Deepfake Fraud in the Contact Center: Verifying Humans in an AI World

The voice on the phone knows the account number. It knows the last four of the Social. It sounds stressed, apologetic, a little rushed; exactly like a real customer whose card was just declined at a restaurant. 

Except it’s not a customer. It’s a model, trained on 30 seconds of audio scraped from a podcast appearance, a company webinar, or a social media post. 

And your agent has about 11 seconds to decide. 

This is the new reality for contact centers, where deepfakes—impersonations of real customers—are becoming increasingly convincing, and more difficult to spot. 

Here’s the part many advisors are overlooking: the contact center has quietly become one of the softest targets in the enterprise, and almost nobody is selling it that way. 

Deepfake Fraud is Outpacing Defenses 

Deepfakes have quickly emerged as a top security threat facing contact centers. What seemed impossible a few years ago can now be accomplished by cybercriminals with alarming speed and efficiency. 

To put this shift into perspective: 

• In a recent Gartner survey of 302 organizations, 62% reported facing a deepfake attack in the prior 12 months, where an actual synthetic voice, video, or identity artifact was used against them. 

• A separate study from Utah Valley University found that it’s becoming harder to detect deepfakes. In the study, only 16% of participants could successfully differentiate between a real video and a fake one.  

• Entrust, looking across more than a billion identity verifications, found that deepfakes now drive roughly one in five biometric fraud attempts. And injection attacks, where a fraudster bypasses the camera or microphone entirely and feeds synthetic media straight into the verification stream, climbed 40% year over year. 

The third stat matters because injection attacks mean the attacker never holds a phone up to a screen. There is no “hold your head to the left” liveness to verify. The fraudulent media is delivered directly into the pipeline. A verification stack that was designed to catch a bad actor in a room is now being attacked by a bad actor in an API. 

Every Technology Moves the World Forward. And Opens a Door for New Threats. 

We have watched this pattern before. For example, cloud gave us speed and scale, and it also gave us multiple years of breaches while security caught up. AI is running the same play, only faster. 

Here is the specific irony in the contact center. Over the last three years, our clients invested heavily in CX modernization: voice biometrics, conversational IVR, AI-assisted agents, and frictionless authentication. Every one of those investments was made to remove friction for the customer. 

Fraudsters simply took the same tools and pointed them in the other direction. 

Voice authentication was the crown jewel of frictionless CX. It’s now, in a growing number of environments, a liability. Knowledge-based authentication was already compromised by a decade of breach data sitting on the dark web; synthetic voice finished the job. 

I remember this feeling from my time as a CISO. When my environment was fully on-prem, I could see the perimeter. Cloud made the perimeter invisible. And now with edge compute, we have decentralized our infrastructure to an ever-expanding attack surface. The security team is accountable for securing what they cannot see. Clients are managing uncertainty layered on top of complexity. 

In uncertainty, we return to what we know. The one constant in any environment is identity. Thus, identity becomes the new perimeter, and it can now be manufactured for the price of a streaming subscription. 

What Identity Verification Looks Like in 2026 

Most leaders don’t want more tools. They want to know whether the person on the line is real.  

That requires moving from a single check to layered signals: 

• Device and network intelligence: Is this call originating from the number and device this customer has always used, or from a spoofed ANI over a VoIP gateway registered last Tuesday? 

• Injection detection, not just liveness: Can the platform tell the difference between a real camera feed and a virtual one? Liveness detection alone no longer clears the bar. 

• Behavioral and conversational signals: Cadence, hesitation, navigation patterns, and how the caller responds to an unexpected question. 

• Step-up authentication tied to risk, not to script: Low-risk balance inquiry moves fast. A high-risk request—a wire change, a port-out, a beneficiary update—triggers a stronger factor automatically. 

• Agent enablement: Agents need a documented, blameless path to slow a call down. If the only metric an agent is measured on is handle time, fraud will win every single time. 

That last point is not a technology gap. It is a policy gap, and it is often the fastest thing to fix. 

If you want evidence that the gap is rarely the platform, look at the Call Center Power win Mike Baillargeon covered on Telarus’s Inside the Win. A home security company in Charlotte already owned Genesys Cloud CX integrated with Salesforce Service Cloud Voice, a real investment, properly sold and installed. But the platform was under-leveraged. Operations were running manual workarounds, and there was no structured support model or improvement roadmap in place. 

As Call Center’s SVP of Channel James Cammareri put it, the technology was there. The expertise and discipline to extract value from it were not. 

That is precisely the condition fraudsters will exploit. Not an absence of tools, an absence of ownership. 

The Advisor Play: Why Contact Center Fraud Prevention is a Two-Practice Deal 

This is where the opportunity gets interesting, and where most advisors leave money on the table: 

Caller verification does not belong to the CX practice or the security practice. It sits between them.  

The CCaaS platform owns the customer journey and the routing logic. The security stack owns identity, fraud analytics, and the risk engine. Neither supplier solves it alone, and neither one will tell you that unprompted. 

So bring both. 

When you scope a contact center refresh, add a fraud and identity layer to the conversation. When you scope an IAM or fraud-prevention engagement, ask who owns the voice channel. You are not stapling two quotes together, you are solving one problem the client has been describing in two separate meetings to two separate teams. 

The commercial effect is straightforward. A CCaaS deal you were competing for on seat price becomes a risk conversation with a different budget line, a different sponsor, and a materially larger scope. Fraud loss and regulatory exposure get funded differently than seat licenses do. 

On a previous Inside the Win, Turning an Exchange Migration Into an $8K Cybersecurity Upsell Opportunity, we walked through a deal that started as nothing more than an end-of-life Exchange 2008 migration to Microsoft 365. Routine infrastructure work. One-time professional services engagement. 

We asked a few simple discovery questions. What are you doing for email security? When someone clicks, what happens? 

They didn’t know. There was no endpoint protection in place. No 24/7 SOC monitoring. No real sense of what the blast radius would be. That discovery turned a one-time migration into professional services plus ongoing MSP and MSSP support; just under $8,000 MRR on a three-year term, with ECI winning the business. 

The conversation framing is the part I’d underline.  

Advisors are hungry to know how to start a cybersecurity conversation. But as it turns out, you usually don’t have to start one. You’re already in one. You just have to ask the next question. 

Now run that same play on the voice channel. “How does a caller prove they are who they say they are?” is the contact center version of “when someone clicks, what happens?” 

Discovery Questions That Open the Door to Contact Center Security Deals 

These are not technical questions. They are leadership questions. 

• How do your agents verify a caller today, and when was that process last updated? 

• If a synthetic voice passed your authentication tomorrow, how would you know? 

• Who owns fraud in the voice channel: the contact center team, or security? 

• What’s your agent’s authority to stop a transaction they think is fraudulent? 

• Are you seeing more account takeover attempts through the call center than you were a year ago? 

That third question is the one I’d lead with. In most organizations, the answer is a pause, because nobody owns it. That pause is your opening. 

Your Moment Is Now: Turn Contact Center Security Concerns into Revenue 

Contact center fraud is a rare kind of opportunity: it’s measurable, it’s urgent, and it has a budget owner who is already worried about it. 

Your clients have spent years making it easier for customers to reach them. They now need help making sure the people reaching them are real, without unraveling the experience they just paid to build their CX environments. That tension is exactly where a trusted advisor earns the seat. 

You don’t need to be a deepfake expert to start this. You need to ask who owns verification in the voice channel, listen to the answer, and bring the right two suppliers to the table. 

Reach out to one CX client this week with a single question: 

“How would you know if the voice on the other end of that call wasn’t a person?” 

Then listen. Loop in your Telarus CX and Cybersecurity teams to help you build the response. 

You’re more ready than you think.