Book an intro call NL

Blog › AI CEO fraud: charges over $129 million

AI CEO fraud: charges over $129 million

Published

Suspects allegedly used AI to imitate executives’ voices and images. Learn how this CEO fraud works and how to check payment requests.

A payment request appears to come from your director. During the call, you even recognise their voice or face. Yet you could be speaking to a scammer. According to the US Department of Justice, suspects used artificial intelligence to impersonate executives and lawyers. They allegedly defrauded companies of approximately $129 million.

The Record reported on the charges against three men on 6 November 2026. The case shows how CEO fraud changes when criminals can also imitate a voice or video image. For employees, this means a familiar face during a call is no longer reliable proof of identity.

What allegedly happened

The US Department of Justice has charged Jacob Soall, Sharon Hazan and Guy Sheetrit. They allegedly approached company employees while posing as their CEO or a lawyer. Their aim was to get employees to transfer large sums to bank accounts the suspects controlled.

According to prosecutors, the operation ran from 2024 to 2026. The suspects allegedly worked with others to use AI to imitate real executives’ voices and images. In several incidents, employees were told their help was needed with a confidential financial transaction.

Employees were then directed to lawyers at a large international firm who were supposedly working on the transaction. According to the charges, the suspects also used AI to impersonate those lawyers. Soall allegedly usually played the lawyer and used a script for the role. Hazan was allegedly responsible for laundering the money.

Losses at several organisations

The charges include a hospital in Missouri that lost nearly $5 million. A technology company in the UK allegedly lost $11.5 million. A bank in Rhode Island and a pet food company in Florida are also named as victims.

According to prosecutors, the money went to accounts in the United States, Europe, the United Arab Emirates and China. The criminals then allegedly converted it into cryptocurrency.

Soall and Hazan were arrested in Spain during the summer. Sheetrit was detained in the UK in September. At the time of the report, extradition requests were still pending. These are allegations in a criminal case, not convictions of these suspects for these offences.

How CEO fraud using an imitated voice or video works

In CEO fraud, an attacker impersonates someone with authority within an organisation. They might send a message or speak to an employee during a phone or video call. The employee is then asked to make a payment they would normally check.

AI can help imitate a voice or face. This is often called a deepfake: artificially generated or altered audio or imagery designed to appear genuine. The article does not describe which tools the suspects used. It does state that, according to prosecutors, they imitated real people’s voices and images.

The request comes from someone you think you know

An instruction from a director naturally gets attention. Employees want to do their job well and generally assume that an internal request is genuine. A familiar voice can reinforce that trust.

This is where the deception works. The employee assesses not just the request itself, but also who appears to be making it. If the attacker imitates that person convincingly, an unusual payment request can still sound plausible.

Being deceived this way does not mean someone was careless. The attacker is deliberately exploiting normal working habits: following instructions and cooperating with a manager.

Confidentiality makes checks harder

In this case, employees were allegedly asked to help with a confidential financial transaction. That explanation might raise questions, but it can also seem like a reason to keep colleagues out of the conversation.

Business transactions can legitimately be confidential. The risk arises when that confidentiality is also used to bypass normal payment checks. An employee may then feel that asking questions would undermine the task.

A supposed lawyer backs up the story

Referring the employee to a lawyer adds what appears to be independent confirmation. The employee speaks not only to the supposed director, but also to someone who seems to be handling the legal side.

Yet the same attackers may control both contacts. The second person proves nothing. They simply repeat the same invented story from a different role.

The payment goes to an attacker’s account

Ultimately, the employee has to transfer real money. The attacker does not need direct access to the company’s banking system. The authorised employee makes the payment.

That distinction makes this fraud difficult for organisations that rely mainly on technical security. A well-protected account cannot prevent its legitimate user from approving a payment under false pretences.

What this means for small and medium-sized businesses

The losses mentioned relate to the organisations named in the charges. The article does not say how often this approach affects Dutch small and medium-sized businesses. The mechanism can still be relevant to a smaller business where employees receive instructions directly from the owner.

If the same employee prepares and makes payments, there may be little scope for an independent check. A personal request from the owner may also carry more weight than an established procedure. Particularly if the employee believes the owner is on the phone.

The damage goes beyond the money transferred

After a fraudulent payment, an organisation needs to establish what happened. Someone must contact the bank, and the employees involved need to reconstruct events. That takes time away from their normal work.

The loss can also affect the funds available to the business. Payments to suppliers may come under pressure. Whether that happens depends on the amount lost and the company’s financial position.

Trust between colleagues can suffer too. An employee may feel ashamed, even though the attacker deliberately exploited their willingness to help. Blaming people afterwards makes it harder for them to report the next suspicious request early.

Payment procedures must apply to the director too

A payment procedure only helps if employees are allowed to follow it when a request comes from the most senior person in the organisation. An exception authorised by the director is exactly the opening an impersonator can exploit.

Make it clear that verification is part of the task. An employee who calls back or puts a payment on hold should not need permission from the person making the suspicious request.

How to recognise AI-assisted CEO fraud

You do not need to be an expert in artificial imagery or audio to recognise a suspicious request. Focus on what the sender wants you to do and which checks they want you to skip.

  • An unusual payment request from a manager. The instruction does not follow the normal payment process, even if the voice sounds familiar.
  • Confidentiality that prevents consultation. You are told not to discuss the request with the colleague who normally checks payments.
  • An unfamiliar lawyer who backs up the story. Their contact details come only from the person requesting the payment.
  • An unexpected account or payee. The recipient does not match your records or the underlying agreement.
  • Resistance to a callback. The person tries to stop you contacting them on a known number you have looked up yourself.
  • Pressure to bypass the payment procedure. The sender wants you to act without the usual approval.

An odd voice or an unnaturally moving face can be a reason to stop. Equally, a smooth conversation does not prove everything is genuine. Even without visible signs of manipulation, an exceptional payment request must be open to verification.

What employees and organisations can do

The most useful check happens outside the suspicious conversation. Do not let the sender decide which channel you use to confirm their identity.

Actions for employees

  1. Put the payment on hold. Do not act on the request while its origin or business purpose remains unclear. Do not start preparing the payment on the assumption that confirmation will follow.
  1. Call the manager on a known number. Use a number from your own contacts or internal records. Do not use a number supplied with the request.
  1. Check the lawyer independently. Find the law firm’s contact details yourself. Use that channel to ask whether the person works there and is genuinely involved in the transaction.
  1. Get approval through the normal payment procedure. Give the person checking the payment the original messages and payment details too. Simply saying that the director has approved it does not provide an independent check.
  1. Report the request internally. Use your agreed reporting channel or contact the responsible colleague. Keep the messages and note when the conversation took place. Do not delete anything until it is clear what the investigation needs.

Actions for the organisation

  • Set a callback procedure for unusual payment requests. Agree which known contact details employees should use.
  • Separate payment requests from payment approval. Have the approver independently check the business reason and the payee.
  • Set rules for confidential transactions. Decide in advance who within the organisation can verify these requests, so confidentiality does not become a reason to skip checks.
  • Practise handling requests that appear to come from a manager. Afterwards, discuss which action helped the employee verify the person’s identity.
  • Document what to do after a payment has been made. Have an employee contact the bank immediately through a known channel and report the incident internally. Ask whether the payment can still be stopped or recovered. Recovery is not guaranteed.

How Lumyo Awareness helps

With Lumyo Awareness, we recreate this kind of deception in a phishing simulation. Employees practise handling requests that appear to come from someone they trust. In awareness training, they learn to recognise the signs and practise checking a request through an independent channel.