400+ AI Apps Conceal Sexually Abusive Deepfake Features Behind Pet Photos and Fitness Tools

Reviewed byNidhi Govil

5 Sources

Share

A WIRED investigation uncovered over 400 AI apps masquerading as pet portrait generators and fitness coaches while secretly offering sexually abusive deepfake features. Apple removed 19 apps after developers used bait-and-switch tactics to bypass app store reviews, swapping approved content for explicit tools that generated nonconsensual intimate imagery.

News article

Network of AI Apps Hides Sexually Abusive Deepfake Features

A

1

exposed a network of over 400 AI apps and connected domains concealing sexually abusive deepfake features behind seemingly harmless services. Apps like Viefur, marketed as a pet portrait generator charging around $4 for 130 digital coins, secretly loaded web interfaces enabling users to create explicit AI-generated images and videos from uploaded photos

1

. The hidden deepfake features included AI-generated scenes labeled "bedroom rape" and "threatening knife," where violent and sexual scenarios could be imposed on victims' photographs

2

.

Researchers from AI Forensics and Infoblox discovered the network after tracing thousands of video ads containing explicit deepfake content published on Meta's Facebook and Instagram platforms. These ads reached tens of thousands of people and linked directly to AI-powered apps distributed through the Apple App Store and Google Play Store

1

. The apps appeared to be operated by Chinese developers and Hong Kong-registered KeLing Technology Limited, according to analysis of web domains and operational infrastructure

4

.

Bait-and-Switch Tactics Bypass App Store Reviews

The network employed sophisticated bait-and-switch tactics to evade detection during app store reviews. Apps initially displayed benign content matching their advertised purpose—language tuition, AI workout coaching, or pet photo editing. After installation, the apps contacted remote servers transmitting information including the user's location, language, and unique identifier

3

. The server then determined whether to show the explicit or non-explicit version, with one apparent trigger being installation after clicking an associated Meta advertisement

2

.

Paul Bouchaud, lead researcher at AI Forensics who reverse-engineered the apps' code, characterized the operation as "organized, commercially run: One operator behind hundreds of disposable apps and domains, cloaked to pass App Store review and pushed through tens of thousands of Meta ads under fake payer names"

1

. The hidden web interfaces contained dozens of pornography deepfake video and image templates, with users able to purchase "diamonds" to generate nonconsensual intimate imagery and make payments through Apple Pay and Google Pay

2

.

Platform Responses and Enforcement Challenges

Apple removed 19 apps, including Viefur, after WIRED shared details of the investigation. Apple spokesperson Adam Dema stated the apps complied with platform policies during initial reviews, but developers subsequently changed the websites supplying content, replacing approved material with pornographic content and promoting it through third-party advertisements

3

. Apple reported suspending tens of thousands of apps over the previous year for similar bait-and-switch practices and updated its guidelines on October 16 to strengthen restrictions against nonconsensual intimate imagery and apps concealing functionality during review

3

.

Meta took action against more than 11,600 advertising accounts, 4,700 business pages, 5,200 users, and 15,000 pages linked to the domains associated with apps with hidden sexual deepfake features

5

. The company blocked 209 apps from advertising on its platforms and had previously banned KeLing Technology Limited from advertising in 2025 after the entity failed to comply with a cease-and-desist letter

4

. Meta advertisements used false names including "Draftmondcreldbro Duhustepre" and "Swaldmungsha Glasirdcroldshold," as well as pages with no followers

4

.

Scale and Technical Infrastructure of the Nudify Network

Maël Le Touz, staff threat researcher at Infoblox, identified almost 500 linked domains sharing code and referencing one another through encryption keys. A JavaScript bundle easily unpacked by researchers contained a configuration list of all domains and apps created by the developers, including app package details, encryption keys, and Facebook advertising trackers

1

. The operation used Chinese-only hosting, Chinese code comments, and watermarks from Chinese-language video-editing software in explicit advertising videos

4

.

By September 2026, 27 related apps accumulated more than 350,000 downloads collectively on Google Play, with many rated suitable for all age groups

3

. DiffAI, created in December 2025, surpassed 100,000 downloads before removal. One user review reported the app contained templates for creating child sexual abuse material and pornography without age verification

4

. Google removed DiffAI shortly before being contacted about the app.

Bouchaud's analysis found the abusive AI deepfake tools began appearing toward the end of 2025 and increased sharply in spring 2026, when approximately 30 nudify app websites were registered each week

4

. The network operated for around a year, at times publishing dozens of apps per week, despite Google, Apple, and Meta enforcing platform policies prohibiting AI-generated harmful content

1

. After being contacted for comment, entities linked to KeLing Technology Limited appeared to erase public connections between the company and the deepfake tools, with the Veyra AI website later removing a KeLing Technology email address from its contact details

4

.

[3]

AppleInsider

|

AppleInsider.com

© 2026 TheOutpost.AI All rights reserved