Social Engineering
Deepfake CEO Fraud and Voice Cloning: How to Stop the Call That Sounds Exactly Like Your Boss
A few seconds of public audio is enough to clone a voice. Spotting a fake is no longer realistic; verifying the request is.
Generative AI has made convincing voice and video impersonation cheap. A few seconds of audio from a conference talk, podcast or social video can be enough to clone an executive's voice. Attackers combine it with research on your org chart to call finance, the help desk or a new joiner with an urgent, plausible request.
How the attack usually runs
- Research: LinkedIn, press releases and vendor lists reveal who approves payments and who your suppliers are.
- Pretext: a confidential acquisition, an overdue supplier, a locked-out executive travelling abroad.
- The call: a cloned voice or a deepfake video meeting adds authority and urgency.
- The ask: a payment to a new account, a password or MFA reset, or a copy of sensitive data.
Why training people to spot fakes is not enough
Detection quality is improving, but so is generation. Industry forecasts for 2026 openly say deepfakes will be impossible for people to spot reliably. The durable defence is to make the process, not a person's judgement of a voice, decide whether money moves or access changes.
Controls that work
| Request | Control |
|---|---|
| New or changed bank details | Call back on a number already on file, never one in the message; dual approval |
| Urgent executive payment | Out-of-band confirmation plus a second approver, regardless of seniority |
| Password or MFA reset | Identity verification the caller cannot fake from public data; no resets on inbound calls |
| Confidential project | A pre-agreed code phrase for genuinely sensitive requests |
Run a realistic exercise: a simulated voice-based pretext against finance and the help desk tells you more than any slide deck.
Test your people and processes safely
We run controlled social engineering assessments, including vishing, and help you redesign approval flows.
Explore social engineering testingFrequently asked questions
What is deepfake CEO fraud?
It is a scam where criminals use AI-generated voice or video of an executive to persuade staff to make payments, share data or reset credentials.
Can deepfake voices be detected?
Sometimes, but not reliably. Verification processes such as call-backs to known numbers and dual approval are more dependable than trying to judge whether a voice is real.
Is deepfake fraud a form of business email compromise?
It is closely related. It uses the same pretexts and goals as BEC, delivered by voice or video instead of, or alongside, email.
PITOWINGS Research
OSCP, CEH and ECES certified testers, SOC analysts and certified forensic investigators at PITOWINGS.
Keep reading
All articlesPITOWINGS · Insights
AI Security
11 min read
Agentic AI Security in 2026: Prompt Injection, MCP Servers and the OWASP Agentic Top 10
AI agents now read email, call APIs and change records on their own. The attack surface is no longer the prompt; it is everything the agent can see, call and remember.
PITOWINGS · Insights
Cryptography
10 min read
NIST IR 8547 Is Final: RSA and ECC Deprecated After 2030 — Your Post-Quantum Migration Plan
The quantum deadline now has a date on it. Here is what NIST's final transition plan changes, and a practical migration path that starts with knowing where your cryptography is.
PITOWINGS · Insights
Security Operations
9 min read
AI in the SOC: What AI Agents Can and Cannot Do for Threat Detection in 2026
AI can read every alert. It still cannot decide, on its own, whether to take your payment system offline at 3 a.m.


