AI-Powered Romance Scams: Incode's CEO Ricardo Amper on AARP's The Perfect Scam
Incode
June 3, 2026•
.png)
June 3, 2026•
Most people assume they could spot a deepfake video. They’re sure they’d notice a glitch in the motion, a voice that sounds distorted, or something that just feels off. But the technology has moved far beyond our perception.
Today, deepfakes can sustain a convincing hour-long video call. AI agents can even carry on a “romantic relationship” for months, responding naturally, building trust, and eventually asking for money. When the scam finally unravels, it’s not obvious who to point a finger at, yet the financial and emotional damage is very real.
That is exactly what happened to David. He met Bonnie online, fell for her over seven months of video calls and messages, wired her money to help with a contract job in Australia, and eventually lost nearly $50,000 to what investigators believe was an AI-generated persona. He never met her in person. But the videos looked completely real.
In the latest episode of AARP's The Perfect Scam, Ricardo Amper, Founder & CEO of Incode, joins host Bob Sullivan to talk about what David experienced, why it is becoming more common, and what can actually be done about it.
During the episode, Ricardo and Bob discuss why romance scams work, how to spot the technology behind them, and what both individuals and companies can do to protect themselves.
Ricardo explains how biology works against us, priming us to trust what we see and hear. Deepfakes now exploit that at scale, with no visible tells.
"Biology taught us to believe the person we are looking at and the voice we are hearing. This is not really about instinct. It is about how technology is evolving in a way that humans can no longer detect it,” said Ricardo Amper, Founder & CEO of Incode.
Two years ago, lip sync issues gave deepfakes away after a few seconds. That problem has now been solved. Attackers can sustain a fully convincing live video conversation for an hour or more.
Romance scams are already a billion-dollar industry. Ricardo warns that as AI agents replace human operators in fraud rings, the volume of attacks will increase dramatically at near-zero cost for the fraudster.
"Before, there was some limit to the scale of these attacks because they required real human operators. That limit no longer exists. A million AI agents can be deployed to trick people simultaneously, at almost zero cost,” said Ricardo Amper, Founder & CEO of Incode.
Ricardo also reminds listeners that shame, while natural, shouldn’t stop victims from reporting these types of scams as soon as possible.
Platforms that implement identity verification (IDV) are the safest places to meet people online, according to Ricardo. Continuous identity verification helps to establish that someone truly is who they claim to be. And these models are specifically trained to catch the deepfake artifacts that humans no longer reliably notice.
For example, Ricardo walks through how Incode’s defensive AI models, trained on millions of generated deepfakes, detect pixel-level inconsistencies invisible to the human eye, in real time, at the point of IDV.
During the conversation, Ricardo implored consumers to stay sharp when interacting with new individuals online. Ricardo advised consumers to:
"It is going to get very difficult in the next two or three years. But in ten years, I believe we will live in a much safer and more productive world,” said Ricardo Amper, Founder & CEO of Incode.
More platforms for listening to the episode are available here.
[Bob]
All they need to rob you is a computer and often a story. But now there are so many tools available to make those stories far more realistic. We have no way of knowing if the videos that David received were AI generated, deep fakes or something else. But we wanted to talk about the possibility with an expert. So we have Ricardo Amper here today. He is the founder and CEO of Incode Technologies, which builds software to detect AI fakes. I asked him first about David's situation.
[Ricardo Amper]
It is very unfortunate, but this is actually very common and online romance scams are happening more and more. It's a $1 billion industry right now, and we think it's gonna be a lot more going forward.
[Bob]
One thing to note about David's story is that after that initial video call, most of the other videos that she sent were one way clips. In other words, not interactions. That's a common tactic. Something I've heard from other people who are victims of romance scams using fake videos is this deal where the criminals generate a few moments of video, enough to be persuasive, and then, oh, the Internet's going out, let's switch to chat. Tell me why that happens.
[Ricardo Amper]
That happened two or three years ago because the technologies were not good enough to be able to convincingly sync your voice to the lips. So what a deepfake is, is AI generated video that has to match your lip, has to change your voice, and has to look like the identity they're trying to portray. And two years ago, you could pull that off for just a little bit of time and then there would be very clear tell signs that your voice was not matching with your lips.
[Bob]
So if someone tries to send you a series of short video clips and offers up excuses to refuse a video call, well that's a sign of trouble. But unfortunately, warnings like that are becoming outdated.
[Ricardo Amper]
But the scary thing is that now you can sustain an hour conversation and you cannot tell. So that tell sign, which happened to us, which is it's not gonna work for a long time, that is now a solved technical problem. And now you can have an hour conversation and not even know.
[Bob]
There was a time when we speculated that there would be these deepfakes, or people would create very realistic video versions of lovers or executives or whatnot, but it was hard to really spot them out in the world. But when you guys contacted me, you referenced that there were 2,000 documented incidents of deepfakes being used in crimes broadly speaking. Has this kind of deepfake video crime now come of age?
[Ricardo Amper]
Deepfakes in general are becoming the number one attack vector in many industries. And what we're seeing is that at least right now, one third of all the attacks that we're defending are deepfakes, where people, or systems, are portraying to be someone they're not and they're trying to fool banks and online sites into thinking there's someone else. They're opening accounts in other people's names, or taking over bank accounts they don't own.
[Bob]
The most alarming part of this story is that criminals are using our very human nature against us.
[Ricardo Amper]
It's scary because biology taught us to believe the person we are looking at and the voice that we're hearing. And this is not really about your instinct. It is more about how technology is evolving in a way that humans cannot even detect it.
[Bob]
Years of evolution have told us: here's how you know that someone's real. And it's by seeing them, their lips move, by hearing their voice. This is a really big change for people to absorb in a very short span of time, that we're going to have to change the way we verify who someone is, right?
[Ricardo Amper]
And that's the problem, right? Because it is not about how sophisticated you are. It's more about making sure that the companies or the sites that you're dealing with adopt this AI technology to verify your identity. And the problem with this is that we work with Homeland Security as well. Even them, they have issues when, for example, pilots are verifying their identity to fly private and all of this is done in a digital way. There's very little they can do unless they adopt some of these new AI technologies.
[Bob]
Deepfake attacks are becoming so common that, well, Ricardo himself was a victim.
[Ricardo Amper]
One of the first ones that we saw was actually a deepfake attack directed at our company. So our head of Treasury, the person that moves the money in the company, got a WhatsApp message from a fake me saying, "Andreas, we need to do an acquisition and I need you to send a million dollars to fund an account so that we can secure this initial acquisition." And of course the person said, let's go to a video, and they went to a Zoom call. Looking at an actual live video, there was a fake me talking about how we needed to do this very discreet acquisition. And it was hyperrealistic. There was no tell sign. What happened was three minutes into the call, it started getting choppy as if you had bad wifi, and then the person said, let's go into a different video. They went into a Teams meeting and the same thing, a very realistic person trying to explain, portraying to be me, trying to explain what we were trying to do. Eventually it went to a phone call, and our treasurer was convinced it was actually me asking him to wire money. At the very end, the reason why he didn't was because I never called him for that. It's really the CFO. So he calls the CFO, and the CFO said, I didn't know about that. They ended up sending someone to my house as I was working from home. And that's how we first saw some of the first deepfakes.
[Bob]
It was very brazen for someone to attack your company this way, wasn't it?
[Ricardo Amper]
It was, yes. That required a lot of guts to do that. But the truth is that it's extremely common.
[Bob]
Just to give you an idea of how realistic these things can sound, we asked Incode and Ricardo if they could send us an example of a deepfake phone call. And they did. What you are about to hear is a recording of Jorge Braniff, VP of Fraud and Operations at Incode, talking with a bot impersonating Ricardo. As you listen, think of this: Jorge told us this is a simple agent trained on a single, short voice clip. The agent was provided with a general prompt to ask for a bank transfer. Nothing was scripted; the agent is making up the words as it goes along.
[AI Bot]
I'm calling today regarding a rather urgent matter that requires your team's immediate attention.
[Jorge Braniff]
What do you need?
[AI Bot]
I need to request a money transfer for a confidential purchase. This is a strategic project that requires immediate action. Could you please let me know what information you need from me to process this transfer?
[Jorge Braniff]
How much do you need for this transfer?
[AI Bot]
I need a transfer of $500,000. This is for a critical, time-sensitive acquisition that will significantly benefit the company.
[Jorge Braniff]
Okay, thank you. I'll let you know when it's done.
[Bob]
Ricardo also told us about a scary incident involving a set of companies in Las Vegas, and the key to that attack was a realistic looking deepfake video.
[Ricardo Amper]
One of the things that happened about a year ago that was really scary was in Las Vegas. Someone was successful at taking an IT manager's identity. The person managing the system for one of the biggest hotel chains was targeted using a live deepfake, meaning the same type of attack we had, where you can see a person you're interacting with live through a video Zoom. The person was saying, look, I lost my phone, I lost my computer, I need to get access to this. They got into a Zoom call with one of their colleagues, gave them access, and that person was actually a highly sophisticated hacker who then took over that system in Las Vegas. For two weeks, essentially half of Las Vegas stopped. They blocked the system until they got ransom, and eventually they got paid more than 30 million in ransom. The loss was in the hundreds of millions.
[Bob]
If all this isn't alarming enough, there's something even more dramatic that's going to happen in the scam world with AI. All those banks of employees who serve as frontline callers and texters making initial contact in many scams, well, they're going to be replaced by AI agents. That will really allow criminal gangs to dramatically scale up their attacks at virtually no cost.
[Ricardo Amper]
And the scary thing is, usually scams were made by humans. We think by next year, half of all the scams in the world in identity, not just in romance, but in every part of identity, is gonna be created by some of these gen AI tools that, by the way, cost almost zero money to make. And what is also changing is that it used to be that you need a live human doing this deepfake, tricking you. But now with the advances of gen AI, this is now starting to be done by AI agents who are responding to you in real time, who can reason like you can reason with a ChatGPT or a Claude. So you're talking to that agent. The agent is reasoning, answering back to you. What it means is that at least before, there was some limit to the scale of the attack because you needed real human beings to do that. And now there's not gonna be a limit. There can be a million agents deployed trying to trick people, and that costs almost nothing.
[Bob]
So what can you do about all this? We showed Ricardo the videos that David had been sent. He said you can't often tell just by looking anymore. Some AI videos can be extremely convincing for the general public, and even trained professionals can get it wrong on the first try. So he doesn't recommend relying on your eyes to spot the fake.
[Ricardo Amper]
The first thing is to recognize that there's a lot of shame involved when this happens, and it happens very frequently. We protect a bunch of companies, including dating sites and similar. I think the first thing I would tell your listeners is that you shouldn't be ashamed. This is happening all around, and technology has evolved far beyond what human intuition can detect. What I would tell those people is, make sure that when you're engaging in some of these websites, you're able to verify your identity first. It's not that you don't trust. As long as that's the modus operandi of the site, then there's a much lower likelihood of having these types of scams, because what they do is they take your picture, they make sure it matches the picture on your driver's license. If they're good sites, they've hired a vendor like us where we would in real time be detecting deepfakes. And so it becomes a much safer environment.
[Bob]
Can you, I realize I'm asking a question that's a years-long answer, but how does technology like yours work? How do you detect deepfakes? How does your computer tell the difference between what's real and what's a computer?
[Ricardo Amper]
We are at an age where it's called defensive AI. It's AI protecting against AI scams. What we do is we have a pretty extensive fraud lab where we have about 120 tools, some of which we develop, some of which we get from the dark web and even just the normal video generation sites that now exist. We create millions of deepfakes. And then what we do is we train AI models to detect things. They detect very little pixel-by-pixel tell signs that a human eye cannot see. And if you train them with enough data, and you have enough samples, including the ones we catch live with customers, eventually you create a really good model. So when someone is trying to verify their identity at these sites, once you take your selfie, those models are looking at those pixels in a very microscopic way. That's how it works.
[Bob]
So you make deepfakes that you then detect.
[Ricardo Amper]
Exactly right. We have to advance the art of how to create some of these, be very active in the dark web, understand what the newest tools are, and then with that capability, train models to protect against it.
[Bob]
Okay, what would you suggest for someone like David, who isn't on a dating site but gets approached through Facebook or Words with Friends? What verification procedure would you recommend?
[Ricardo Amper]
It is very tricky because when you have a more freeform way of meeting someone, the social media companies haven't necessarily adopted that yet. The first thing I would say is some of these social media websites already have what they call verified profiles. Make sure it's a verified profile. The second thing is, if they're from the area, encouraging a meeting is important, because the more you get involved and build connection without verifying, the more likely you are to get scammed. And out of all the fraud that we see, the one with the highest cost per person is unfortunately romance scams. But the industry is changing. Most video conferencing companies are adopting technologies like this. It wouldn't be rare, maybe in a year, that you'd have a check mark on your Zoom that says you're the real Bob Sullivan, I'm the real Ricardo Amper. You verify once, show your ID once, and then every subsequent conversation shows that you did. And it's a safer world.
[Bob]
Despite all these scary implications, Ricardo actually remains optimistic about the future of AI, but it's going to be a bumpy ride.
[Ricardo Amper]
I think gen AI is a wonderful revolution that is going to make all of us more productive. But in the short term, there's gonna be a lot of fraud. I think people don't understand the scale at which the combination of perfectly impersonating someone in video and audio, combined with AI agents doing the conversation, is going to become a big mess. There are three things people have to be aware of. Number one, make sure the website you're interacting with has implemented this technology. Number two, we are working with government at the local and state levels, partnering with DMVs, so that in the future you won't have to show your ID. It'll just be first name, last name, date of birth, and take a selfie. And number three, the more you encourage companies to use these technologies, the more of a trust graph it becomes, where all these companies are connected so that we can together fight fraud. It's gonna get very bad in the next two or three years, but I think in 10 years it's gonna be a much safer and very productive world.
Transcription courtesy of AARP.
%20(1).avif)
