It’s been in the air for some time, but now it seems it’s finally here: the first major scam with the help of deepfake videos. This all concerns an unnamed multinational company, which has been defrauded of 200 million Hong Kong dollars.
According to CNN Fraudsters have made clever use of deepfakes. AI software digitally attaches one head to another, making it appear like someone is talking when that’s not actually the case. This software ensures that lip and facial expressions are copied.
About the Author
Laurens Verhagen is the science editor for de Volkskrant. He writes about technology, the internet, and artificial intelligence. Previously, he served as editor-in-chief of nu.nl.
Due to the availability of accessible software, anyone can now perform this trick. Politicians singing a song is not an innocent act, but there are also more dangerous applications such as the creation of disinformation. And recently Taylor Swift became the victim of a fake image showing herself in a sexual scene.
However, using deepfake videos in a live setting is much more difficult, but according to CNN this has now been done successfully. The finance employee was already suspicious when he received an email from his finance director asking him to transfer a large amount of money.
However, a subsequent video encounter convinced him. The employee put aside his initial doubts after this video call, because “the participants looked and sounded like coworkers he knew,” CNN quoted a police official as saying. The finance director is also among these colleagues. After the conversation, the employee still transferred the requested millions. However, everyone present turned out to be fake.
Criminals have been using AI to seduce people for some time, but so far there have only been successful attempts to imitate voices alone.
Bunq
The case in Hong Kong is reminiscent of a similar effort in the Netherlands last fall. Here, the fraudster pretends to be Bunq CEO Ali Niknam during a video conference. In this case the employee was not fooled because he smelled trouble.
Theo Gevers, a computer vision professor at the University of Amsterdam, did not dare say whether this was a true deepfake because police have not released the video.
On its own, deepfake live conversations can be done realistically, Gevers said, but only in certain settings. ‘I have a lot of questions. For example, is the voice also cloned? And it said there were some people there. Is it in the same room, or separately, through different screens?’ In short: ‘There are still many questions that cannot be answered definitively.’
For popular AI scientist and author, Gary Marcus, it is now clear: ‘Black Mirror has arrivedearlier than planned,” he wrote in response to the news, referring to the popular tech dystopian series on Netflix.
“Coffee trailblazer. Analyst. General music geek. Bacon maven. Devoted organizer. Incurable internet ninja. Entrepreneur.”
