Apple has removed 19 iPhone apps from its App Store after investigators uncovered a coordinated setup of over 400 apps and domains hiding deepfake tools designed to create sexually explicit imagery using photos of real people. These apps claimed to offer benign services—such as pet portraits, language lessons, and fitness coaching—but some were secretly loading abusive AI deepfake features. The network successfully evaded review processes by concealing these harmful tools behind external servers.
How the concealment worked
Researchers found that after installation, these apps reached out to remote servers sending data like the user’s language, where they came from, and a unique identifier. Based on this info, the server decided whether to show the same innocent features advertised or instead trigger access to a hiddenサイト running explicit deepfake generators. Some users got the advertised harmless services, while others—particularly those who clicked ads on networks like Facebook or Instagram—were exposed to the illicit content.
One striking case involved an app called Viefur: publicly, it offered AI-generated pet portraits. Behind the scenes, though, the same app could place real people into sexual or violent scenes through hidden deepfake technology. The investigation also exposed shared infrastructure—hundreds of connected apps and domains, a common JavaScript bundle, encryption keys, and ad trackers—suggesting the effort was centrally managed rather than being a collection of isolated bad actors.
Scale, affiliations, and reach beyond Apple
In September 2026, 27 of the connected apps were found on Google Play, with over 350,000 downloads combined. Many of them were labeled suitable for all age groups, indicating this misuse extended well beyond Apple’s platform.
Investigators also traced infrastructure connections to Chinese-language developers and a company registered in Hong Kong called KeLing Technology Limited. Shared codebases, hosting services, and contact info bridged apps across platforms. While this points to coordination, evidence did not determine who held full control over the entire network.
Apple’s response & policy changes
After being alerted by investigators, Apple removed 19 apps—including Viefur—that were part of the deceptive network. Apple claims these apps had passed review under their declared functionality. Only later did the content supplied via external websites change to include pornographic or nonconsensual deepfake materials.
Apple reports taking action against thousands of similar apps over the past year for ‘bait-and-switch’ behavior—in which an app presents one function at approval but delivers something else after launch. To counter this, on October 16, 2026, Apple updated its app guidelines: adding explicit language banning AI-generated nonconsensual intimate imagery (Guideline 1.2), and clarifying that hidden or undocumented features (Guideline 2.5.19) aren’t allowed. Yet, the core issue persists—apps can still supply disallowed content from external servers without needing to re-submit through Apple’s review process.
These findings reinforce concerns that app stores struggle to keep pace with abuse tactics—especially when external content is dynamically loaded. Users may encounter unlawful deepfake content long after an app has been approved, with minimal visibility into what’s happening behind the scenes.
What this episode proves: policy updates alone can’t stop every abuse. Apple’s latest guideline changes help, but detecting hidden features served from outside the app remains a major enforcement challenge. For the industry, it’s essential to improve runtime monitoring and enable users, researchers, and platforms to trace what content is actually being served—not just what’s claimed by sellers. Vigilance must shift from pre-approval to continuous oversight if we want to protect against nonconsensual AI misuse in the wild.”
