Viefur presents itself as an AI pet-portrait app, but research by WIRED indicates it is connected to a much larger network of apps and domains that can generate explicit deepfake images and videos, including material depicting sexual abuse. The network reportedly uses seemingly harmless apps, targeted advertising, and conditional behavior to evade detection by app stores and platforms.
At first glance, Viefur appears to be another AI image generator. Users can purchase digital coins—130 coins cost about $4—and spend them turning pet photos into “anthropomorphic portraits” that place animals in “human clothes” and unusual settings. A hamster might appear beside a pool in beach shorts and sunglasses, while a cat might be shown drinking from a coconut in a Hawaiian shirt.
According to WIRED’s research, however, these features appear to conceal the app’s apparent primary purpose: generating deepfake images involving sexual abuse.
The app is reportedly part of a network comprising more than 400 connected domains and apps. Like Viefur, many of the services present themselves as ordinary products, including language-tutoring tools and AI workout coaches. Under certain conditions, researchers found, they can secretly load a web interface that uses uploaded photos to create explicit AI-generated sexual images and videos.
That interface includes scenarios labeled “Bedroom Rape” and “Blackmail Knife,” the latter depicting a man holding a knife to a woman’s neck.
Network discovered through explicit deepfake ads
Researchers uncovered the network after tracing links from thousands of video advertisements containing explicit deepfake material on Meta’s Facebook and Instagram platforms. WIRED has reported in recent months that hundreds of those advertisements included AI-generated child sexual abuse material, including images of real children and members of European royal families.
The advertisements reached tens of thousands of people and prompted investigations by regulatory authorities around the world.
In each case examined, the ads directed users to fraudulent apps available through Apple’s and Google’s app stores. The apps enabled users to generate explicit images and make payments through Apple Pay or Google Pay.
The network behind the nudify apps appears to have been publishing software for roughly a year, sometimes releasing dozens of apps in a single week. This activity continued despite Google, Apple, and Meta stating that their policies prohibit harmful deepfake content on their platforms.
Companies and identities linked to the operation
WIRED’s investigation into the operation of the relevant domains and services indicates that the apps and websites may be run by a Chinese developer and a Hong Kong-registered company called KeLing Technology Limited.
Email addresses associated with the entities, apps, and related services did not respond to multiple requests for comment from WIRED. After WIRED made contact, the apparent public connection between KeLing Technology Limited and the deepfake tool was removed.
“This was a coordinated, commercially executed operation, with a single operator behind hundreds of single-use apps and domains, hiding their true identity in order to pass App Store review, and pushing through tens of thousands of meta ads in fake payer names,” said Paul Bouchot, principal investigator at the nonprofit AI Forensics, which investigated the network. “They don’t expect enforcement to catch up, but so far they haven’t.”
How the apps hide their behavior
When users download one of the apps in the network, it may initially behave exactly as advertised—for example, by generating AI images of pets. Researchers found that the apps can also make a hidden decision about whether to show harmless content or load a web interface offering explicit deepfakes.
“Each time an application is downloaded, information such as the user’s hometown, language, and unique identifiers are called to the server, and the server responds by displaying an explicit or non-explicit version of the application,” said Bushau, who reverse engineered the app’s code and related websites.
Bushau said there is no clear list of factors determining which version a user receives. However, downloading an app after clicking on a related advertisement on Meta’s platforms appears to be one possible trigger.
Source: www.wired.com