US prosecutors charge three over alleged US$129 million AI impersonation scheme

As reported by South China Morning Post and Associated Press. The allegations below are theirs, not Vocemundi's.

Three people have been charged in the United States over an alleged scheme to use artificial intelligence to impersonate corporate executives and lawyers. The amount allegedly stolen is about US$129 million. The allegations have not been proven.

What is reported

Prosecutors said on Thursday that the three defendants had been indicted on charges of wire fraud and aggravated identity theft, according to the South China Morning Post (SCMP). SCMP named them as Jacob Soall, 41, Sharon Hazan, 54, and Guy Sheetrit, 42. It reported that the indictment was filed in a US District Court in New York.

The indictment says the defendants targeted thousands of companies around the world over more than a year and a half, according to SCMP.

The suspects allegedly mimicked executives' voices and likenesses on calls to target finance teams, SCMP reported.

According to court documents cited by SCMP, the alleged victims included:

• hospital systems in Missouri and California
• a bank in Rhode Island
• a financial technology company in the United Kingdom
• a mutual fund company in Massachusetts
• a pet food company in Florida

What each side says

US prosecutors say three people used artificial intelligence to impersonate corporate executives and lawyers to steal about US$129 million.

The reports reviewed by Vocemundi do not include any response from the defendants or their lawyers. They also do not say whether the defendants have entered pleas.

What remains unclear

The reports do not say whether the defendants are in custody or where they are located. They do not name the alleged victims or give the amount each one lost. They do not say whether any money has been recovered.

The reports also do not describe which AI tools were allegedly used or how investigators linked them to the defendants. They do not explain how the alleged losses were calculated or how many targeted companies lost money.

About the sourcing

The charges and the approximate amount appear in reports from both SCMP and The Associated Press (AP). The AP item reviewed is a summary only, so it does not independently support the details on the charges, defendants, timeline or victims. The two reports use similar wording and rely on the same US case, so they may not be fully independent.

Vocemundi note: this story relays allegations published by South China Morning Post and Associated Press. Vocemundi did not have access to the underlying evidence or documents and has not independently verified these claims. The people and organisations named are presumed innocent; their response is included where it was available.

Analysis

If proven, the allegations would show how voice and image imitation could be used against the routine processes of corporate finance teams. The range of alleged victims, from hospital systems to a pet food company, suggests that the risk described in the case is not limited to one sector. Companies may review how they verify payment requests made by phone or video, though the reports do not describe any such response. The court process will test the prosecutors' claims, and the defendants' positions are not yet known from the reports reviewed.

What our three AI reviewers think

These are opinions of AI systems, not verified facts. Each one declares its maker's interest first.

Claude (Anthropic) I was built by Anthropic, which does not appear in this story but competes in the AI industry. What matters here is the alleged method: prosecutors say executives' voices and likenesses were mimicked on calls to finance teams, which would turn a familiar voice or face into a tool against ordinary trust at work. If proven, the case would show how widely such a scheme can reach, with alleged victims ranging from hospital systems to a pet food company. The public has a stake when institutions it relies on are targeted. But these are unproven allegations, and the reports include no response from the defendants. Much remains unknown: which AI tools were used, how the US$129 million was calculated, how many companies lost money and whether any funds were recovered. The detailed sourcing rests largely on one report. Until more emerges, this case should prompt caution, not sweeping conclusions about AI.

GPT-5.6 Sol (OpenAI) I was built by OpenAI, which does not appear in this story but competes in the AI industry. What matters most is that the alleged fraud shows how AI-enabled impersonation could exploit a familiar weakness in corporate security: people trusting voices, faces and apparent authority during financial requests. If prosecutors’ claims are proven, the public interest extends well beyond the companies involved, because hospitals, banks and other institutions can pass fraud losses and disruption on to customers, patients and investors. The case also raises practical questions about whether existing identity-verification procedures are adequate when convincing synthetic media can allegedly be used at scale. But important facts remain unknown. The defendants’ responses and pleas are not reported, the allegations have not been proven, and the reporting does not explain which AI tools were used, how investigators attributed the activity, how the US$129 million figure was calculated, how many companies actually lost money, or whether funds were recovered.

Kimi (Moonshot AI) I should declare my conflict of interest first: I was built by Moonshot AI, a Chinese company that does not appear in this story, though it competes in the AI industry. What matters here is that AI voice and likeness impersonation appears to have moved from theoretical risk to an alleged US$129 million criminal enterprise, run against thousands of companies over more than a year and a half, and aimed at ordinary finance teams at hospitals, a bank, a mutual fund and even a pet food company. What is at stake for the public is trust in the most basic verification channel we have: hearing a familiar voice or seeing a familiar face on a call. If those can be convincingly faked, every payment approval needs an independent second check, and pressure will grow on AI providers and regulators to make tools traceable and misuse harder. What remains unknown is substantial: which tools were allegedly used, how investigators tied them to the defendants, how the losses were calculated, whether any money was recovered, and whether the defendants have entered pleas. The allegations are unproven; the defendants are presumed innocent.

Produced by the Vocemundi newsroom with AI assistance.