Developments · · 2 min read · Source: U.S. Attorney's Office SDNY
AI CEO voices allegedly stole $129M. The best defense is a callback
Accounting gets a call from the CEO. Confidential deal, needs to move fast. According to a US indictment, the voice was AI, and roughly $129 million walked out the door. If your first reflex is to buy a deepfake detector, you're solving the wrong problem.
Related tool · AdKrisp – Quiet callsTry it free
You know the setup. Email from the boss, subject line "confidential." Then he calls. The voice fits, the tone fits, so does the pressure.
That's how, according to US prosecutors, one crew allegedly cleared about $129 million.
What the indictment says
On Thursday, November 5, the US Attorney's Office for the Southern District of New York announced charges against three men: Jacob Soall, Sharon Hazan and Guy Sheetrit, from Spain, the UK, Israel and the UAE. Charges: wire fraud, aggravated identity theft, money laundering. Over more than a year and a half, they allegedly targeted thousands of companies worldwide.
The playbook, per prosecutors, in three steps:
- An email posing as the CEO asks accounting to help with a confidential transaction.
- A phone or video call where AI mimics the CEO's voice, and sometimes his image.
- A handoff to a supposed attorney at a major law firm, also faked, who directs wires abroad. The money was converted to crypto.
Alleged victims include hospital systems in Missouri and California, a Rhode Island bank, a UK fintech, a Massachusetts mutual fund company and a Florida pet food maker. The three were arrested between June and September in Spain and the UK; extradition requests are pending. These are allegations; the defendants are presumed innocent.
The tech is new. The trick is old
CEO fraud has been around for years. What's new: the verification call, once the scammers' weak spot, is now their strongest argument. The voice sounds real. So people believe it.
Which means any process built on "I'd recognize my boss on the phone" is broken. Not weaker. Broken.
Then step three. The scammers borrow trust from two brands at once: your own CEO and a big law firm. Prosecutors say one defendant even had a handwritten script titled "Lawyer Call." Social engineering is brand abuse with a screenplay.
The uncomfortable part for marketers
In my view, this lands on us too. For years we've told executives: show your face. Podcasts, LinkedIn video, keynotes. I do it myself and I'll keep recommending it. But every hour of public audio is also material a voice can be rebuilt from. The indictment doesn't say whether these defendants worked that way. The risk is real either way.
The fix isn't less visibility. It's designing visibility and payment processes together. If you turn the CEO into a brand, tell finance: this voice no longer proves anything.
What I'd do now
- Callback rule. Payments above a set amount only after a callback to a number from your own directory. Never the number in the email, never within the same call.
- "Confidential" as a tripwire. Any request for secrecy automatically triggers a second approver. No exceptions, not even for the CEO.
- Permission to say no. The CEO tells finance once, in writing: if you hang up on me and call back, you did everything right.
Bottom line: Better software won't beat a perfect voice. A rule that trusts no voice will.
Key facts
- On November 5, 2026, the US Attorney's Office SDNY announced charges against three men accused of taking about $129 million using AI-cloned CEO voices.
- Per the indictment, a CEO email was followed by an AI-faked call and a fake attorney at a major law firm; the money was converted to crypto.
- The three were arrested between June and September in Spain and the UK; extradition requests are pending, and they are presumed innocent.
Source
More sources
- CBS News / AP – Bericht vom 6.11.2026
- SecurityWeek – Bericht vom 5.11.2026
This article was researched and written with AI assistance. Responsible: Luigi Lugmayr.




