The boss on the call was not real

A finance worker joined a video call with the CFO and colleagues, then sent about US$25 million. Every face on the call was fake. How AI impersonation became a business attack, and the simple checks that stop it.

Analysis November 9, 2026 3 min read

A carved Noh theatre mask of a young woman’s face — Noh mask, Waka-onna, Japan, 1400s. Cleveland Museum of Art, CC0.

In January 2024, an employee in the Hong Kong office of a multinational firm reported to police that the company had been tricked out of about HK$200 million, roughly US$25.6 million. She had been on video calls with people who looked and sounded like the company’s senior officers, including its chief financial officer, and had been told to move money to designated accounts. She made 15 transfers to five local bank accounts.

“(In the) multi-person video conference, it turns out that everyone [he saw] was fake.”

Baron Chan Shun-ching, Hong Kong police, quoted by CNN

Police said the fakes had been built from video and audio already available online. In May 2024 the British engineering firm Arup confirmed it was the company, and that fake voices and images had been used. It said its financial stability and operations were not affected and no internal systems were compromised. There was no break-in. The attackers went straight to a person.

The same move, again and again

Arup was the expensive one. It was not the only one.

WPP, May 2024: scammers set up a WhatsApp account with a public photo of the chief executive, Mark Read, then a Teams meeting with a cloned voice and YouTube footage. The attempt failed.
LastPass, April 2024: an employee received calls, texts and a voicemail on WhatsApp with a deepfaked voice of the chief executive. The employee ignored them and reported it. The company said: “To be clear, there was no impact to our company.”
Ferrari, July 2024: an executive received WhatsApp messages from an unfamiliar number, then a call imitating the chief executive’s voice about a confidential deal. He asked which book the chief executive had recently recommended. The caller hung up.
US officials, 2025: impostors contacted foreign ministers, a governor and a member of Congress using the identity of the Secretary of State, and officials and executives received messages posing as the White House chief of staff. The FBI has warned that the campaign uses AI-generated voice messages.

A carved Noh theatre mask with a fierce scowling face — A mask was always a performance. Now it can join the meeting.Noh mask, Ko-beshimi, Japan, 1392 to 1573. Cleveland Museum of Art, CC0.

Now at industrial scale

In November 2026, federal prosecutors in New York charged three men over a scheme they say took more than $129 million from companies, hospital systems and a bank between July 2024 and February 2026. Prosecutors allege the men emailed victims as a chief executive, then held calls in which AI imitated the executive, and passed them to fake lawyers. These are charges, not findings, and the men are presumed innocent unless proven guilty.

The FBI has been warning about this since at least December 2024, when it said criminals were using AI to impersonate executives and authorities “during live video calls”. In 2025 it warned that AI voice cloning can “sound nearly identical”, and it has said North Korean IT workers use face-swapping during video job interviews. The US Treasury’s financial crimes unit, FinCEN, issued its own alert in November 2024 about deepfakes used to get past identity checks.

Why it works

Every one of these attacks borrows the trust a company already runs on: a familiar face, a familiar voice, a request from someone senior, an urgent reason. The technology removes the last cue people relied on, which was seeing and hearing the person. What it cannot copy is something only the real person knows, or a phone number the company already has.

What stops it

The defences that work are not sophisticated, and that is their strength. They come from the FBI’s own guidance and from the attempts that failed.

Call back on a number you already have, never the one in the message.
Confirm any request for money through a second, separate channel.
Agree a code word or a question only the real person can answer, as the Ferrari executive did.
Treat a new number, WhatsApp, or a request to move to Signal or Telegram as a warning sign.
Never share a login code, whoever asks.
Verify identity properly when hiring remotely, at interview and after.

The FBI’s latest advice fits on one line: “If you receive a message claiming to be from a current or former senior U.S. official, do not assume it is authentic.” Inside a company, the same is now true of anyone on a call.