--- title: Trust is under attack: Inside the world of deepfakes, with Andrew Austin source_page: https://www.sardine.ai/media/fraud-forward/episodes/deepfake-scams canonical: https://www.sardine.ai/media/fraud-forward/episodes/deepfake-scams format: text/markdown date: 2026-08-05T11:00:00.000Z description: Deepfake scams are changing fraud by turning protection tools into risks. The vulnerabilities financial institutions face as voice cloning fraud expands... --- **Quick links:** [Human page](https://www.sardine.ai/media/fraud-forward/episodes/deepfake-scams) · [Home](https://www.sardine.ai) · [Customers](https://www.sardine.ai/customers) · [Blog](https://www.sardine.ai/blog) · [Demo](https://www.sardine.ai/demo) --- # Trust is under attack: Inside the world of deepfakes, with Andrew Austin **Published:** 2026-08-05T11:00:00.000Z Deepfake scams are changing fraud by turning protection tools into risks. The vulnerabilities financial institutions face as voice cloning fraud expands... What’s up fraud fighters, and welcome back to Fraud Forward! This episode opens a little differently. And if you are not paying close attention, you will probably believe it was me. It sounds like me, it looks like me, but every bit of it is a deepfake created with artificial intelligence in about twenty minutes for less than five dollars. That is not a future threat. That is today. Andrew Austin is one of my favorite people to talk to because he does not do hype. He spends a lot of time exploring where emerging technology meets fraud risk. And this is one of the most practical conversations I have had on the show. We are not here to scare anyone. We are here to make sure fraud fighters understand exactly what they are up against and how organizations need to start thinking differently about deepfake fraud and AI voice cloning. We walk through the full landscape of deepfake technology together, from voice cloning to live face swapping to lip sync. We talk about what each one costs, how convincing each one is, and where criminals are already using deepfake scams today. Andrew breaks down why identity verification is the most vulnerable point for financial institutions right now, why real time detection of AI-generated content is still an unsolved problem, and why fraudsters do not need Hollywood quality deepfakes to be effective. They just need to be convincing enough. We do not stop at the threat. We spend the second half of the conversation talking about defenses, device intelligence, behavioral biometrics, consortium data, friction design, and why employee awareness and organizational culture still matter as much as any technology investment. Andrew also shares what excites him about where AI is headed on the fraud fighting side, and what he thinks every CISO should already have on their roadmap to combat AI fraud. Fraud has always been built on manipulating trust. AI does not change that. It just gives criminals a far more convincing disguise. What you’ll hear in this episode: How Andrew created a convincing deepfake in twenty minutes for less than five dollars, and what that means for fraud teams The difference between voice cloning, live face swapping, and lip sync technology, and the distinct fraud risk each one creates Why deepfake scams do not need to be perfect to work, and how criminals are already exploiting that reality Where financial institutions are most vulnerable to AI-generated content right now, and why onboarding and IDV top the list Why real time detection of AI-generated video content is still not a solved problem How device intelligence, behavioral biometrics, and consortium data work together to close the gaps IDV alone cannot cover Why employee awareness and organizational culture are just as important as any technology investment What Andrew thinks every CISO should already have on their roadmap to combat AI fraud Where AI is creating real operational efficiency gains on the fraud fighting side, and why Andrew is genuinely excited about it You should listen to this episode if you: Lead fraud, risk, compliance, or financial crimes at a bank, credit union, fintech, or neobank Are responsible for identity verification, onboarding design, or authentication strategy Want a practical, non-sensational breakdown of what deepfake scams and AI voice cloning actually look like today Are building or updating your fraud program and want to understand where AI fraud risk fits into your existing controls Are new to fraud and want to understand how emerging technology is reshaping what fraud fighters need to know ### Episode notes What you just heard was a scam The episode opened with a deepfake. Andrew cloned a voice from a thirty second clip, synced it to an existing video, and produced something convincing enough after just one round of adjustments. The point was not the trick. It was the demonstration. Seeing and hearing it lands differently than being told about it. Voice cloning is cheap, fast, and already being used A thirty second audio clip is enough to clone a voice. From there, pitch, speed, and variability can all be tuned until the output is convincing. Andrew has heard credible reports of voice clones bypassing IVR-based voice authentication at financial institutions, the kind where your voice is your password. Vendors in this space have evolved, but anyone claiming to be perfect should be questioned. Live face swapping exists but has real limitations Live face swapping requires significant GPU power, usually a high-end gaming machine, and works best when the source and target have similar facial geometry. Where it falls down is mismatched face shapes, hair, beards, and build. For that reason it is not as widely used in scams as other deepfake methods. The barrier to entry is higher, and the margin for error is more visible. The more significant risk right now is animating a still image into a convincing video, which is far simpler and cheaper. Onboarding and IDV are the most vulnerable points For the past five to ten years, identity verification has relied heavily on document scans and liveness checks. Turn your head left, turn your head right. Andrew makes the case that this approach is increasingly vulnerable because real time detection of AI-generated video content is not yet reliable, and fintechs competing on frictionless onboarding are inadvertently making themselves easier targets. The answer is not just better IDV. It is layering in device intelligence, behavioral biometrics, and consortium data to build a more complete picture of who is actually on the other end. Fraudsters do not need perfection, they need enough The creator's purpose drives how much time and money goes into a deepfake. Scamming strangers who have never seen or heard you is easy. Bypassing a voice authentication system requires more precision. The key insight from this conversation is that criminals are not trying to win awards. They are trying to get paid. Good enough is good enough, and the trust vector, the voice, the face, the urgency in a message, does the rest of the work. Technology alone does not solve this Andrew's father was scammed out of twenty thousand dollars over the phone, no AI involved, by someone who knew how to manufacture urgency and emotional pressure. The point landed clearly. Fraud teams can invest millions in technology and still lose to a well-trained human scammer if the frontline culture, training, and procedures are not there. Behavioral biometrics and device intelligence are critical. So is a teller who asks why an 82-year-old is withdrawing cash for the first time. Where AI helps fraud fighters Andrew is genuinely excited about what AI can do on the defense side. Not replacing investigators, but giving them better tools. Consolidating alerts around related parties, surfacing connections that would take thirty minutes of manual work to find, and bringing everything into a single view so investigators can spend their time investigating instead of clicking between twenty screens. He sees the greatest near-term opportunity in operational efficiency, not automated decisioning. ### Key takeaways Deepfake scams do not require Hollywood quality. They require enough trust to get someone to act. Voice cloning from a thirty second clip is cheap, fast, and already being tested against financial institution controls Onboarding and IDV are the most exposed parts of a financial institution's fraud stack right now Real time detection of AI-generated video is not yet a solved problem. Layered controls are the answer Device intelligence, behavioral biometrics, and consortium data are essential complements to IDV Employee awareness and organizational culture are just as important as any technology investment AI is creating real efficiency gains on the fraud fighting side, particularly in investigation workflows Fraud has always been built on manipulating trust. AI gives criminals a more convincing disguise, not a new playbook ### Final takeaway Today's conversation was not really about artificial intelligence. It was about trust. Fraud has always been built on manipulating trust. AI does not change that. It just gives criminals a far more convincing disguise. If the opening of this episode made you a little uncomfortable, good. That is exactly how your customers are going to feel if we do not start preparing for this now. Episode resources & links: Connect with Andrew Austin | LinkedIn Fraud & Risk SME Product Strategy & Solutions, Sardine Connect with Hailey Windham, CFCS | LinkedIn Host of the Fraud Forward Podcast Banking Community Lead at Sardine Certified Financial Crimes Specialist (CFCS) 2023 Credit Union Rockstar, CU Magazine Continuous Improvement Award, SAFE Federal Credit Union, 2023 Top 20 Professionals Under 40, The Sumter Item, 2022