More than 400 apps and linked websites hid tools for generating sexually abusive deepfake imagery behind ordinary store listings, and the cleanup is still catching up with the network. Apple removed 19 apps after being shown the findings, Google had already pulled most of the flagged titles, and Meta said it acted against thousands of advertising accounts tied to the operation.
How the switch worked
The apps presented themselves as pet photo editors, language tutors, workout planners and other everyday utilities. When reviewers or ordinary visitors opened them, they showed a harmless tool or a blank screen. Users who arrived through targeted video advertising on Meta platforms, however, received server instructions that loaded a hidden interface for creating explicit imagery from uploaded photographs.
Researchers at the nonprofit AI Forensics and the cybersecurity firm Infoblox traced the cluster to Chinese-language codebases and Chinese hosting services, and to a Hong Kong-registered entity named KeLing Technology Limited. Infoblox researcher Maël Le Touz mapped almost 500 linked domains, finding shared JavaScript configurations and cross-referenced encryption keys.
Registrations began in late 2025 and peaked at about 30 new app-and-website pairs per week. The scale on Google Play was large: 27 apps found live in September had more than 350,000 combined downloads, including one app, DiffAI, with more than 100,000 installs.
What the platforms said
Apple spokesperson Adam Dema said the apps complied with its rules when reviewed, and developers later changed the remote websites feeding content into them. Apple said it had suspended tens of thousands of apps for similar bait-and-switch tactics and tightened its rules in October. Google did not respond, but had removed most titles before being contacted.
Meta said it had banned KeLing from advertising in 2025 after a cease-and-desist letter. After receiving the researchers’ dossier, it said it acted against 11,600 advertising accounts and roughly 4,700 business pages, and blocked 209 apps from its advertising platform.
Our read: review checks the package, not the server
Store review inspects the submitted app, not the server that can change what it shows later. That gap made this network possible: a clean listing, a quiet switch, and payment through Apple Pay and Google Pay. Closing it requires checking what an app loads after approval, not only what it shows on day one.
That is the same scale problem other AI warnings are now describing. Britain’s cyber agency has warned that AI can turn long-tolerated weaknesses into incidents faster, and this network shows the commercial version: ordinary distribution, mainstream billing, and abuse visible only after downloads pile up.
Why It Matters
For parents and users, the danger is that usual safety signals failed. Many apps carried all-ages ratings on Google Play, and their store pages gave no hint of the hidden interface. A listing and an age rating are not evidence about what an app does once it contacts its own server, especially when it arrives through a social media advertisement.
For regulators, the timing is sharp. AI Forensics notes that the European Union’s ban on nudifier apps takes effect in about a month, and Europe has already moved to test deepfake safeguards before AI systems ship. App stores may now be forced to prove, continuously, that approved apps stay approved.
Background
Nonconsensual deepfake tools have moved from fringe websites into mainstream app stores by borrowing the trust those stores carry. The network described here did not need to break into phones or bypass security controls; it used advertising, store listings and payment rails that millions of people already trust every day.
Source: WIRED investigation report and second full report
0 Comments