Tuesday, 4:40 p.m. The controller at a food distributor in Mississauga gets a call from the company president, who's on the road in Montréal. It's his voice, his pace, the way he says "listen, this is important." He needs a $48,000 wire transfer to close a confidential acquisition before 5 p.m., and she's not to mention it to anyone.
The president was in a meeting two blocks away. His voice had been generated from interviews and videos he'd posted online.
That's an audio deepfake, or cloned voice. The story is fictional, but the technique is real, and you can't beat it by listening harder. You beat it with process.
TL;DR
- A voice no longer proves identity: in June 2025, the Canadian Centre for Cyber Security and the Canadian Anti-Fraud Centre flagged AI-generated voice messages mimicking senior officials, with business executives among the targets.
- The most reliable defence: hang up and call back on a number you already know, never the one the caller gives you or the one on your screen.
- For money: dual approval, and no "urgent" exceptions, not even for the president.
- A code word shared by finance and leadership adds a simple check.
What is deepfake voice fraud?
Deepfake voice fraud is a scam in which the caller uses an AI-generated voice to pass as someone you know: an executive, a supplier, a colleague.
The Competition Bureau of Canada defines deepfakes as realistic-looking fake images, videos or audio impersonating real people. The raw material is often public: interviews, podcasts, webinars, social media videos.
The call is all the more convincing because the number displayed can be faked. As the Canadian Anti-Fraud Centre points out, scammers use spoofing to make a call appear to come from a trusted number. This is the old CEO fraud, moving from email to the phone.
What Canadian authorities say
Canadian authorities treat deepfakes as an active threat, not a curiosity.
- Canadian Centre for Cyber Security. Its National Cyber Threat Assessment 2025-2026 notes that threat actors use generative AI to "create realistic audio and visual content impersonating trusted individuals," which helps them establish legitimacy and persuade targets.
- Joint advisory, June 23, 2025. The Cyber Centre and the Canadian Anti-Fraud Centre describe a campaign of text messages and AI-generated voice messages impersonating senior officials. Targets include business executives. Warning signs: urgency, requests to move to another messaging app, suspicious links, and a tone that seems out of character.
- Competition Bureau, March 9, 2026. It warns that scammers use voice cloning and AI-generated messages to impersonate government officials.
The best-known case happened abroad. According to CNN (May 2024), a finance employee at engineering firm Arup in Hong Kong sent about HK$200 million (roughly US$25 million) in 15 transactions after a video call in which the chief financial officer and other colleagues were deepfakes. Arup confirmed that fake voices and images were used.
The scenarios that target small businesses
In a small business, a cloned voice is mostly used to get money, access or an exception to procedure, by leaning on urgency and authority.
| Scenario | Typical request | Verification |
|---|---|---|
| Fake president | An urgent, confidential wire transfer | Call back the executive's known number, then dual approval |
| Fake supplier | New banking details | Call the number in the vendor file, never the one from the call |
| Fake employee calling the help desk | A password or MFA reset | Call back the number in the HR file and get manager approval |
| Fake technician calling an employee | An MFA code, or installing remote access software | Nobody gives out a code over the phone, ever |
| Voice message from an executive | Continuing the conversation on another messaging app | Call back through the usual channels |
The fake supplier scenario is covered in detail in our article on vendor banking change fraud.
The defence: a process that doesn't depend on the voice
The only reliable defence is to make the voice insufficient: no sensitive request should be carried out on the strength of a phone call alone, however familiar the voice.
The June 2025 joint advisory puts it this way: verify independently by contacting the person or organization "through alternate, previously confirmed channels." As internal rules:
- Mandatory call-back to a number you already have (staff directory, vendor file), never the one the caller gives or the one displayed.
- Dual approval for any transfer above a set threshold, and for any change of banking details.
- Waiting period for a new payee, for example until the next business day.
- Code word between finance and leadership, never written in an email, changed regularly.
- Written commitment from leadership: "I will never ask you to bypass the process, even in an emergency."
- Help desk: no password or MFA reset based on a phone call alone.
- MFA codes: never shared over the phone.
The code word isn't a movie trope: in December 2024, the FBI advised families to create a secret word or phrase to verify a caller's identity. At work, it complements the call-back; it doesn't replace it.
Leadership's commitment matters most. Scammers count on fear of displeasing the boss. A controller who knows she'll be thanked for hanging up on the "president" won't give in.
Train your team to hang up
Learning to hang up without embarrassment matters more than learning to spot a fake voice, because the best clones leave no audible clues.
The joint advisory lists a few things to listen for: a flat tone, unexpected background noise, choppy sentences. Useful, but not enough. Build reflexes instead:
- any urgent, confidential or unusual request triggers the process, no matter who's calling;
- "I'll call you right back" is always an acceptable answer;
- nobody gets blamed for checking.
Add a fraudulent call scenario to your phishing simulations, especially for finance, payroll, reception and the help desk.
The FBI also suggests limiting public content featuring your voice. A small business can't erase the president's interviews, which is exactly why the process has to hold even when the voice is perfect.
Finally, a fake call can be followed by a "confirmation" email that appears to come from your own domain. Check that your domain is protected against spoofing with our DMARC, SPF and DKIM checker.
If you think you've been targeted
Hang up, call the real person back on a known number and, if money has left, call your financial institution immediately: every hour counts when trying to stop a transfer.
Then:
- Report the incident to your local police, the Canadian Anti-Fraud Centre and the Canadian Centre for Cyber Security, as the joint advisory recommends.
- Write down the number displayed, the time, the length and content of the call, and keep any voicemails and texts.
- Warn your staff: the scammers may try other people in the organization.
- If personal information was disclosed, assess whether it's a privacy breach you must record or report (in Quebec, a confidentiality incident under Law 25), and confirm with legal counsel.
- Review the step that failed, without looking for someone to blame.
Want to know whether your payment and help desk procedures would hold up against a well-imitated call? Our cybersecurity audit reviews them with your team. To talk it through first, book a free consultation.
Frequently asked questions
Can AI really clone someone's voice?+
Yes. The Competition Bureau of Canada defines deepfakes as realistic-looking fake images, videos or audio impersonating real people, and warns that scammers use voice cloning. In June 2025, the Canadian Centre for Cyber Security and the Canadian Anti-Fraud Centre also warned about AI-generated voice messages mimicking senior officials, with business executives among the targets.
How can you tell if a call uses a cloned voice?+
You can't rely on your ears. Canadian authorities mention a few clues, such as a flat tone, unexpected background noise or choppy sentences, but a good clone may show none of them. The only reliable check is to hang up and call the person back on a number you already know.
Does a code word actually help against deepfakes?+
Yes, as an extra layer. In December 2024, the FBI advised families to create a secret word or phrase to verify a caller's identity. At work, a verification phrase known only to finance and leadership, never written in an email and changed regularly, adds a simple barrier. It doesn't replace call-backs or dual approval for payments.
What should we do if we sent a wire transfer after a fake CEO call?+
Call your financial institution right away to try to stop or recall the transfer, then your local police. Report the incident to the Canadian Anti-Fraud Centre and the Canadian Centre for Cyber Security. Write down the number displayed, the time and what was said, and warn your team, since others may be targeted next.
Sources
- 1.Canadian Anti-Fraud Centre: Joint Advisory, Cyber officials warn Canadians of malicious campaign to impersonate high-profile public figures (June 23, 2025)
- 2.Competition Bureau Canada: Watch out for AI-generated government impersonators (March 9, 2026)
- 3.Canadian Centre for Cyber Security: National Cyber Threat Assessment 2025-2026
- 4.Canadian Anti-Fraud Centre: Emergency scams
- 5.FBI IC3: Criminals Use Generative Artificial Intelligence to Facilitate Financial Fraud (December 3, 2024)
- 6.CNN: British engineering giant Arup revealed as $25 million deepfake scam victim (May 16, 2024)