Three Guys Cloned Voices of CEOs With AI, Cost Firms Over $129 Million

U.S. prosecutors have charged three men in a fraud case that they say cost companies in the U.S. and Europe more than $129 million.

The men allegedly used generative AI to impersonate chief executives and lawyers, persuading employees to wire money.

The fraud allegedly ran for about 18 months, starting in July 2024.

In one case, an accountant at a Missouri hospital system received what sounded like a call from the chief executive, who requested an immediate wire transfer and told the employee to keep it confidential.

A caller posing as a lawyer followed up. The hospital system lost $4.8 million, and federal prosecutors in New York say both voices were generated by AI.

Prosecutors say the men also tried to draw employees into video calls using deepfakes of chief executives.

Among the unnamed victims, a U.K. financial technology company lost $11.6 million, a Florida pet food company lost $5 million, and a Massachusetts mutual fund company lost $4.7 million.

Authorities arrested Jacob Soall and Sharon Hazan in Spain, and Guy Sheetrit in the U.K. All three await extradition to the U.S., where they face charges of wire fraud, aggravated identity theft and money laundering.

How the Calls Worked

The approach often began with an email to an accountant or finance manager.

In October 2024, an employee at a New York investment bank was told to sign a nondisclosure agreement.

Once the employee signed, they were asked to call a lawyer. Prosecutors say the defendants assumed the identities of real lawyers, often at some of the world’s largest firms.

Prosecutors say Soall kept a handwritten script that invoked the risk of insider trading to persuade employees to keep the conversations from colleagues.

The men allegedly directed employees to wire money to accounts controlled by conspirators in several countries, including Singapore, Spain and the United Arab Emirates.

Prosecutors say the group then converted the proceeds into cryptocurrency to obscure their origin and ownership.

Hany Farid, chief science officer at GetReal Security, said a short recording can now be enough to reproduce a voice.

“I would comfortably say with less than a minute, you can clone somebody,” he said.

The FBI logged more than 22,000 AI-related complaints totaling over $890 million in losses in 2025.

The agency said generative AI was used in schemes ranging from voice-cloning scams involving supposed family emergencies to phishing and romance fraud.