Chinese app for movie fans raises the specter of "deepfake" technology
A new app allows people to put their faces into scenes from movies and TV shows by uploading a single photo. But there could be a potential cost to a user's privacy.
One user of the Chinese app Zao posted a video on Twitter in which he was able to "become" Leonardo DiCaprio. The app is currently only available in China, but it raises new concerns that fake videos are becoming too realistic, and that your face could be used for anything.
CBS News contributor and "Wired" editor-in-chief Nick Thompson visited "CBS This Morning" to discuss the privacy implications of what is known as "deepfake" technology.
He admitted that the Zao app is a lot of fun: "That's why it's become so popular so quickly. The concerns are that people can take your image and do whatever they want with it. In China, in particular, where an image is used for financial transactions, having somebody create realistic simulacrums of your face is scary."
Thompson said what makes this app different from similar programs is that "it's really good, and they came up with a very clever idea of creating a bunch of set movie scenes or TV scenes and allowing you to put your face in the scenes. It was a brilliant idea for an app."
In case you haven't heard, #ZAO is a Chinese app which completely blew up since Friday. Best application of 'Deepfake'-style AI facial replacement I've ever seen.
— Allan Xia (@AllanXia) September 1, 2019
Here's an example of me as DiCaprio (generated in under 8 secs from that one photo in the thumbnail) 🤯 pic.twitter.com/1RpnJJ3wgT
"My feeling about this is there are real privacy concerns," Thompson said. "And the terms of service initially when the app launched were quite bad. They said, 'We can do whatever we want with your face for as long as we want it. We can even give it to the government for whatever purpose they might have.' So, that's concerning."
Thompson said privacy concerns would be greater for those trying to keep themselves anonymous. "For a public figure or for someone who shares their face all over the place — as all four of us do — I don't see why it's particularly bad. If you're a dissident, if you're an undercover operative, if you're concerned about Chinese state surveillance, don't use this app!"
Some have already raised concerns about the impact technology like this could have in the upcoming presidential campaign. Co-host Tony Dokoupil said, "Adam Schiff, a Democratic congressman, said that there are nightmarish scenarios tied to this technology into 2020. We used to believe what you see with your eyes or hear with your ears. Now you can't anymore?"
"Absolutely," said Thompson. "In 2020, it is highly likely that someone will make some kind of a very realistic fake video of one of the candidates saying something that they didn't say or doing something they didn't do. That's concerning. But candidates also have cameras with them all the time. It's relatively easy to disapprove.
"What I worry about the most is bullying, and kids creating videos of classmates. There's a lot of bad stuff that you can do with this technology to really hurt people's feelings that may be worse than what's going to happen in politics."
Thompson says the company behind the Zao app has agreed to change their terms of service. "They've had a lot of problems because of the privacy backlash," he said.
But the threat of deepfake technology runs both ways, by allowing someone who is in a genuinely compromising video to be able to claim, "I don't think it's me. That's not me in blackface, or that's not me singing that particular song or making those particular comments."
"Absolutely," said Thompson. "We're going to need real verification technology, which fortunately some people are working on."