AI Clones of CEOs Stole $129 Million: Inside the FBI Case That Shows How Voice Fraud Went Corporate

An accountant at a hospital system in Missouri picked up the phone and heard a voice she knew well. It sounded like her organization's chief executive, asking for an urgent and confidential wire transfer. She sent the money. According to federal prosecutors, the voice was not the CEO at all. It was a clone built with AI, and the hospital lost about $4.8 million in a single transaction.

That call was one small piece of a much larger operation. On November 5, 2026, federal prosecutors in Manhattan unsealed an indictment charging three men with using AI to impersonate corporate executives and attorneys, tricking employees into wiring money to accounts the group controlled. The government says the scheme ran about a year and a half and took over $129 million.

It is the largest AI impersonation fraud prosecution the Justice Department has brought so far, and a warning to every finance team: a voice that sounds right and a face that looks right prove nothing on their own.

Who was charged and how the scheme worked

The three defendants are Jacob Soall, 41, of Spain and the UK; Sharon Hazan, 54, of Israel; and Guy Sheetrit, 42, of Israel and the UAE. They were arrested in Europe between June and September 2026, and extradition requests were pending when the indictment was unsealed. The indictment contains seven counts: conspiracy to commit wire fraud, wire fraud, two counts of aggravated identity theft, and three counts of money laundering.

Prosecutors allege the operation ran from about July 2024 to February 2026, targeting thousands of banks, financial technology firms, hospitals, and other businesses in the United States and other countries. The group allegedly used AI tools to mimic the voices and, in some cases, the faces of chief executives and attorneys at major law firms.

An employee who handled payments would get a call or a video call from someone who sounded and looked like the boss, backed up by a message that appeared to come from a real lawyer at a real firm. The request was always urgent and confidential, and the money would move through accounts the group controlled, in many cases converted into cryptocurrency.

The six losses prosecutors spelled out

The indictment lists six victim companies whose losses add up to about $31.9 million. A UK financial technology company lost about $11.6 million, the largest single loss in the list. A pet food company in Florida lost about $5.2 million, the Missouri hospital system about $4.83 million, and a mutual fund company in Massachusetts about $4.6 million. A hospital system in California lost about $3 million, and a bank headquartered in Rhode Island about $2.68 million.

Two of those payments landed on the same day in November 2024, handled by the same supposed lawyer from the same firm: the Missouri hospital wired $4,829,580 and the Rhode Island bank wired $2,684,580. That is a playbook, not a series of one-off scams. The FBI did not name the targeted companies, but it said the scheme reached thousands of businesses.

Why this case is different from the scams that came before

Business email compromise is not new: criminals have long sent fake invoices and spoofed emails pretending to be a supplier or a senior manager. What changed, according to prosecutors, is the quality of the impersonation. A spoofed email can be checked, but a voice on a live call that sounds exactly like your CEO is much harder to doubt.

The indictment does not say how the voices were generated, or how much of the $129 million has been recovered. What is clear is that the tools are no longer exotic: voice cloning now works from short clips, some pulled from public earnings calls and videos posted online, and video deepfakes have moved from research labs to consumer apps. The Justice Department announced its Super Intelligence Task Force the same day, a unit created to set enforcement priorities for the criminal misuse of advanced AI.

What companies should take from the indictment

The lesson is not about the technology but about the process that failed: every transfer went through because a single employee believed a single voice. The fix is a rule that no large payment moves on a voice instruction alone, no matter how familiar that voice sounds.

A call-back procedure, where the employee hangs up and calls the executive back on a number they already have, would have stopped these transfers, as would a second approval or extra scrutiny for any request marked urgent and confidential. If a boss calls from an unfamiliar number, pause. If the request involves a new bank account or a first-time vendor, verify through a separate channel. If the caller pressures you not to check with anyone else, that is the clearest warning sign of all. None of this requires new software. It requires a habit: treat a voice as a claim, not as proof.

The charges are allegations, and the story is not over

The charges are allegations, and the three defendants are presumed innocent unless and until proven guilty in court. Extradition proceedings were still pending when prosecutors announced the case.

AI impersonation has moved from a curiosity to a tool of organized fraud. A hospital system that loses nearly $5 million to a fake phone call has less money for patients, and the Justice Department has signaled that more cases are coming.

Frequently Asked Questions

How did the alleged $129 million AI impersonation scheme work?

According to prosecutors, the defendants used AI tools to clone the voices and, in some cases, the faces of executives and attorneys at major law firms, then talked employees who handled payments into wiring money to accounts the group controlled. The scheme reportedly ran from July 2024 to February 2026.

Who are the three men charged in the case?

Jacob Soall, 41, Sharon Hazan, 54, and Guy Sheetrit, 42, were charged in the Southern District of New York with wire fraud, aggravated identity theft, and money laundering. All three were arrested in Europe between June and September 2026, and extradition requests were pending when the indictment was unsealed.

How can a company protect itself from AI voice cloning fraud?

The best defense is a rule that no large transfer moves on a voice instruction alone. Employees should hang up and call the executive back on a number they already have, get a second approval for unusual payments, and treat urgency and secrecy as reasons for extra checks. Unfamiliar numbers, new bank accounts, and pressure not to verify are all warning signs.