DashboardBook a demo
Blog
Platform
Red teaming
Use cases
Resources
PricingDashboardBook a demo
Summary

In January 2024, a Hong Kong-based finance employee at global engineering firm Arup authorized HK$200 million in wire transfers to fraudsters after a video call. Fraudsters used AI-generated video and audio to impersonate senior Arup figures on the call. This article reconstructs how the attack was built, why every layer of Arup’s control architecture failed, and which forensic signals could have interrupted it before a single transfer was authorized.

This attack is one of several examples covered in our broader guide on AI social engineering.

Could you tell a deepfake from the real thing?
Real-or-AI rounds · voice, video, and images · no sign-up
Key takeaways
  • Traditional security controls can’t stop attacks that exploit trust rather than technology.
  • Video calls are no longer reliable proof of identity because of AI-powered deepfakes.
  • Deepfake fraud exploits human trust, so every high-value authorization workflow now needs built-in identity verification.
  • Simple out-of-band verification checks can prevent fraud at virtually no cost.
  • Liveness detection alone is insufficient against modern deepfakes and synthetic media injection attacks.
  • Deepfake detection tools rapidly lose effectiveness unless they are continuously updated.
  • Most organizations expect deepfake attacks but remain unprepared to respond effectively.
  • Verifiable digital identity is evolving from a security requirement into a competitive advantage.

What happened to London-based engineering firm Arup in January 2024 may be the most significant corporate deepfake scam confirmed to date, and finance teams have studied it closely ever since. Arup, with 18,000 employees spread across 35 countries and a dedicated in-house cybersecurity team, fell victim to a cyberattack that cost the firm HK$200 million over the course of about a week. The incident was an eye-opener for many organizations because it did not breach the network or steal credentials. Instead, attackers targeted a specific finance employee, who joined a video call, saw the CFO and several senior management colleagues, heard familiar voices, and did exactly what they asked him to do.

Among documented deepfake scam examples targeting enterprise finance teams, the Arup case stands out for what the attackers bypassed: a control architecture that worked as designed but failed because it rested on an assumption synthetic media had made obsolete.

Every other participant on the video call was a deepfake of a real colleague. Arup’s IT environment remained fully intact throughout the attack; the attackers never tried to hack the network. They only needed to look like people who were already trusted. Nothing in the network logs would have caught this, so stopping it required deepfake video detection on the call itself, before the transfer was approved.


Watch the 6-minute breakdown of how the attack worked:

Video: A Fake CFO Stole $25 Million on a Video Call | Arup Deepfake Scam
Chapters

What Arup is, and why that raises the stakes for every enterprise

Arup Group, founded in 1946, has a portfolio of landmark projects, including the Sydney Opera House, the Beijing National Stadium, the London Millennium Bridge, and substantial portions of the UK’s HS2 rail program. Its work routinely involves multinational financial coordination, sensitive government contracts, and complex project financing across jurisdictions. It maintains a dedicated in-house global cybersecurity and information assurance team.

Arup didn’t skip the security basics. It had the resources, the mandate, and the expertise to run a mature control environment.

This case is relevant to your security team because your controls likely rest on the same assumption Arup’s did: that a familiar face on a live video call is a reliable signal of identity. That assumption no longer holds, and the Arup incident is the clearest evidence yet that organizations that haven’t updated it are exposed.


How the attack was built, stage by stage

Stage 1: Reconnaissance and source material harvest

A CISO does not need to attend a conference for an attacker to learn their voice. Executives with a large public profile are at higher risk because they’ve generated years of raw training material that anyone can find.

The attackers never touched Arup’s internal systems to build what they needed. Hong Kong police confirmed the deepfakes were constructed entirely from public material, including conference recordings, virtual meeting footage, and media appearances. The CFO and several colleagues had enough publicly available video and audio to train convincing face-swap and voice-clone models.

What a layered detection stack could flag: Reconnaissance itself is invisible, but what it leaves behind isn’t. Diopter’s media authentication layer reads C2PA provenance signals and cross-references them against known synthetic generation fingerprints. When a face-swap or voice-clone model built from harvested material enters a live session, missing or unverifiable provenance isn’t proof of fraud on its own, but paired with synthetic-media signals, it can trigger additional checks.

Stage 2: The spear-phishing entry point

In January 2024, a phishing email that appeared to come from the UK-based CFO reached a finance employee in Arup’s Hong Kong office. It cited a confidential transaction, was marked urgent, and asked for discretion.

Initially, the employee was skeptical about the email. That skepticism was the one moment in the sequence when stopping the attack required no technical tool, policy update, or additional budget.

Many deepfake fraud attacks follow this pattern. The phishing email’s job is to set up the follow-up call that closes the deal. The video walkthrough of each stage of the Arup deepfake attack shows how this email set up the call that followed.

What a layered detection stack could flag: The email pairs urgency framing with a confidentiality instruction, a recognizable attack pattern. Requests like this are designed to route around standard authorization channels without leaving a calendar trail behind. Under NIST SP 800-63-4’s identity assurance framework, these are pre-authentication attack precursors: signals that appear before any credential is tested and before any verification is attempted. Diopter’s deepfake detector has a behavioral analysis layer that reads these sequences in combination. An executive impersonation signal paired with an authorization bypass request can produce an alert before your employee joins the call, not after the wire clears.

Stage 3: The synthetic video conference

When the employee joined the meeting, he saw the CFO on screen, along with other familiar colleagues. He didn’t raise an alarm because the call looked like every other executive meeting he had attended.

In fact, every other person on that screen was synthetic.

The attackers used deepfaked video and voices built from the footage harvested in Stage 1. Hong Kong police later said the meeting was pre-recorded, so there was no real interaction between the employee and the fake participants. The employee did what most people would do. He saw and heard people he recognized and trusted the instruction. He approved the first of 15 transfers to five accounts and authorized more over the following days.

The Arup case is so widely studied because it put multiple AI-generated executives on one live call, each built entirely from footage that was never private.

What a layered detection stack could flag: A live real-time face-swap pipeline usually needs a virtual camera driver to feed the synthetic stream into the conferencing application. That driver leaves a fingerprint. Diopter’s injection-attack detection layer reads device telemetry at the session layer.

Synthetic or injected streams may lack a stable, sensor-consistent photo response non-uniformity (PRNU) signature, or may show metadata that does not match the claimed capture device. On the audio side, voice synthesis still struggles to replicate the breath sounds, micro-variation between words, and slight inconsistency of unrehearsed speech. A layered detection stack can evaluate these signals in combination. These signals were likely present on the Arup call, but no tool was reading them. A detection layer positioned before authorization could have escalated the session for step-up verification before the first transfer instruction went through.

Stage 4: Discovery through routine follow-up

The fraud did not trigger traditional security alerts because it operated outside the systems those controls were built to monitor.

The employee contacted Arup’s head office after the transfers to follow up on the confidential transaction he had just authorized. The head office confirmed that no meeting had taken place and no transfer had been requested.

By the time the investigation started, the money had already been moved through the five accounts. As of 2026, the HK$200 million has not been publicly reported as recovered.

In June 2026, threat-intelligence reporting from @BushidoToken noted that Arup had been listed on FulcrumSec’s Tor leak site. FulcrumSec claimed it had gained initial access in September 2025 through a GitHub personal access token hardcoded in a JavaScript file on a forgotten subdomain, giving access to more than 10,000 private GitHub repositories. The claim has not been independently confirmed by Arup, and there is no public evidence linking this alleged breach to the 2024 deepfake fraud.


The Numbers: HK$200 million (about US$25.6 million) transferred over roughly a week. 15 wire transfers across five separate bank accounts. 0 systems breached throughout the entire incident. 0 funds publicly reported as recovered as of 2026. 0 arrests announced by Hong Kong authorities.

Would your stack have flagged the Arup call?Diopter maps your current controls against every stage of a real deepfake attack arc and shows you exactly where the gaps are. Start with a simulated attack on your own team to see where your process breaks.

See how it works


Why every control your security team relies on failed

Every control in Arup’s security architecture (firewalls, multi-factor authentication, endpoint protection, transaction monitoring, and segregation of duties) was working on the day of the attack, and none of them were relevant.

That’s what separates deepfake fraud like the Arup case from conventional business email compromise. There are no network intrusions to detect, no malware signatures to flag, and no anomalous logins to trigger an alert. The attacker never attempted to enter the system. Your entire technical control layer is built to defend a boundary this attack never approached.

Here is what that looked like, layer by layer.

The perimeter nobody crossed

Arup ran firewalls, intrusion detection and prevention systems, network segmentation, and secure VPN access. Your organization almost certainly runs the same stack, but none of it applied. Perimeter security is designed to stop an attacker who tries to get inside, and since the attackers did not need to infiltrate the network, no one caught it.

How Diopter’s layered detection stack addresses this: Diopter defends the identity layer that sits in front of the perimeter. The attack surface in the Arup case was a human authorization decision, made during a video call, with no tool positioned to read the media stream it was made on.

MFA protects credentials, not judgment

Multi-factor authentication stops an attacker who has stolen a password and is trying to use it. In the Arup case, the credentials were never at risk, and no password was stolen. Instead of stealing access, the attackers convinced the employee to authorize a transfer he believed was legitimate, and neither MFA nor your IAM platform can check for that.

How Diopter’s layered detection stack addresses this: Diopter’s behavioral analysis layer reads the communication sequence that precedes a financial authorization, not the credential that enables it. An executive impersonation signal combined with an authorization bypass request is a detectable pattern.

Segregation of duties: Satisfied from the inside

Multi-person approval is designed so that no single employee can both initiate and sign off on a significant financial transaction without independent confirmation. It’s a sound control, but in Arup’s case it failed without being technically bypassed.

From the employee’s perspective, the multi-person approval had already happened. The CFO was on the call, senior colleagues were present, and leadership had issued the instruction directly. The control was satisfied from inside a synthetic video session by people who did not exist.

How Diopter’s layered detection stack addresses this: A segregation of duties control cannot verify whether the people in a video call are genuine. A layered detection stack can add that missing verification layer by checking media, device, and session-level signals before authorization proceeds. The injection-attack detection layer reads device telemetry at the session layer before a single word of instruction is delivered. A virtual camera pipeline feeding synthetic video into a conferencing application resolves differently than a physical sensor. That discrepancy is readable before any authorization decision is made.

Video call authentication

This is where the architecture failed most completely, because this is the control the attackers specifically targeted.

For years, a live video call with a recognizable face and a recognizable voice has been treated as strong identity confirmation. If you can see your CFO and hear them, the assumption has always been that you can trust the instruction. That assumption is the attack surface. Arup’s attackers didn’t need to defeat this control. They exploited it.

Most documented enterprise deepfake scams follow the same path, exploiting the quick judgment employees use to decide whom to trust.

Video call platforms were built to transmit faces, not verify them. That gap between what the platform shows and what it can confirm is exactly where synthetic media operates. Your employee’s brain fills the verification void with pattern recognition wired for a world where familiar faces mean safe people. The Arup incident changed that assumption for all employees.

How Diopter’s layered detection stack addresses this: Diopter’s media forensics layer operates on the live video stream itself. Since real-time face swaps carry no consistent PRNU signature, a detection layer reading the stream could have flagged the session for review before the first transfer instruction landed.


The shift your board has not yet made

Traditional cybersecurity architecture is built around one objective: to stop an external actor from getting inside your systems.

The Arup incident showed that an attack can arrive as a deepfake identity that never touches the perimeter. Organizations that recognize this can build something competitors can’t easily copy: a verification architecture that does not depend on a human employee correctly identifying a synthetic face under pressure.


The forensic signals that were present but not read

Every AI impersonation at this level of sophistication exhibits forensic signals before the financial transfer is authorized.

The Arup case was no exception. What stands out is that no trained person or detection layer was positioned to read them.

The Arup deepfake scam case is particularly instructive because the signals span both behavioral and technical categories. Several required no tooling at all.

A 2025 study out of Beijing University of Posts and Telecommunications analyzed 482 phishing emails and found that authority and urgency cues beat every other manipulation tactic when attackers want someone to act before they stop to verify. Arup’s attackers didn’t need a careless employee. They needed him to do exactly what authority bias and manufactured urgency are designed to produce in anyone: defer to a senior figure under time pressure, and mistake that deference for verification.

A familiar face on a video call and a confidential, time-sensitive request work together. The face provides authority, and the urgency closes the window in which the employee would normally pause to confirm it through a second channel. The researchers concluded that removing either half could stall the attack. That’s why “out-of-band verification, no exceptions” works against this category of fraud regardless of how convincing the synthetic media gets.

Behavioral signals available to any trained employee

  • The meeting was initiated by email, not a calendar invite from an internal scheduling system. Legitimate executive meetings at Arup would likely generate an internal calendar entry with an auditable creation trail. This one did not.
  • The request was designed to bypass standard approval workflows. These three framing elements (confidential, urgent, and discretionary) appear consistently across documented deepfake fraud because they move a target past the natural hesitation that multi-step approval processes are designed to trigger.
  • The session terminated the moment the financial authorization was obtained. A genuine executive meeting ends with pre-scheduled follow-ups, informal conversations, or a natural wind-down. This one closed as soon as the specific action was completed.
  • After the call, none of the participants were reachable for confirmation. The employee who discovered the fraud did so by contacting the head office directly after the fact. Any attempt to reach participants through standard channels during the call’s follow-up window would have produced the same result.

Technical signals detectable by a positioned detection stack

The behavioral signals above are available without any specialized tooling. The technical signals below require a detection layer operating against the media stream. They are the signals that a forensic classifier, if properly positioned, would have flagged before a transfer was authorized.

  • PRNU absence: A wholly synthetic video stream, unlike one generated by a genuine camera, carries no consistent PRNU. The absence of this signature is one of the most reliable signs of synthetic content. It’s invisible to the human eye and detectable only by a media forensics classifier.
  • Facial edge artifacts: Real-time face-swap rendering produces characteristic softening and warping around facial boundaries, particularly at the hairline, the ear-to-jaw transition, and at oblique angles beyond approximately 30 degrees from center. These artifacts are not always visible at casual viewing resolution but are consistent and measurable.
  • Prosodic anomalies in the audio: Voice synthesis flattens the spectral and temporal variation that characterizes natural speech. A synthesized voice lacks the breath sounds, the micro-hesitations, the self-corrections, and the natural cadence variation of an unscripted human speaker. These are detectable through spectral analysis of the audio stream.
  • Virtual camera device metadata: Real-time face-swap pipelines require a virtual camera driver to pipe synthetic video into the conferencing application’s media input. The device identifier for a virtual camera resolves differently than a physical sensor. A detection layer reading device telemetry would see this discrepancy before the call began.

None of these signals required experimental tooling to act on in January 2024. The PRNU absence analysis, spectral audio detection, and facial artifact classification were all operational techniques by then. They just weren’t positioned against this channel in Arup’s environment.


The Arup attack is part of a pattern

The Arup case is the largest enterprise deepfake fraud on record, but it wasn’t the first. Three other cases from 2019 to 2024 show the same attack logic, with the tools improving each time.

Energy firm CEO voice fraud, March 2019

In March 2019, attackers used a cloned voice of the CEO of a German parent company to instruct the CEO of its UK subsidiary to wire $243,000 to a Hungarian supplier. The call sounded fine, and the framing was urgent and confidential. The transfer went through. The funds were moved through multiple accounts and never recovered. The target grew suspicious only when the same cloned voice called again asking for another transfer.

In this incident, voice alone was enough. Five years later, the Arup attackers used the same logic of voice confirmation, urgency framing, and a request engineered to bypass normal authorization.

WPP CEO impersonation, May 2024

In May 2024, attackers built a fake WhatsApp account using publicly available photographs of WPP CEO Mark Read, then used it to set up a Microsoft Teams meeting with senior executives. During the meeting, they used a voice clone and YouTube footage of Read, and impersonated him in the chat to ask for money and personal details for a new business venture.

The executives didn’t engage, and no funds were transferred.

Ferrari CEO impersonation, July 2024

In July 2024, a senior Ferrari executive received calls from a cloned voice impersonating CEO Benedetto Vigna, referencing a confidential acquisition requiring immediate action. The voice clone was convincing. The attack ended when the target asked an offhand question about a book Vigna had recently recommended, which the attacker couldn’t answer.

The trend data

According to research published by Entrust, deepfake attempts were happening at a rate of one every five minutes by 2024.


A 2024 Medius survey of businesses in the US and UK found that 53% had already been targeted by deepfake-powered financial scams.

The numbers show how common these attacks already are. CISOs need to ask whether their detection architecture will catch an attack at the right stage of the execution chain.

The pattern wasn’t confined to finance teams

The same month Arup’s finance employee joined that call, AI-generated audio of a different kind reached thousands of New Hampshire households: a synthetic Joe Biden robocall telling voters to stay home from the primary. Eight months earlier, a fabricated image of a Pentagon explosion had briefly moved the stock market before officials confirmed nothing had happened. The targets and payoffs differed, but the mechanism was the same: synthetic media, built from public material, used to manufacture trust or panic faster than anyone could verify it.

That’s why the lesson from Arup reaches beyond corporate treasury. Synthetic media is a general-purpose attack capability, and finance was simply where the money was.

See which attack surfaces your stack is leaving uncovered.Diopter scores every detection layer against your live environment and walks a real attack arc with your team. Start with a simulated attack on your own team to see where your process breaks.

Get a free assessment


What adequate defense looks like

Each recommendation below responds to a specific failure in the Arup case or a pattern seen across documented deepfake fraud. These controls hold up under pressure.

Out-of-band verification, no exceptions

Any request involving a wire transfer, credential reset, or vendor banking change requires confirmation through a second channel, using contact details your organization already holds, not what the caller provides. Urgency and confidentiality don’t override this rule. They appear in most documented deepfake fraud cases, so treat them as escalation triggers. This kind of control maps directly to Diopter’s wire fraud prevention.

Arup introduced these controls across all offices in March 2024, one month after the incident was confirmed. A pre-agreed callback to a verified number, applied consistently to any financial authorization originating from a remote interaction, could have stopped all 15 wire transfers before the first one cleared.

Behavioral verification that AI cannot pass

Visual and audio recognition are no longer reliable identity controls. Attackers struggle to defeat a spontaneous, unscripted challenge based on shared knowledge, like a reference to a specific recent conversation, a detail from an internal meeting, or a time-bound code exchanged through a separate channel. The Ferrari impersonation failed because the target asked about a book they had recently discussed. One unscripted question, requiring knowledge no public source could supply, was enough to stop the attack at its critical moment. The WPP attempt failed because the executives didn’t act on the request.

Your employees do not need to be deepfake experts. They need to know that any high-value authorization request, however convincing the caller appears, requires a verification step the caller cannot have prepared for.

No voice-only authorization

Wire approvals by phone alone are not authorizations. Neither are password resets based solely on a call, or vendor payment changes without written, independently verified confirmation. If your policy does not say this explicitly, it leaves the gap open.

Hard controls on treasury and accounts payable

Require dual authorization on all wires, lock pre-approved vendor banking details before a transfer request arrives, and set a mandatory cooling-off period for any new payment instruction, whoever issued it. These controls could have stopped the Arup transfers at the process layer.

Media forensics detection at the stream level

For organizations processing high-value authorizations over video, a media forensics classifier operating on the live stream can flag PRNU absence, spectral audio anomalies, and facial edge artifacts before a transfer instruction is issued. The classifier doesn’t need a definitive verdict, only enough to flag the session for escalation, which shifts the burden of proof from the employee to a secondary verification process.


Detection also degrades over time. The Deepfake-Eval-2024 benchmark (arXiv 2503.02857) recorded a 50% drop in area under the curve (AUC) for video detectors tested against real-world content compared with the academic datasets they were trained on. A classifier certified a year ago but not updated since has already lost ground against current tools. Detection that holds up in production has to be maintained continuously.

Injection-attack detection is a separate compliance requirement

Presentation attack detection (PAD) under ISO/IEC 30107-3 was built to stop an attacker holding a photograph in front of a camera. It does not stop an injection attack, where a virtual camera driver pipes synthetic video directly into the application’s media stream, bypassing the physical sensor entirely. CEN/TS 18099, published by the European Committee for Standardization in late 2024, is the first technical specification written specifically to evaluate injection-attack detection. NIST SP 800-63-4 now requires both PAD and injection-attack detection as separate normative controls. A single ISO/IEC 30107-3 certificate does not satisfy this requirement.

The same kind of synthetic media used on Arup’s video conference would work just as well against the systems that open accounts. A 404 Media investigation in 2024 found that a website called OnlyFake could generate a convincing fake passport for roughly $15, enough to clear the identity-verification stage at the crypto exchange OKX, which relied on a third-party verification provider, Jumio. Federal prosecutors later charged the site’s operator with selling more than 10,000 fraudulent IDs between 2021 and 2024, used to bypass KYC checks at banks and exchanges.

The same distinction applies here as on video calls. A presentation attack holds a fake photo up to a camera, and liveness detection is built to catch it. A digital injection attack pipes synthetic video or a forged document directly into the verification pipeline through a virtual camera or emulator, bypassing the physical sensor the liveness check was designed around. Most KYC products were built for the first kind of attack, and injection attacks are the harder one to catch.

Help desk verification that AI cannot guess

Knowledge-based questions are insufficient. An attacker with LinkedIn access, a company website, and two minutes of publicly available audio can answer most of them. Require ticket-based workflows, manager approval for executive account changes, and step-up authentication for anything sensitive. Ask verification questions that require real-time, unscripted knowledge the help desk can check, like what was discussed in yesterday’s meeting. These are not questions a voice clone can answer from a prepared script. If you have a recording of a suspicious call, our AI voice detector can check whether the voice is synthetic.

Executive conduct is a security control

Last-minute urgent payment requests that bypass the process are a social engineering template, not a leadership prerogative. Your finance team needs explicit backing to push back on an executive instruction that skips verification, including when the instruction appears to come from the CEO. That backing has to be established before the call arrives, not negotiated during it.

Train for the actual threat

A finance team that has experienced a convincing synthetic video call in a controlled environment responds differently in a live incident than a team that reads a policy document. Train your employees on what a deepfake call feels like and the specific step they take when one appears. The WPP and Ferrari cases both demonstrate that one practiced, instinctive challenge is enough to interrupt the attack, and that instinct comes from running drills, not reading security bulletins.

When it happens: Speed is the control

Escalate any deepfake attempt, successful or not, to legal and security immediately. Preserve call logs, recordings, and transaction records before they’re overwritten, and contact your financial institution right away, because wire recall windows are narrow. The organizations that recover funds are the ones that move in hours, not days.

No single vendor covers every one of these controls, and the differences go deeper than feature lists suggest. We broke down how two approaches diverge in Diopter AI vs Resemble AI.


Trust is now a technical architecture decision

Organizations drawing the right conclusions from cases like Arup aren’t just asking how to prevent this from happening to them. That framing keeps the conversation defensive, siloed inside the security budget, and reactive to an attack surface that changes faster than policy cycles.

What your board should be asking and probably isn’t

If your leadership team can’t answer all five of these questions, your authorization architecture has a gap that synthetic media can move through.

  1. Does your organization have a standing policy that voice-only confirmation is insufficient to authorize a financial transfer? If that policy is not written, not trained, and not enforced, you do not have it.
  2. When an executive issues an urgent, out-of-band instruction, what is the verification step? Who calls back, on what number, through what channel, and who owns that procedure when the executive in question is the one being impersonated?
  3. Your help desk is the most consistently exploited entry point in enterprise social engineering. Are the controls your team follows built for today’s threat actors or for the attack surface that existed five years ago?
  4. When a transfer request arrives under urgency and confidentiality framing, what is the override mechanism that stops it moving forward without independent verification?
  5. Has your organization run a live-voice or video impersonation drill against this specific scenario with your finance team?

Here’s what organizations must get right.

  • Deepfake fraud attacks do not compromise your systems. They compromise your people using AI-assisted realism to fool their senses.
  • If your authorization controls treat a familiar voice as sufficient confirmation, your organization is exposed.
  • The organizations that manage this risk best don’t rely on detection alone. They have clear verification procedures, enforced callback protocols, and executive teams that understand the threat well enough not to be its weakest link.

Diopter recommends

In 2026, the security question is no longer just what got into your systems. It is who your people trusted and whether your organization gave them any way to verify it.

For your clients and counterparties, verifiable trust is becoming a key selection criterion. Since Arup, enterprises that can’t demonstrate detection-layer coverage are a harder sell to sophisticated partners.

Diopter builds deepfake detection as a layered architecture that combines artifact forensics, media authentication, and injection-aware biometric verification, and we maintain it continuously so it doesn’t decay the way static tools do.

Find out which attack patterns your current stack covers. If you do not know whether your current controls would have flagged the Arup call, that is the right question to start with.

Browse our detection products, or talk to us for a tailored assessment of your stack.

Walk the Arup attack arc with Diopter.

In 30 minutes, we replay the real incident, show the signals Diopter would score, and map the verdict your team could act on before a transfer clears. This is precisely the exposure Diopter’s executive impersonation is built to close. Start with a simulated attack on your own team to see where your process breaks.

Get a free walkthrough


Also read: Deepfake Detection for Video Conferencing: Zoom, Teams, Meet, and Webex

FAQs

How convincing are current voice and video deepfakes?
Convincing enough that employees at Arup, WPP, and Ferrari were deceived or nearly deceived by them. The WPP and Ferrari attacks failed not because the synthetic media was detectable but because targets applied spontaneous, unscripted verification challenges the attackers had no prepared answer for.
What is the single highest-impact control an organization can implement today?
Mandatory out-of-band verification. Any financial authorization that originates from a remote interaction should be confirmed through pre-established channels, using contact details your organization already holds. This is the control Arup implemented one month after its incident. It could have stopped all 15 wire transfers before the first one cleared.
Does multi-factor authentication protect against deepfake threats?
No. MFA stops an attacker who has stolen a credential and is trying to use it. In a deepfake fraud scenario, no credential is stolen. The attacker convinces an authorized employee to take an action they believe is legitimate. MFA has no mechanism to read that transaction.
Are there laws against non-consensual deepfakes?
The TAKE IT DOWN Act (2025) criminalizes non-consensual sharing of AI-generated intimate images at the federal level but does not cover audio deepfakes. At least 26 states have laws addressing AI-generated intimate depictions. Missouri’s House Bill 1887, passed in April 2026, makes it a felony to share AI-generated depictions used to harass or harm, with enhanced penalties where minors are involved. Multi-state operators face conflicting jurisdictional requirements that federal preemption has not yet resolved.
What should employees do if they suspect a deepfake call in progress?
Apply one unscripted verification challenge before taking any action. Do not terminate the call immediately; that tips off the attacker. Escalate to security after the call ends. Preserve every available record: call logs, screen recordings, and transaction data. If a transfer has already been initiated, contact your financial institution immediately since wire recall windows are narrow.

More questions answered in our deepfake detection FAQ.

Get the Diopter threat brief

Monthly analysis of AI social engineering, voice fraud and deepfake attacks on enterprises.

One email a month. No spam, and we never share your address.

Ask AI about this articleClaudeChatGPTPerplexityUsing Claude or ChatGPT? Check any image, video or voice clip with Diopter for Claude or Diopter for ChatGPT.
Cite this articleAPA · MLA · BibTeX
APA 7
Gupta, S. (2026, June 22). Famous deepfake scams: The 2024 Arup cyberattack. Diopter AI. https://diopter.ai/blog/arup-deepfake-scam/
MLA 9
Gupta, Surojoy. "Famous deepfake scams: The 2024 Arup cyberattack." Diopter AI, 22 June 2026, https://diopter.ai/blog/arup-deepfake-scam/.
BibTeX
@misc{diopter20263de757, author = {Surojoy Gupta}, title = {Famous deepfake scams: The 2024 Arup cyberattack}, year = {2026}, month = {jun}, howpublished = {Diopter AI}, url = {https://diopter.ai/blog/arup-deepfake-scam/} }
SG
Security Researcher & Writer

Surojoy Gupta is a security researcher and writer with 8 years embedded in the cybersecurity industry, specializing in deepfake fraud, social engineering, and AI-driven threats. His work covers APT threat analysis, ransomware, and the evolving tactics attackers use to exploit enterprise trust at the human layer.