🎭 Scammers Can Now Look and Sound Like Someone You Trust
Your son calls you. His voice is familiar, but he's crying and speaking quickly:
"Mom, I've been in an accident. I need money right now!"
Or your boss invites you to a video conference. You recognize their face, hear their voice, and see familiar colleagues. After the meeting, you're instructed to transfer money.
Or a buyer sends you a bank receipt: "I've already paid. Please ship the item."
In all three situations, what you're seeing or hearing could be a convincing fake.
This isn't science fiction. With generative artificial intelligence becoming widely accessible, criminals can now create convincing material that previously required specialized skills, expensive equipment, and considerable time.
Even more troubling, some of these scams have already caused multimillion-dollar losses.
💸 1. A Video Meeting With a Boss Who Wasn't Really There
In January 2024, an employee of the international engineering company Arup took part in a video conference in Hong Kong.
The company's chief financial officer and other familiar colleagues appeared on screen.
The meeting was convincing enough for the employee to follow instructions to transfer money.
The result: 15 transactions totaling approximately US$25 million.
Investigators later established that the criminals had used artificially generated images and voices impersonating real employees. According to police, the attack involved previously prepared video material.
The employee trusted what they saw and heard.
That's what makes this case so disturbing: scammers no longer have to impersonate an executive using nothing more than a fake email.
They can create a digital imitation of that person.
📞 2. A Familiar Voice Is No Longer Proof of Identity
Generative AI can reproduce a person's vocal tone, intonation, and speaking style using samples of their voice.
Those samples may come from publicly available videos, social media, or other accessible recordings.
Criminals then combine the technology with a familiar psychological tactic: an accident, a hospital emergency, police detention, or an urgent demand for money.
But now the victim doesn't hear an unfamiliar scammer.
They hear a voice that sounds like their own child.
And this is no longer just a hypothetical threat.
In February 2025, scammers in Italy used an AI-generated imitation of Defence Minister Guido Crosetto's voice to deceive prominent businesspeople.
Massimo Moratti, the former owner of football club Inter Milan, transferred nearly €1 million before the deception was uncovered. Italian authorities subsequently froze the transferred funds.
If criminals could convincingly imitate a government minister's voice, imagine how persuasive a call supposedly coming from someone you know might sound.
🧾 3. Fake Bank Receipts, Messages, and Websites
Imagine you're selling a laptop online.
A buyer sends you a screenshot of a banking app. It shows the amount, date, recipient, and confirmation of a successful transfer.
Everything looks genuine.
But the money never reaches your account.
In May 2025, British police warned about scams involving fake banking applications. Criminals were showing sellers screens that appeared to confirm successful payments, even though no money had been transferred.
In this particular case, the use of AI was not confirmed.
However, modern image-generation tools make it much easier to produce convincing visual forgeries, including:
- Screenshots of bank transfers and account balances.
- Conversations supposedly taken from Telegram, WhatsApp, and other messaging platforms.
- Images of documents and identification cards.
- Pages resembling online stores and investment platforms.
- Fake order confirmations and delivery notifications.
For someone accustomed to trusting screenshots, a convincing forgery can look like genuine evidence.
But an image of a bank transfer is not proof that you've received payment.
Always verify incoming payments directly through your own bank or payment provider.
That's a small detail that can make an enormous difference.
🕵️ 4. A Message From a Scammer Who Knows Too Much About You
In the past, phishing emails often revealed themselves through spelling mistakes, awkward wording, or unnatural translations.
Today, language models can produce fluent, personalized messages in numerous languages.
Scammers can use publicly available information about your job, friends, and interests to make their messages resemble ordinary business correspondence.
For example, an employee receives an email supposedly from a manager.
It mentions the correct project name and identifies real colleagues.
Or someone receives a message from an "old friend" referring to actual events in their life.
In 2026, the FBI also warned about scammers impersonating law enforcement officials. These schemes involved fraudulent websites and AI-generated videos.
Particularly troubling was the targeting of people who had already lost money to fraud and were now looking for help.
Even a website resembling an official government service, or a video featuring someone in uniform, no longer guarantees authenticity.
🤖 5. Scammers Can Communicate With Large Numbers of People at Once
Another major change is automation.
Modern speech-recognition systems, language models, and voice synthesizers can be combined to create conversational systems capable of responding to people in real time.
Criminals may use these tools for initial contact, identifying potential victims, and preparing personalized responses.
Instead of sending the same generic message to thousands of people, they can generate individual conversations.
Such methods are particularly concerning in investment scams and romance fraud.
A victim may be shown an attractive profile, told a convincing personal story, and gradually encouraged to trust the person behind it.
In some schemes, automated systems can support parts of the interaction, while human operators become involved later.
AI didn't invent these scams.
It has made them faster, cheaper, and potentially much more convincing.
🔍 The Devil Is in the Details: How Can You Spot an AI Fake?
Generative AI still makes mistakes. Sometimes those mistakes provide useful clues.
But no single visual or audio feature can reliably determine whether something was created by artificial intelligence.
🖼️ Images, Receipts, and Screenshots
Look carefully for:
- Unnaturally smooth, almost plastic-looking skin.
- Patchy colors or strange transitions between shades in otherwise uniform areas.
- Blurred edges around hair, clothing, or objects.
- Shadows and reflections that don't match.
- Incorrect characters, inconsistent fonts, or unusually positioned interface elements.
- Contradictions involving dates, payment amounts, or other document details.
Pay particular attention to screenshots of banking applications and payment confirmations.
However, modern image generators are becoming increasingly good at reproducing text, lighting, and interface details.
The absence of visible mistakes does not prove an image is authentic.
🎙️ Voices and Phone Calls
Some synthetic voices may contain:
- Metallic or robotic tones.
- Unusual pauses.
- An unnaturally consistent speaking rhythm.
- Strange transitions between words.
- Unusual breathing patterns or repeated audio artifacts.
But poor cellular reception and audio compression can produce similar effects.
Conversely, a high-quality synthetic voice may sound completely natural.
Recognizing someone's voice is no longer enough to verify their identity.
🎥 Video Calls
Deepfake videos sometimes display:
- Slight differences between lip movements and spoken words.
- Blurring around the face or hair.
- Unnatural changes in facial expressions.
- Brief distortions when someone turns their head.
- Strange changes in lighting or facial contours.
Some deepfakes may also produce visual distortions when part of the person's face is covered by a hand.
But asking someone to turn sideways or touch their face is no longer a reliable verification method.
Instead of relying on visual imperfections, independently confirm who you're actually speaking to.
🎵 If AI Can Create Chart-Topping Music, How Convincingly Can It Fake a Voice?
Here's a fact that demonstrates how quickly these technologies are evolving.
In November 2025, Walk My Walk, a song by the AI-based virtual country music project Breaking Rust, reached number one on the US Billboard Country Digital Song Sales chart.
This was the country digital song sales chart, not the overall Billboard Hot 100.
Still, the achievement is significant.
Artificial intelligence can already contribute to music that attracts millions of streams and enters established industry charts.
Meanwhile, technologies for generating text, images, video, and voices are available far beyond major recording studios.
Tools such as ElevenLabs for speech synthesis, image generators, video-generation services, and large language models are used for many legitimate creative and professional purposes.
But criminals can exploit similar capabilities.
And if a synthetic voice can convincingly perform a song, the familiar sound of someone's voice during a phone call can no longer be treated as sufficient proof of identity.
🛡️ How Can You Protect Yourself and Your Family?
One of the most important habits in the age of generative AI is to verify people's identities and requests independently of the images, voices, or documents presented to you.
If a family member calls urgently asking for money: End the call and contact them yourself using a number you already know. A private family code phrase can provide an additional safeguard, but it shouldn't be your only method of verification.
If your boss demands an urgent payment: Confirm the request through another established work channel. For significant transfers, require approval from more than one authorized employee.
If a buyer sends you a receipt: Check that the money has actually arrived in your account. Never release goods solely because someone sends you a screenshot.
If someone sends you an official-looking website or document: Navigate to the organization's website independently and verify the information through its official channels.
If someone pressures you, frightens you, or insists that everything must remain secret: Treat that as a serious warning sign. Stop the conversation and verify what's happening.
💚 What Should We Expect as AI Becomes More Powerful?
Online fraud isn't going to disappear. But the methods used to deceive people are becoming increasingly convincing.
Fake images will contain fewer obvious visual defects.
Synthetic voices will sound more natural.
Digital characters will be able to sustain increasingly complex conversations.
And fabricated documents and websites will become harder to distinguish from genuine ones.
The familiar idea of "I saw it with my own eyes and heard it with my own ears" can no longer be considered sufficient grounds for making a financial decision.
We need to develop a different habit:
Don't judge information solely by how convincing it looks or sounds. Ask whether you can independently verify where it came from.
That habit may become one of the most important defenses against fraud in the age of artificial intelligence.
🔎 About Axon OSINT
Axon OSINT is a platform for digital investigations, open-source intelligence analysis, and organizing information related to potential fraud.
If you encounter suspicious calls, forged documents, fraudulent websites, or attempts to obtain your money through deception, preserve any messages, images, links, and other available evidence.
If needed, you can report an incident through Axon OSINT.
Reporting an incident through the platform does not replace contacting your bank or law enforcement when necessary.