What to Do When Someone Fakes Your Brand

Executive voice cloning for payment fraud is the most financially damaging vector. Attackers clone a voice from earnings calls, conference talks, podcast appearances or social video, then call a finance team member impersonating the CEO to authorise an urgent transfer. These calls tend to arrive late on Friday afternoons or during known travel, when independent verification is inconvenient. The compliance rate is the uncomfortable part: a CrowdStrike study found that 65% of employees who received a cloned voice call from their "CEO" complied with the request before seeking independent verification.

Real-time video call impersonation has escalated the same technique. The largest single deepfake scam on record used live video impersonation of multiple colleagues simultaneously on the same call, which defeats the instinct that a face on a video call is verification.

Other recurring patterns include fake job candidates using real-time face-swapping in interviews (flagged by the FBI's Internet Crime Complaint Center in 2022 and systematic by 2025), internal impersonation of IT staff, HR directors or team leads to extract credentials or authorise access changes, and brand or advertising impersonation, where synthetic executives or spokespeople appear in fraudulent ads. Each pattern targets a different form of institutional trust — a finance approval, a hiring decision, an internal help-desk request, or a public endorsement — which is why a single detection tool can never be a complete answer. Researchers identified more than 100 deepfake video ads impersonating a single UK prime minister on one platform in a single month, which gives a sense of how cheaply this scales. Cyble's Executive Threat Monitoring reporting found AI deepfakes involved in over 30% of high-impact corporate impersonation attacks in 2025, and CEO fraud using deepfakes now reportedly targets around 400 companies per day.

Two sources were notably honest about their own limitations, and their caution is worth adopting. One noted that a full-year 2025 reported loss total exceeding $1.28 billion should not be converted into an average loss per incident, because most incidents had no disclosed financial figure, the largest public cases dominate the total, and media datasets are biased toward events that become visible. Another stated plainly that the available datasets cannot be combined into one global total, because each measures a different stage of the fraud process. The figures worth putting in front of a board are the ones with a named source and a defined measurement — the CrowdStrike compliance rate, the Cyble corporate impersonation share, and the detection accuracy figures below — while the billion-dollar totals and percentage surges are best left out.

A practical habit follows from this: when a statistic arrives without a named source, a defined denominator and a stated measurement window, treat it as marketing rather than evidence, and ask the vendor presenting it to produce the underlying methodology before it goes into a board paper or a public statement.

Humans fare worse. iProov's research puts human detection of deepfakes at 0.1%, and separate work by Veriff and Kantar found people performing only slightly above chance when asked to identify manipulated visuals. Training staff to spot visual artefacts is not a viable control, because it asks people to perform a task they cannot reliably do. The asymmetry is also widening: detection improves incrementally, while generation quality improves with every model release. The practitioner conclusion is that detection is a supporting control, not a primary one — the primary control is process that makes the fraud fail even when the fake is convincing. The Ferrari executive did not detect anything; he simply required verification through a channel the attacker did not control.

Layer one — process controls. This is where most of the protective value sits, and it costs almost nothing. Out-of-band verification means confirming an unusual or irreversible request through a channel the requester did not supply, such as a call-back on a known number rather than one given in the message. Wire transfers, payee changes, credential resets, access grants and contract signatures arriving through an unusual channel should require this by policy. Pre-agreed passphrases or challenge questions for executive-level requests formalise the Ferrari method, because they require knowledge an attacker cannot synthesise. Removing urgency as an authorisation shortcut matters too: the attack mechanics are consistent — establish authority, manufacture urgency, restrict independent verification, push toward an irreversible action — and a policy that no genuine request is ever too urgent for verification removes the mechanism the attack depends on. Rehearsing deepfake scenarios in tabletop exercises and red team engagements is now standard practice at some security firms, which tells you the threat has moved from theoretical to operational.

Layer two — provenance. Rather than trying to prove a fake is fake, this makes genuine content provably real. Content provenance is a cryptographic record, embedded in media at capture or export, that verifies its origin and edit history. The relevant standards are C2PA, Adobe Content Credentials and Google SynthID. Adoption has reached meaningful scale: Google reported in May 2026 that SynthID had watermarked more than 100 billion images and videos plus 60,000 years of audio, with verification used 50 million times globally. The practical strategy is to tag all genuine corporate content, so unlabelled media claiming to be from you becomes automatically suspect — inverting the burden from proving a fake is fake to pointing at the absence of credentials. Two caveats apply: adoption is strongest in enterprise creative tools and weakest in open-source generation pipelines, so an attacker using open tooling produces unlabelled media without effort, and provenance strengthens evidence about origin without addressing the wider scam process on its own.

Layer three — monitoring and response. Content monitoring platforms scan for media referencing a brand, its executives or its products, and liveness detection integrated into identity verification workflows protects hiring and KYC processes from face-swapped candidates. None of these three layers substitutes for the others: monitoring catches what process and provenance miss, provenance gives monitoring something concrete to check against, and process is what keeps a convincing fake from causing loss in the minutes before either of the other layers has a chance to act.

The TAKE IT DOWN Act, signed in May 2025, criminalises certain publication of non-consensual intimate imagery including AI-generated material, and from 19 May 2026 covered platforms must provide a removal process and take down validly reported content and known identical copies within 48 hours. The FTC began enforcing the platform requirements and sent warning letters to a dozen websites on 20 May 2026. The DEFIANCE Act creates a federal civil cause of action for deepfake victims, with damages up to $150,000 per violation or actual damages, whichever is greater, and holds platforms liable where they fail to act. A brand cannot rely on either law directly for most commercial deepfakes, but both establish precedent and functioning notice-and-takedown infrastructure at platforms, which is what gets used when reporting brand impersonation. EU AI Act obligations phase in through 2027, adding transparency duties around synthetic content, and US state-level deepfake laws continue to expand.

Report through platform channels immediately, since the notice-and-takedown infrastructure described above is now subject to regulatory attention. Notify staff internally before communicating externally, because employees are a target audience for the fake, and a message explaining what happened and restating verification procedures prevents a secondary attack. Communicate to customers with specifics rather than a generic warning — naming the platform, the fabricated claim and the fact that the company does not solicit business that way is what makes a correction usable. Point to your verified channels, which only works if provenance and channel verification were established beforehand. Involve legal early for preservation orders and platform escalation, and debrief the process afterwards rather than the technology — the useful question is which control should have made the fraud fail, and whether it existed.

A multinational running finance operations across several countries has staff receiving instructions in a second or third language routinely. The subtle cues that might make a synthetic voice feel wrong — idiom, register, regional phrasing, hesitation patterns — are precisely the cues a non-native listener has least access to. A cloned voice speaking English to a team in Manila, Jakarta or Nairobi loses the anomalies a colleague in the CEO's home market might half-notice. Voice cloning quality also varies by language: cloning is generally strongest in high-resource languages with abundant training audio, so an executive with hours of public English speaking is more cloneable in English than in a language they rarely use publicly, which creates an underused control — verification in a language the attacker's model is less likely to handle well. Verification procedures need writing in every operating language, not translated as an afterthought, because a challenge-question protocol that exists only in the English employee handbook does not protect a finance team in another market. Organisations that produce synthetic media legitimately also carry a responsibility here: applying provenance discipline to their own output avoids contributing to an environment where unlabelled synthetic media looks normal, which is the environment attackers depend on.

Write the out-of-band verification policy for wire transfers, payee changes, credential resets and access grants as a single page. Establish executive challenge questions and brief the people who would receive such a call. Audit executives' public audio footprint, not to reduce it, which is impractical, but to know what an attacker has to work with. Start tagging genuine content with C2PA or Content Credentials, beginning with executive video and official announcements. Add a deepfake scenario to the next tabletop exercise. Write the response playbook before it is needed, including who preserves evidence and who authorises public statements, and translate all of it into every operating language. None of this requires a detection platform, and a specialist AEO and GEO provider can help fold verified-channel and provenance signals into how a brand's official presence is described across the web, which is part of the same defence covered under AI publisher deals and brand liability.

This is where Lifewood's own work sits, so the interest is worth declaring. Lifewood operates across 50+ languages with human-in-the-loop verification as the core discipline, and builds the multilingual speech and content data that underpins both generation and detection systems.