Chinese Viral App ZAO Forced To Backtrack After WeChat Ban And Privacy Backlash

iOS App Store / ZAO

It was inevitable—and now it has happened. China’s viral sensation ZAO has bowed to intense media pressure and changed its privacy approach to the use and ownership of the millions of user images uploaded to the app since its launch last week.

I reported on the rise and rise of ZAO on Monday [September 2], describing it as China’s own “FaceApp moment.” ZAO’s users can upload images and then watch as its AI engine swaps their faces with those of celebrities in a selection of video clips. The privacy backlash was instant—and now the backtrack has been almost as fast.

The app soared to the top of China’s iOS store download chart over the weekend, and so this backtrack will affect the millions of users who have spent the last 72 hours in its deepfake playground.

Just as with FaceApp, it was ZAO’s privacy policy as much as its technology that raised concerns. Again, over the rights granted by users to the developers over the uploaded images. And it’s not helped by the security risks inherent in identity manipulation.

Such was the strength of those concerns, that China’s leading messaging platform WeChat restricted access to the app. Invites and links were prohibited, with the TenCent-owned platform citing “security risks” in its decision.

Now CNN has reported that ZAO (owned by NASDAQ-listed Momo Inc) has responded to this backlash and changed its end-user agreement to reflect all the “concerns about privacy and safety issues.” As a publicly listed company, this was likely an easy decision to make given the strength of feeling around the world.

ZAO’s user agreement was criticized for asking users to cede too many rights over their images and data than were necessary to run the service. Exactly the same issue that caused such a backlash against FaceApp. Now ZAO “will try its best” to only use content “within a reasonable, necessary and expressly stated extent.”

“This is a new product,” the developers said on Tuesday [September 3] on the Chinese site Weibo. “We were indeed inconsiderate about people’s core concerns.” The developers have also made clear that ownership of the image and data rights is not changed by the agreement—it stays with the user.

There is always going to be enhanced sensitivity when a viral application harvests facial images and then deploys AI to manipulate those images—again, just as with FaceApp. The security risks associated with the interchangeable identities of deepfakes has already raised material concerns, making such manipulation more accessible just makes that worse.

And, on this subject, ZAO has assured that no “facial biometric data” will be stored and all other information will be deleted in line with data regulations. “We’ve also adopted several safety measures,” the developers added, “including storage encryption.”

“One can expect the privacy policies surrounding the app to adapt and change to reflect user concerns,” I said on Monday—which is exactly what has now happened. “And then,” I added, “we can focus on the ease of mass deployment of a powerful deepfake application, which in itself raises all kinds of concerns.”

And that’s now where we find ourselves—and this debate will be much harder to resolve, with significantly more serious implications.

Original Source