Deepfakes: How to spot them and protect your business
Deepfake impersonation accounted for 45% of AI-driven attacks in IBM’s 2026 Cost of a Data Breach research, more than any other type. Knowing how to spot deepfakes is worth a few minutes of your team’s time, but the habit that protects your money is simpler: verify any request through a channel you already trust.
The typical business version is a phone call. Your controller hears your voice, slightly rushed, saying you’re at the airport, a vendor needs a wire before end of day, and the details are coming by text. Widely available AI tools can clone a voice from recordings that are easy to find online, such as a video, a podcast or a webinar.
Talk to an advisor
Ask about your own setup.
We ask what you’re trying to accomplish and tell you what we’d look at first. It costs nothing to talk, and no price comes up on this call.
What a deepfake is
A deepfake is synthetic media, audio, video or images created or altered with machine learning so it appears to show a real person saying or doing something they never did. The same tools that power legitimate AI voice and video features can impersonate an executive, a client, a family member or a government official. In business, they show up as:
- 01Cloned voices in phone calls and voicemails, usually asking for a payment, a password or a multi-factor code.
- 02Fake video calls where an “executive” or “vendor” appears on camera to approve a transaction.
- 03Fake images and documents, such as AI-generated IDs, profile photos or invoices.
- 04AI-written messages that copy a person’s writing style, which our post on AI-powered phishing covers.
How criminals use deepfakes against businesses
The FBI has warned that criminals use AI-generated audio to impersonate people and request payments, generate video for real-time calls with supposed company executives or authority figures, and create fake identification documents for impersonation schemes.

$893M
in losses tied to more than 22,000 IC3 complaints involving AI in 2025, according to the FBI’s 2025 Internet Crime Report.
The FBI’s 2025 report describes how this shows up in practice:
- Business email compromise, with fake messages and cloned voices of a CEO or official asking for a wire transfer. Businesses reported more than $30 million in losses to BEC scams involving AI in 2025, out of more than $3 billion in total BEC losses.
- Fake job candidates using voice or video deepfakes in online interviews, where lip movement and audio don’t quite line up. The goal is often access to the company’s network.
- Impersonated officials. In 2025 the FBI warned of a campaign using AI-generated voice messages that claimed to come from senior US officials, aimed at gaining access to accounts and contacts.
IBM’s 2026 report points the same way: AI-driven attacks rose 56% over the prior year.
How to spot a deepfake
These signs still catch many fakes, and the FBI’s own lists include several of them:
| Where to look | Warning signs |
|---|---|
| Video and images | Faces that look slightly smooth, blurry at the edges or oddly lit; shadows that don’t match; distorted hands, teeth, eyes, glasses or jewelry; lips that drift out of sync; stiff head movement, unnatural blinking or a person who avoids turning sideways |
| Voice | Flat emotion, odd pauses or a delay before every answer; word choices the real person wouldn’t use; a caller who won’t answer a personal question or switch to video |
| The request | Urgency and secrecy (“don’t loop anyone else in”); a new phone number, messaging app or bank account; a push to skip the normal approval process |
The FBI says AI-generated content has advanced to the point that it is often difficult to identify. Treat these signs as a bonus. The request is usually a bigger giveaway than the pixels.
Controls that stop deepfake fraud
- Call back on a known number. Any request involving money, payment details, passwords, MFA codes or sensitive data gets verified by calling the person at a number you already have on file. The FBI recommends exactly this.
- Require two people for payments. Wires, ACH changes and new vendor bank details need a second approver, no matter who asks.
- Agree on a verification phrase for executives and finance staff, and change it periodically.
- Never share MFA codes. No legitimate IT team, bank or executive needs your code.
- Tighten hiring checks. For remote hires, verify identity documents and references independently, and keep new accounts on limited access until checks are complete.
- Limit what’s public. Long recordings of executives on websites and social media give attackers plenty of voice and video to work with.
- Train with real examples. Staff who have heard a cloned voice demo take the threat more seriously.
None of these needs special software. They are policy decisions, written down and followed every time, and they sit alongside the technical layers in our cybersecurity services.
Write it down if you keep patient, client or financial records
If you handle patient records, client funds, donor information or tax data, a deepfake that tricks someone into sending data or changing a payment can become a reportable incident. Your payment verification and identity-check procedures belong in your written security program and your compliance documentation, and in your annual staff training.
The same goes for AI inside your own walls. Our AI services include writing AI use policies and controlling unapproved AI tools, and our post on shadow AI risks covers the rest.
A cloned voice only works if one phone call can move money. Make the callback rule universal, then book a 15-minute call and we’ll help you put it in writing.
Frequently asked questions
What is a deepfake?
A deepfake is audio, video or an image created or altered with machine learning so it appears to show a real person saying or doing something they never did. In business, the common versions are cloned voices on phone calls asking for a payment or MFA code, fake video calls with an executive or vendor, and fake IDs or invoices.
How can you spot a deepfake?
In video, look for faces that seem too smooth or blurry at the edges, shadows that don’t match, lips that drift out of sync and stiff head movement. In audio, listen for flat emotion, odd pauses and word choices the real person wouldn’t use. The FBI says AI-generated content is often hard to identify, so the request itself is the bigger clue: urgency, secrecy and a new number or bank account.
How do you protect your business against deepfakes?
Verify any request involving money, payment details, passwords or MFA codes by calling the person back at a number you already have. Require a second approver for wires and bank detail changes, agree on a verification phrase for executives and finance staff, and never share MFA codes. None of this needs special software. Write it into policy and follow it every time, whoever seems to be asking.
What should we do if we think a call was a deepfake?
Hang up, contact the real person through a number you already have, and tell your IT or security contact. If money moved, call your bank immediately to request a recall and file a report at ic3.gov.
Sources
- FBI IC3, 2025 Internet Crime Report
- FBI IC3 PSA, Criminals Use Generative Artificial Intelligence to Facilitate Financial Fraud, December 2024
- FBI IC3 PSA, Senior US Officials Impersonated in Malicious Messaging Campaign, May 2025
- IBM, Cost of a Data Breach Report 2026
- Baker Donelson, Ten takeaways from IBM’s 2026 Cost of a Data Breach Report
Next step
Let us handle this for you.
We keep staff off AI tools nobody approved and steer them to a tool we can secure.
- 15 minutes
Tell us which question brought you here and how your office runs.
We’ll tell you how it applies to you and what we’d look at first.
If we’re not the right fit, we’ll say so.
