It looks like the future: the ability to seamlessly transplant yourself into the title roles of some of the biggest films of the last few decades, all from your smartphone. Little wonder that Zao, a Chinese faceswapping app, has become a runaway success.
“It’s a fairly simple, intuitive interface,” explains Matthew Brennan, a Chinese tech expert who has tried out the app. “You just pick from a selection of clips, scroll down the page, it’s mostly Chinese movies but some western ones and things such as Game of Thrones.”
Users pick a scene or meme they’d like to transplant themselves into, record footage of their face including movements of their head, and upload it to the app. “Once you upload it, it’s super fast and generates the clip for you,” Brennan says. “The thing that was most impressive was the speed with which it was able to turn one picture into a video.” The app works by training AI on pre-existing clips, and then mapping users’ faces onto that footage.
It’s the most downloaded free app on China’s iOS App Store – despite only being released on Friday – and its developers, a company called Momo, has said its servers were stretched to breaking point because of its viral popularity over the weekend. Rumours abound that mentions of the app are banned on popular chat app WeChat, which Brennan says often indicates the competitor views it as a threat, or users are posting so much of the same content it is rendering people’s feeds boring.
But almost as quickly as word of the app spread across Chinese social media, concerns followed about the app’s terms and conditions, and its eerily accurate ability to map users’ faces onto pre-existing video clips.
The controversy is reminiscent of concerns about FaceApp, a Russian-developed app that aged photographs of users but was alleged to have been a front for a mass Russian hoovering up of all our faces. (Spoiler alert: it was not, but its terms and conditions were troubling for different reasons.)
So what are people’s worries? For one thing, there’s a worry that technology is reaching the uncanny valley, with the alarm that people felt around the advent of deepfakes becoming more pronounced now the technology is in consumer’s hands (and can be used so quickly and easily). It’s a small step, some reckon, from pre-packaged video clips into more freeform deepfakery. But there’s no evidence that the impressive results of Zao’s AI could be replicated in footage outside its core library of clips.
“The privacy implications for this app in particular are less significant than they first appear, since there is a limited library of originals that the user-provided content can be inserted into,” says Rowenna Fielding, a privacy and data protection expert at Protecture.
“However, the mainstreaming of deepfake technology into the consumer space is significant, both for personal and political life,” she adds. “It seems likely that investment into forensic examination of digital images will be needed, to detect and prevent this technology from being used in harmful ways.”
Just as people were concerned FaceApp was uploading images of their face to Russian servers, so people are worried Zao is doing the same. It’s not, according to French security researcher Baptiste Robert.
Another concern that Zao users share with those of FaceApp back in July are the broad terms of service to use the app, which grant the developers global rights to use any images created on the app – for free and in perpetuity. But as with FaceApp, such legalese is par for the course for using apps nowadays.
“Many apps and websites already have similar terms, where user-generated content becomes the intellectual property of the company providing the service,” says Fielding. “The potential for harm to individuals arising from uncontrolled commercial re-use of their likeness varies enormously depending on the context, but clearly shows the need for terms and conditions to be made accessible and understandable to the average person, in order for them to be fair.”
The similarities to FaceApp are something Robert is concerned about. “The concerns are exactly the same,” he says. “We have an app which produces something very cool, but if the user uploads his face to the app he directly loses rights on his photos. Zao can do whatever they want with the user’s face.”
The biggest brickbat thrown against Zao is that it’s a potential tool of the Chinese government, allowing it to hoodwink users into giving over crystal-clear photographs of their face. It’s a worry in light of recent news that a massive hack of iPhones was likely to be instigated by the Chinese government to target Uighur Muslims living in the country.
But that overlooks three key elements in the Chinese technology ecosystem: one, citizens already willingly share their face with the government to pay for things like train tickets and to board flights seamlessly. And two, Momo is an established company. “It started around the same time as WeChat, at the birth of the mobile revolution in China,” says Brennan. This isn’t a fly-by-night app developer that has come from nowhere to produce a stunning service.
Which leads on to the final reason why Zao isn’t likely to be a Chinese conspiracy to gather the facial features of its citizens. “There are much easier ways to get that data,” says Brennan. “You could scrap that from Weibo. Besides, like most apps, while you put all that effort into building it, there’s no guarantee it will go viral. That doesn’t really make much sense to me at all that gathering data would be the purpose.”
More great stories from WIRED
🍔 World-class chef rates the best vegan burgers in the UK
😡 TikTok is fuelling India’s deadly hate speech epidemic
🍫 The foods you’ll really need to stockpile for no-deal Brexit
♻️ The truth behind the UK’s biggest recycling myths
🤷🏼 How is the internet still obsessed with Myers-Briggs?
📧 Get the best tech deals and gadget news in your inbox