AI Deepfakes Trick Staff Out of $129 Million in CEO Scam

Three men were arrested for using AI voice and video clones of CEOs to steal $129 million, proving you can no longer trust your own eyes.

Business ยท Source: U.S. Department of Justice (SDNY)

What happened

Federal prosecutors unsealed charges against three men for a massive AI fraud scheme. Jacob Soall, Sharon Hazan, and Guy Sheetrit allegedly used artificial intelligence to clone the voices and faces of corporate executives. They targeted thousands of companies across the globe for over a year and a half. The US Department of Justice says they successfully stole roughly $129 million from unsuspecting workers.

The attackers ran a highly coordinated playbook to bypass corporate defenses. They started with an email pretending to be a CEO needing help with a secret financial transaction. Next, they jumped on a phone or video call using AI to mimic the executive. The indictment states they used AI to clone both the voice and sometimes the image of the CEO. They then directed the employee to a fake attorney. Soall even used a handwritten script titled Lawyer Call to play the part of a UK attorney and sound authentic.

The fake lawyers instructed accounting employees to wire massive payments to international bank accounts. The stolen funds were quickly converted into cryptocurrency to hide the trail and distribute the money. Victims included hospital systems in Missouri and California, a Rhode Island bank, a UK fintech company, a Massachusetts mutual fund, and a Florida pet food brand. The three suspects, who used nicknames like Rocky Balboa and Superman Cash, were arrested between June and September.

Key facts

Why it matters

The trust model for corporate communication is officially broken. A live video call with your boss is no longer cryptographic proof of identity. Founders and engineers building internal tools must now design systems assuming audio and video can be spoofed in real time. Authentication requires out-of-band verification, not just visual recognition. If anyone asks you to move money, you must verify the request through a second channel before you send anything.

This scales social engineering to an industrial level. Previously, whaling attacks required massive manual effort and relied mostly on text-based phishing. Now, attackers can weaponize emerging technologies to bypass human intuition entirely. As the FBI noted, criminals are rapidly adapting to these tools. Enterprise security budgets will be forced to pivot toward deepfake detection and strict multi-channel payment approvals. The sheer scale of this attack shows that current corporate training is entirely inadequate for the AI era.

For builders

Mandate out-of-band verification for payments

Never trust a single channel for financial requests. If a CEO asks for a wire transfer on video, employees must verify it via a separate channel like a phone call to a known number. Companies that fail to implement this will lose millions.

Build deepfake detection enterprise tools

Security vendors have a massive opportunity here. Companies will pay a premium for tools that can analyze live video and audio streams to detect AI manipulation. Founders who build reliable real-time detection will capture massive enterprise security budgets.

Redesign accounting workflows with hard stops

Software engineers building fintech and accounting platforms need to implement multi-signature approvals for large transfers. Relying on human judgment alone is a massive liability when AI can perfectly mimic authority figures. Build software that forces compliance.

My take

I keep warning founders that AI is a dual-use technology, and this $129 million heist proves it. If you are building financial workflows, you are negligent if you still rely on human eyes and ears for authentication. We have to engineer zero-trust systems because seeing is literally no longer believing.

Original reporting: U.S. Department of Justice (SDNY). This is my rewrite and opinion.

More AI news for builders