WPP deepfake scam: a fake Mark Read, a cloned voice, and a target who said no (2024)

The WPP deepfake scam is a case that failed, which is exactly why it's useful. In 2024, fraudsters set up a WhatsApp account using a public photo of Mark Read, chief executive of the world's biggest advertising group. From it, they arranged a Microsoft Teams meeting that appeared to include Read and another senior WPP executive. On the call they played an AI clone of that second executive's voice over old YouTube footage of them, while "Read" chipped in through the text chat. The target, a WPP agency head, was asked to help set up a new business. Read's internal email said the goal was personal details and money. The agency head didn't bite. Public reporting: 10 May 2024. Status: Attempt failed; nothing lost; attackers unidentified.

How it worked

WPP never said what tipped the executive off. No glitchy video, no telltale audio artefact was reported. A spokesperson credited "the vigilance of our people, including the executive concerned". In other words, the request itself, or the way it arrived, felt wrong.

Tip: a meeting invite from a senior leader that arrives via a personal messaging app rather than the corporate calendar is a red flag on its own, before anyone says a word.

Whose voice was cloned?

Here's a genuine ambiguity in the record. The Guardian, which obtained Read's email, wrote that the impostors "deployed a voice clone of the executive as well as YouTube footage of them". Which executive? Coverage split. Some outlets (the FT headline, the New York Post) read it as Read's voice. TechRadar, quoting the same email, said the second executive was the one cloned and that the fake Read only took part via chat.

This page follows the second reading. It fits The Guardian's next sentence, that Read was impersonated "off-camera using the meeting's chat window", which would be odd if his voice were also live on the call. WPP never named the second executive or the agency head, and never gave the date of the attempt.

Why a failed attempt matters

This came a few months after the Arup deepfake, where a video call full of synthetic colleagues led to HK$200 million in transfers. WPP shows the same playbook aimed inward, at a company's own senior staff, and shows it failing.

Think about where it failed. The attackers got a target onto a call and delivered their story. That's all they managed. They never reached a payment request, so payment thresholds, callback rules and dual authorisation never came into play. Those controls can't be what saved WPP. A person did, at the earliest and cheapest moment to catch it. A case that ends in a wire teaches you to harden the wire. This one teaches you that the decisive control sits upstream, in whether an odd request on an odd channel gets checked before anyone acts.

Read's email put it plainly to staff: "just because the account has my photo doesn't mean it's me". He also noted that senior leaders face more tailored, psychological attacks than the general public does.

What stops it

• Verify unusual requests out of band, on a channel you already use with that person, no matter how convincing the voice or video.
• Treat personal-app meeting invites as suspect. Real executives use the corporate calendar.
• Train leaders specifically. Generic phishing awareness doesn't cover "your CEO is on a call asking for a favour".
• Assume your public footage is attacker material. Interviews, panels and press photos are enough to build a clone.
• Reward the person who says no. WPP publicised the near miss. That's how the next target knows what to look for.

See also deepfake examples, voice cloning and CEO fraud.

Common questions