Back to Articles
Industry Insights

$129 Million AI Impersonation Case: Hospitals, a Bank and a Pet Food Maker Lost Millions to Fake CEOs and Fake Lawyers

By DeFake ResearchNovember 7, 20267 min read
Companies that move millions have safety procedures. According to federal charges, this scheme had an answer for each one. Here is how a $129 million AI impersonation scheme allegedly walked authorized finance employees past the controls meant to stop them.

On November 5, 2026, the U.S. Attorney's Office in Manhattan unsealed an indictment charging three men, Jacob Soall, Sharon Hazan and Guy Sheetrit, with a scheme that allegedly took at least $129 million from companies around the world between July 2024 and February 2026. According to the indictment, the group used AI to copy the voices, and at times the faces, of real CEOs, and posed as lawyers at some of the world's largest law firms.

According to the charges, no systems were hacked and no passwords were stolen. Instead, employees who were authorized to send money were convinced to send it.

This article walks through the scheme the way a finance team would experience it, and shows how each common safety procedure could be turned around.

Who They Targeted

The indictment says the group "typically targeted" people "who worked in accounting roles" and had "sufficient authority" to approve a large wire transfer. They did not need to break into the payment system. They needed the person who already had the keys.

The Playbook, Step by Step

  • Step 1: An e-signature email from "the CEO." According to the indictment, the approach usually began with an email sent through an electronic document-signature service, such as DocuSign or a similar service (DocuSign is the example the indictment names), made to look like it came from the company's own CEO. One real example, sent to a major New York investment bank in October 2024:
  • "We need to loop you into a deal we've been working on. The companies involved require you to sign an NDA - see attached. Once you've signed it, could you please call [the attorney], our legal representative at [the law firm] on [phone number]. He will assist you in scheduling a compliance call for later today, and will provide further details."

  • Step 2: A call with "the CEO." The group then tried to get the employee on a phone or video call, where AI was used to mimic the CEO's voice and, at times, the CEO's face. The fake CEO told the employee to work with the lawyer and to send the wire the lawyer would request.
  • Step 3: "The lawyer." The lawyers carried the names of real attorneys at real firms, often among the 50 largest law firms in the world and usually in their European offices. The "lawyer" explained that the company was working on an important transaction, often an acquisition, and that the employee had to keep it secret from colleagues.
  • Step 4: The wire. The "lawyer" directed the employee to wire a large sum to a bank account the group controlled. The money went to accounts in the United States, Spain, Portugal, Greece, the United Arab Emirates, Singapore, Austria, Slovakia and China, and was then converted into cryptocurrency.
  • Every Control, and How It Was Turned Around

    Companies that move millions of dollars have safety procedures. The victims here included two hospital systems, a bank and a mutual fund company, organizations expected to have real controls. The indictment does not describe each company's procedures, but based on the scheme it describes, this is how common controls can be turned around:

  • Email security blocks fake senders. The request came through a legitimate document-signature service, from that service's own servers. It looks like any routine signature request, so email filters have little reason to stop it.
  • Only authorized people can approve a wire. The scheme targeted the authorized person directly. The control worked exactly as designed: the right person sent the payment.
  • Call back to verify. The first email supplied the lawyer's phone number. From that moment, any "verification" call to that number would go back to the scammers. And the CEO call itself felt like verification: you heard your CEO's voice and, sometimes, saw their face.
  • Check who the counterparty is. The lawyers used the names of real attorneys at real firms. Search the name and it checks out.
  • Escalate to legal or compliance. The secrecy was presented as a legal duty. Telling anyone felt like breaking the rules, not following them.
  • Ask a colleague. The script had an answer ready for "who else knows about this?"
  • Review new payees. In an acquisition, paying a new account at a law firm or an escrow agent looks normal.
  • Bank fraud monitoring. A real employee, using their own login, sending a payment they were authorized to send. There is little for a bank to flag.
  • Why a Signature Request Proves Nothing About the Sender

    E-signature services in general are built to protect the person sending a document. They prove a document wasn't changed after signing and keep a record of who signed it and when. Identity checks, such as ID verification or SMS codes, are there to verify the signer, and the sender decides whether to use them. That is how these services are designed to work.

    In this scheme, the scammers were the sender. The checks were never pointed at them, and the employee's signature was genuine. A sender's display name is simply what the account holder enters, so a signature request shows that a document was sent and signed, not who really sent it.

    The signature step wasn't there to prove anything. It made the request look formal, and signing an NDA committed the employee to secrecy before money was ever mentioned. Before signing, check the sender's actual email address in the request details, and whether it matches your company's own domain.

    The Script: Insider-Trading Rules as the Cover Story

    According to prosecutors, Soall, who mainly posed as UK attorneys, kept a handwritten script titled "Lawyer Call." The indictment quotes it:

    "Target company, company you are acquiring is publicly listed. Legal + regulatory perspective confidentiality is a greater priority . . . target company project shareholders + also greater security concerns around insider trading. Sure you understand all this? [] XYZ said you'd probably ask if anyone else had been NDA'd . . . answer is not yet . . . although that is likely to change in coming days . . . internal memo coming days."

    ("XYZ" stands for the impersonated CEO.)

    This is the center of the attack. Confidentiality around the acquisition of a public company is real: people who learn about a deal early are routinely told not to discuss it, because of insider-trading rules. Anyone who has worked in finance has heard that instruction. That is exactly why it works as cover. The script even anticipates the obvious question and answers it before the employee can act on it.

    The Pressure From Above

    There is one more control the scheme defeated, and it is the one most companies rely on without writing it down: an experienced employee's sense that something is wrong.

    In finance, the person who doubts a payment is often the person with the least power in the conversation. When the instruction comes from the CEO, saying "I don't feel right about this" means arguing with the boss, on a deal you've been told is confidential and urgent. The scheme was built to win that argument. The voice on the call sounded like the CEO's. The lawyer used the name of an attorney at a famous firm. The secrecy was presented as a legal requirement. An employee who hesitated had nothing objective to point to, only a feeling.

    The One That Got Away

    Among the victims, the indictment lists six companies that lost money:

  • Financial technology company, United Kingdom: $11,578,474
  • Pet food company, Florida: $5,242,747
  • Hospital system, Missouri: $4,829,580
  • Mutual fund company, Massachusetts: about $4.64 million
  • Hospital system, California: $2,968,497
  • Bank, Rhode Island: $2,684,580
  • A seventh case is the most useful lesson. In October 2024, the group tried to take about $2.85 million from a major investment bank in New York. The employee was convinced and authorized the wire, but the bank's back office stopped it before it was sent.

    The person on the call was fooled. A second check, by people who were not in the conversation and not under the CEO's pressure, was not.

    Why Procedures Alone Aren't Enough

    Companies like these have safety procedures. A scheme like this one is designed to get around each of them, because each procedure depends on one person making the right call under pressure, with nothing objective to rely on.

    Rules still matter. Confidentiality should never replace approval, and the second approver should never be part of the conversation. But rules can't tell a cloned voice from a real one, an edited document from a genuine one, or a real lawyer from someone using his name.

    That takes technology: tools that check the calls, documents and identities behind high-value payments, automatically and continuously, so the decision never rests on one person's gut feeling. The New York bank was saved by a second check. The right system makes that second check happen every time.

    How DeFake Helps

    DeFake is built to fight this new generation of fraud: AI-cloned voices, faces, documents and pictures. DeFake is a forensic and provenance engine that gates fraudulent transactions.

    Protect your payments before the next call comes in. Talk to us about DeFake for your finance team

    If you think your organization was targeted by the scheme described here, the FBI asks that you report it at tips.fbi.gov.


    Sources

  • Superseding indictment, United States v. Soall, Hazan and Sheetrit, S3 26 Cr. 168 (S.D.N.Y.): PDF
  • U.S. Attorney's Office, Southern District of New York (November 5, 2026): Three Foreign Nationals Arrested And Charged With Scheme To Use Super Intelligence To Defraud Companies Of $129 Million
  • CyberScoop (November 6, 2026): Authorities accuse trio of using AI deepfakes to defraud $129M from businesses
  • Legal note: The charges in the indictment are accusations. All defendants are presumed innocent unless and until proven guilty in a court of law. Attorneys for Guy Sheetrit have said he has no connection to the matter, according to news reports.

    Share this article:

    DeFake - AI Detection Resources