400 Hidden Deepfake Apps Slipped Past App Stores
WIRED found 400-plus apps posing as pet editors and tutors that hid deepfake tools from Apple and Google reviewers. How the trick worked and what stores did.
A WIRED investigation published on November 2, 2026, found more than 400 connected apps and websites that looked like harmless everyday tools but could secretly load tools for making abusive deepfakes from photos of real people. The apps posed as pet photo editors, language tutors, workout planners and similar utilities, and they reached both Apple's App Store and Google Play. Researchers say the network used a server-side trick to pass app review, then switched on its hidden features for selected users.
How the hidden switch worked
The apps did not show their hidden features to everyone. After installation, each app contacted a remote server and sent details such as the user's language, referral source and a unique identifier. The server then decided what the user would see: the advertised feature, or a separate web page with deepfake generators. Someone opening the app normally might see nothing suspicious. One apparent trigger was arriving through an advertisement on Facebook or Instagram. Researchers traced thousands of ads promoting the abusive tools to apps listed in Apple's App Store and Google Play, and users who entered through those ads could buy credits through Apple Pay and Google Pay to generate imagery from uploaded photos. The shared code tied the apps together. Researchers found a JavaScript bundle that referenced hundreds of connected apps and domains, along with shared encryption keys and advertising trackers. That shared infrastructure points to a coordinated operation, not a set of unrelated developers.
The scale of the network
The nonprofit AI Forensics and cybersecurity firm Infoblox traced the cluster to Chinese-language codebases, Chinese hosting services and a Hong Kong-registered company called KeLing Technology Limited. Infoblox researcher Mael Le Touz mapped nearly 500 linked domains, while AI Forensics lead researcher Paul Bouchaud traced registrations back to late 2025 and found the network registering as many as 30 new domain and app combinations per week at its peak. The downloads were large. In September 2026, researchers found 27 related apps live on Google Play with more than 350,000 combined downloads, many rated suitable for all ages. One app, DiffAI, passed 100,000 installs on its own. One example named in the reporting is Viefur, a pet portrait app that appeared to turn animal photos into human-like characters.
What Apple, Google and Meta did
Apple removed 19 apps, including Viefur, after WIRED contacted the company and shared details of the network. Apple spokesperson Adam Dema said the apps complied with App Store rules during their initial reviews, and that developers later changed the websites supplying the apps' content. Google had removed most of the flagged titles before WIRED contacted it. Meta said it had banned KeLing Technology Limited in 2025 after a cease-and-desist letter, and that after receiving the researchers' dossier it acted against 11,600 ad accounts, roughly 4,700 business pages and blocked 209 apps from its advertising platform. Despite that enforcement, WIRED reported that similar promotional campaigns and mirrored apps kept appearing as its reporting ended.
Why app review missed it
The case exposes a basic limit of app review. Reviewers check the app a developer submits, but they cannot see features that a remote server switches on later. Developers can change content delivered from external servers without submitting a new app version for approval, so banned features can appear after an app has passed review. Apple has tightened its rules in response to this kind of abuse. On October 16, it updated its App Review Guidelines to explicitly cover AI-generated nonconsensual intimate imagery and to ban hidden, dormant or undocumented features, including features activated after approval. Apple also said it suspended tens of thousands of apps over the past year for similar bait-and-switch tactics. Rules alone cannot close the gap, because the abusive content in this case was served from remote servers, not baked into the approved app itself. Detecting those changes after approval remains the hard part for both Apple and Google.
What happens next
The removals do not end the story. Researchers and platforms will be watching whether the network reappears under new names, domains and app listings, as WIRED says it already has. The case is also likely to feed into wider debates about app store safety, advertising rules and how platforms police AI tools that can misuse real people's photos. For users, the lesson is simple: an app that looks harmless in a store listing can behave very differently once it phones home to a remote server. Checking the developer, the reviews and the permissions an app asks for is still worth doing, but this case shows even that may not be enough when the real behaviour is hidden behind a server switch.
Sources
- AppleInsider - https://appleinsider.com/articles/26/11/02/hundreds-of-ai-apps-were-part-of-a-network-hiding-sexually-abusive-deepfake-tools - Silicon Report - https://www.siliconreport.com/wired-finds-over-400-domains-and-apps-hid-explicit-ai-features