How deepfakes are becoming a serious threat to companies

HostI was looking at a news story the other day that kind of blew my mind. We used to think of fake videos as just things people made to mess with famous actors or politicians, but it seems like the target has shifted. Now it's the person sitting in the cubicle next to you or the person on your morning video call. Why should we be worried about this in our normal jobs?

GuestIt's a big shift because for a long time we thought we could trust what we saw with our own eyes on a screen. But there was this huge case recently in Hong Kong that really shows how things have changed. A worker at a big firm got an email that looked like it came from the head of the company. It told him they needed to move a lot of money for a secret deal. At first, he was a bit unsure. But then he got invited to a video call with the boss and other people from the team he knew. He saw them, he heard their voices, and they all told him the same thing. So he sent the money. It turned out he was the only real person on that call. The rest were all fakes made by AI. He ended up sending over twenty-five million dollars to people who were basically ghosts.

HostWait, a whole video call? I can see how one person might be faked, but an entire room of people you know? That feels like something out of a movie. How do you even pull that off without the video looking like a bad video game?

GuestWell, the people doing this are getting very smart about it. They don't have to build these people from scratch. They go out and find old videos of company meetings or talks the bosses gave online. They take those real clips and use AI to make the faces move and say whatever they want in real time. And on a video call, if the connection is a little fuzzy or the light is kind of dim, it's really hard to tell that the person's mouth isn't quite right. They kept the call short, they gave quick orders, and they relied on the fact that when your boss tells you to do something and you can see them right there on the screen, you usually don't stop to ask if they're actually a bunch of code.

HostBut what about the voices? Usually, when I hear those AI voices online, they sound like robots. They have that weird, flat way of talking. Surely you would notice if your boss suddenly sounded like a toaster?

GuestThat's the thing, the tech for sound has moved even faster than the video. Nowadays, you only need a few seconds of someone talking to make a clone of their voice that sounds almost perfect. If you have ever posted a video of yourself on a site like LinkedIn or given a talk that ended up on the web, your voice is out there. These tools can pick up on the way you breathe, the way you pause, and the little accent you might have. It's not just about the words. It's about the feeling. They can make the voice sound stressed or like it's in a hurry, which makes the person on the other end feel like they have to act fast without thinking.

HostSo it's less about making a perfect movie and more about just being good enough to trick someone who's in a rush. But I still feel like I would know if it was my friend or my boss. There are those little things, like a joke they always make or a way they sit.

GuestYou would think so, but when you're at work, you're often tired or just trying to get through your list. We have spent years being told to watch out for bad spelling in emails, but we haven't been trained to doubt a face. And it's not just about stealing money directly. Think about what happens if a video comes out of a boss saying something awful about their own company, or maybe saying their latest product is a lie. If that video goes around on social media, the company's stock price could drop in minutes. By the time the real boss comes out to say it was a fake, the damage is already done. People have lost money, and the brand is hurt.

HostThat sounds like a nightmare to try and stop. Is there not some kind of tool or software that can just scan the video and tell us if it's real? Like a filter that catches the fakes?

GuestPeople are trying to build that, but it's a bit of a cat and mouse game. Every time a new tool comes out to catch a fake, the people making the fakes find a way to get around it. It's like a race that never ends. Some companies are actually going back to basics. They're telling their staff that if they get a weird request for money or data, even on video, they need to use a safe word. It sounds old fashioned, like a secret club, but having a word or a phrase that only the real team knows is one of the only ways to be sure. Others are saying you should ask the person on the screen to do something weird, like turn their head all the way to the side or wave their hand in front of their face. Sometimes the AI has a hard time keeping the fake face glued on when there's a lot of movement.

HostIt's wild that we have all this high tech stuff and the best fix is asking your boss to wave their hands around like a fan.

GuestThe biggest risk isn't the tech itself, but how much we want to believe what we see.

HostThe office meeting might never feel the same now that we have to wonder if the person on the screen is actually there.