D models and deepfake faces, as well as passing video verification with fake documents created with AI. According to bank representatives, the key change is that familiar fraud techniques have gained new technological tools: audio and video content is produced faster, voices are cloned, correspondence is automated, chatbots imitating support services are configured, and phishing messages and websites are generated on a mass scale.

### The Target Is Human Behavior

It is important to understand: AI itself does not gain access to money. In most schemes, criminals need to convince the victim to perform an action on their own — click a link, enter card details or passwords, install a third-party app, or transfer funds to a specified account. The Cyber Police emphasize that the target of the attackers is not the banking app or service, but human behavior, for which social engineering techniques are used: artificial urgency, psychological pressure, and persuasion to disclose confidential information. AI has not replaced social engineering — it has given fraudsters tools to create deception faster and tailor it more precisely to the victim.

### The Numbers: Fewer Transactions, but Higher Losses

According to data cited by RBC-Ukraine, referencing the NBU, the number of illegal transactions with payment cards in 2025 decreased by 5% — to 256 thousand, while the amount of losses grew by 24% and reached 1.4 billion hryvnia. The average amount of a single fraudulent transaction increased by 30% — to 5,536 hryvnia. This means that attacks are becoming less frequent, but more "targeted" and costlier for victims.

### Contradictory Data

There are two notable inconsistencies here. The first is in the assessment of the scale of voice cloning: the Cyber Police note that they record isolated cases of voice cloning technology being used in fraudulent activities, but emphasize that these do not yet have a mass character, whereas PrivatBank reports regular client inquiries with similar messages. The second is in the dynamics of the statistics: with a 5% decrease in the number of transactions, losses and the average ticket grew by 24% and 30% respectively, which may indicate either a change in attackers' tactics or differences in counting methodology. Both versions are presented openly, and a final assessment of the scale requires additional data.

### How to Recognize the Danger

Recognizing such deception is helped less by the characteristics of the voice itself than by context and behavior: atypical urgency, a request not to call back or verify, pressure on emotions, and a demand to transfer money to a "safe" account or install a third-party app. Banks and law enforcement agencies recommend that in case of any suspicious request to transfer funds, you should break off the correspondence and contact the loved one by a known phone number, as well as not click on links or install apps on the instructions of "bank employees."

## 🔍 Fact-Check Verification

- [AI Has Learned to Steal Money: What New Financial Fraud Schemes Have Emerged in 2026](https://www.rbc.ua/ukr/news/shi-navchivsya-krasti-groshi-ki-novi-shemi-1787924424.html) - Первичный источник: заявления руководителя fraud-менеджмента ПриватБанка и инфографика по данным НБУ (256 тыс. операций, 1,4 млрд грн убытков, средний чек 5536 грн). Оговорка: оценка масштаба клонирования голоса расходится с позицией Киберполиции.
- [Financial Fraud in Armenia: How Fraudsters Steal Money and How to Protect Yourself](https://newsarmenia.am/news/koshelek/finansovoe-moshennichestvo-v-armenii-kak-moshenniki-kradut-dengi-i-kak-zashchititsya/) - Подтверждено по источнику newsarmenia.am
- [AI and Deepfakes Are Used in Every Eighth Successful Fraud Scheme](https://3dnews.ru/1144471/ii-i-dipfeyki-ispolzuyutsya-vkagdoy-vosmoy-uspeshnoy-moshennicheskoy-sheme) - Контекстный источник: подтверждает глобальный тренд применения ИИ/deepfake в успешных мошеннических схемах (примерно каждая восьмая), что согласуется с украинской динамикой роста среднего чека.
- [Deepfake: A Trap for the Trusting](https://www.vedomosti.ru/partner/articles/2026/06/10/1204725-dipfeik-lovushka) - Подтверждено по источнику vedomosti.ru

## ❓ FAQ

### Q: Can AI itself withdraw money from my card?
**A:** No. In most schemes, AI does not gain direct access to money: criminals need to convince the victim to independently click a link, enter card details, install an app, or transfer funds. The central link remains user behavior.

### Q: How do fraudsters create a voice message with a loved one's voice?
**A:** According to PrivatBank's description, they hack an account, process publicly available audio and video recordings with neural networks, and generate a message reproducing the owner's timbre and intonation, after which they send it to contacts with a request to urgently transfer money.

### Q: What signs help recognize AI deception?
**A:** More by context than by the voice itself: atypical urgency, pressure on emotions, a request not to verify or call back, and a demand to transfer funds to a "safe" account or install a third-party app. In case of suspicion, it is worth contacting the person by a known number.