Wired found that more than 400 connected apps, available on the App Store and Google Play, hid a sexual and violent deepfake generator behind completely harmless facades: language tutor, fitness coach, pet photo filters. The apps would be the work of a single Chinese organization, attributable to a company based in Hong Kong, the KeLing Technology Limitedand would have passed unscathed through the controls of Apple, Google and Meta despite the policies of each platform explicitly prohibiting content of this type.
The most cited case concerns Viefurpresented as a simple app to “dress up” pets in photos. Behind that facade, under certain conditions, the app loaded a separate web interface capable of generating explicit sexual images and videos from user-uploaded photos. The available templates also included violent scenarios labeled “bedroom rape” and “threatening knife”, the latter depicting a man holding a knife to a woman’s throat.
The mechanism used to overcome Apple and Google’s review is a classic bait-and-switch: At the time of publication the apps only show the declared functions. A server-side check then activates the hidden layer: one of the triggers identified is the installation after clicking on a deepfake-themed ad that appeared on Facebook or Instagram. Once unlocked, the feature allows you to download dozens of pornographic templates and purchase credits to generate content from user-uploaded faces.
Thousands of adverts, checks passed for a year
The network, according to what was reported in the original investigation, is supported by thousands of advertising videos spread on Facebook and Instagram, where the real nature of the apps remains hidden behind reassuring claims. Some applications have been active for about a year and have passed the review checks of Apple, Google and Meta several times, even though each company has long prohibited the generation of non-consensual deepfakes or sexual content.
After the details have been shared with the companies involved, Apple removed 19 apps from your store. Cupertino’s official response admits the problem, but places much of the blame on the developers:
“We review every app before publishing, and these apps were compliant with our rules when we reviewed them. After approval, the developers changed the site from which the app retrieved content, replacing what was approved with pornographic material, promoted via ads hosted on a third-party platform.”
Meta claims to have already strengthened the filters against this type of circumvention and to have taken action against it over 11,600 ad accounts and 15,000 pages connected to the domains associated with the offending apps. Google, for its part, has made it known that it does not allow apps with sexual content on the Play Store, that it has removed those identified and suspended hundreds of other apps for similar violationsas well as narrowing the search terms that led to discovering them.
The underlying problem is not completely resolved: controls designed to block prohibited content at the time of submission do not intercept an app that changes behavior after distribution has already taken place. A structural limit that removal alone, however timely, is not enough to truly resolve.
