AI impersonation fraud is no longer limited to a suspicious voice message asking somebody to send a gift card. US prosecutors have charged three men over an alleged scheme in which criminals posed as company bosses and international lawyers, using synthetic voices or images to persuade employees to approve large transfers. The Justice Department says victim businesses lost more than $129 million. The allegation is serious, but it remains a criminal case: charges are not convictions, and the defendants are presumed innocent.
The US Attorney’s Office announcement, issued on 5 November 2026 and updated the next day, describes a deception built around ordinary corporate routines. A supposed chief executive would tell a member of staff that a confidential transaction required urgent help. The employee would then be introduced to someone claiming to be a lawyer at a recognised firm. That lawyer, also allegedly an impersonator, supplied instructions to send the money. Every step was intended to make the next one look more legitimate.
How the alleged AI impersonation fraud worked
According to prosecutors, the alleged operation ran between 2024 and 2026 and targeted businesses in different countries. The imitation could involve a senior executive’s voice, an image or both. An employee receiving the call might think they were speaking to a familiar boss, then regard a second call with a seemingly independent law firm as confirmation. In reality, investigators allege the same criminal network controlled the story and the account receiving the funds.
The sums in the indictment show how expensive a single deception can become. The Justice Department cites individual transfers from organisations including a UK financial technology business, US healthcare providers and other companies. It also alleges that money was moved into accounts across several countries and converted into cryptocurrency to obscure where the proceeds had gone. These details come from the prosecution’s account of the case; a court will ultimately decide what has been proved.
The clever trick was not only the fake voice
A striking feature of the alleged method is the choreography. The caller did not simply demand money in a badly written email. The supposed boss introduced a legitimate-sounding business reason, the need for secrecy and a second professional authority. Prosecutors say one accused person even kept a handwritten script for impersonating a lawyer. The technology could make the exchange more believable, but the manipulation relied on familiar pressures: seniority, urgency and the feeling that someone else had already checked the decision.
That is why training staff only to recognise an unnatural voice will not solve the problem. Some fraudulent calls may sound odd; others may be convincing or use genuine recordings mixed with fabricated material. A reassuring screen image, a familiar name or a polished legal vocabulary is not proof of identity. Previous LiveAIWire coverage of deepfake caller checks points to the more dependable principle: verification should use a channel that the caller cannot choose or control.
A second approval channel is worth more than a perfect ear
The FBI’s earlier guidance on business email compromise recommends confirming unusual payment requests through a secondary channel and checking changes to bank details independently. That might mean telephoning a director on a known number saved in company records, or asking the bank to verify the destination through an established process. The key is that the staff member chooses the checking method, not the person asking for the transfer.
A business could also separate the authority to initiate a transfer from the authority to approve it, particularly for unexpected or high-value payments. That is not an accusation against genuine colleagues. It is recognition that the same process ought to protect everybody, including the managing director whose identity may have been copied. A caller’s request for secrecy should never override the organisation’s payment-control rules. If someone really is arranging a confidential transaction, an independent identity check can still happen without broadcasting the details.
What to do when something feels wrong
The first useful move is to interrupt the transaction, not debate with the caller about whether their voice is real. Ask for the request to be sent through the company’s normal payment process and independently contact the appropriate person. Treat any sudden change in beneficiary details, refusal to allow a callback or insistence that ordinary authorisations be bypassed as a reason to stop. Where money has already left, the FBI advises contacting the financial institution immediately to seek recovery and reporting the incident without delay.
The problem extends beyond finance departments. AI voice-cloning scams targeting older people use a similar appeal to trust, only the supposed authority may be a relative rather than a chief executive. In both settings, the safest question is not whether the voice sounds right. It is whether the request has been verified independently and whether the proposed payment makes sense under normal circumstances.
The lesson from a case still before the courts
The alleged $129 million scheme has not been proved against the charged individuals, and it would be wrong to write about their guilt as settled fact. The Justice Department’s description nevertheless identifies a clear vulnerability in the way organisations trust voices, faces and professional titles. The emerging risk is that those signals can be reproduced at scale while a human employee is still expected to make a quick, consequential decision.
The most effective defence is deliberately unglamorous. A company that insists on known telephone numbers, normal approval chains and an independent check before transferring money removes much of the alleged scheme’s leverage. The fake boss can sound exactly like the real boss and the fake solicitor can know the right jargon. Neither should be able to authorise a bank transfer merely by sounding convincing.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, cybersecurity and the social impact of emerging technology. LiveAIWire is an independent, human-led technology publication using AI-assisted research, editorial production and original AI-assisted editorial illustrations under his direction.