Glonce

WIRED finds 400+ AI apps hiding sexual deepfake generators

securityThe Glonce editorsWired🔥65Glonce rating65Impact62Novelty55Relevance80Surprise60Players40Credibility85
WIRED finds 400+ AI apps hiding sexual deepfake generators

WIRED reports that the pet portrait app Viefur looks like ordinary AI slop: for around $4 you buy 130 in-app coins and turn pet photos into "anthropomorphic portraits", such as a hamster in beach shorts. According to the investigation, that is cover. Viefur is part of a network of more than 400 connected domains and apps that often present themselves as everyday services, including language tuition and AI workout coaches. When accessed under certain conditions, the apps stealthily load a web interface for generating explicit AI sexual images and videos from photos that users upload. The scenes on offer include ones called "bedroom rape" and "threatening knife", where a man holds a knife to a woman's neck.

Multiple researchers found the network after following links from thousands of video ads with explicit deepfake content on Meta's Facebook and Instagram. WIRED has earlier reported that hundreds of those ads contained AI-generated child sexual abuse material, including images of real children and a member of a European royal family; the ads reached tens of thousands of people and prompted investigations from regulators around the world. The ads linked to apps in Apple's and Google's app stores, where people could generate explicit images and pay using Apple Pay and Google Pay. The network appears to have been publishing apps for around a year, at times dozens per week, even though Google, Apple and Meta say they enforce policies against harmful deepfake content.

How the trick works: Paul Bouchaud, a lead researcher at the nonprofit AI Forensics, reverse engineered the apps' and websites' code. Each time an app is downloaded, it calls a server with information such as where the user came from, the language and a unique identifier, and the server replies whether to show the explicit or the non-explicit version. There is no clear list of reasons a user gets the explicit version, but one trigger appears to be downloading the app after clicking an associated ad on Meta's platforms. The linked websites usually show a black screen, and the only visible change is that the URL gains the word "risk". They host the hidden layer: dozens of pornographic deepfake video and image templates, where users upload images of people to be inserted into sexual scenes, plus the option to buy "diamonds" to generate content and links to social accounts for contacting the developers.

The same websites expose the network. A JavaScript bundle holds a list of all domains and apps the developers created, including app package details, encryption keys and Facebook advertising trackers. Maël Le Touz, a staff threat researcher at security firm Infoblox, who identified almost 500 linked domains, said the domains run the same code and reference one another.

Scale and timing: Bouchaud's analysis of domain registrations found the apps began appearing toward the end of 2025, with a spike in spring 2026. At one point around 30 nudify app-websites were registered each week. In September, 27 were live on Google's Play Store, downloaded more than 350,000 times collectively, and many were rated suitable for all ages. One of the most popular, DiffAI, was created in December 2025 and has been downloaded more than 100,000 times from the Play Store. A user review there read: "This app contains templates for creating CSAM and other types of pornographic material. It performs no age verification." Since the start of the year, Bouchaud's research has found tens of thousands of explicit deepfake ads linked to the network on Meta's platforms, often with similar copy. Ads used apparently false names and pages with no followers. Danish journalists reported on portions of the network in September.

Who appears to be behind it: WIRED says, based on a review of web domains and how the services operate, that the apps and domains appear to be run by Chinese developers and a Hong Kong registered entity called KeLing Technology Limited. Le Touz points to exclusive Chinese hosting and code written, and commented, in Chinese; explicit ad videos carried watermarks from Chinese-language video editing software. The apps used generic privacy policies, free Gmail or Outlook addresses and obscure developer names. KeLing's name was listed as the contact on the first website registered in the network and in the brand name within the deepfake template; Bouchaud also found an admin login page mentioning "Keling", and KeLing-domain emails appeared on at least one app since removed from the stores. Toward the end of September, users of the explicit interface were notified of a new AI model called Veyra (tagline: "More real. More natural. More wild."), and the Veyra AI site listed a KeLing address as its contact. KeLing emails did not reply to WIRED, but after contact someone renamed some social channels, removed KeLing emails from the Veyra AI site and removed the KeLing address from the first registered site.

The platforms' responses: Apple removed 19 apps, including Viefur, after WIRED asked. Spokesperson Adam Dema said nudification apps and other pornographic content are strictly prohibited, that every app is reviewed before publication, and that these complied when reviewed; afterwards, he said, developers changed the website the app pulls content from and promoted it through pornographic ads on a third-party platform. Apple says it has suspended tens of thousands of apps for such bait-and-switch tactics over the last year and recently updated its rules to further restrict nudify tools. Google did not respond to a request for comment, though it had removed most of the apps before being contacted; DiffAI came down shortly before. Meta says it banned KeLing Technology Limited from advertising in 2025 after it failed to comply with a cease and desist letter (Meta declined to share the letter, citing legal reasons), and that KeLing keeps trying to advertise by other means. After WIRED shared the 400-plus domains, Meta said it had identified more than 3,000 associated ads, two-thirds already removed. It later said it had acted against 11,600 ad accounts, more than 4,700 business pages, 5,200 users and 15,000 pages linked to the domains, and blocked 209 apps from advertising on its platforms. As WIRED finished reporting, similar ads and apps were still being published.

Key facts

  • WIRED found a network of more than 400 connected domains and apps, posing as pet portrait, language and workout apps, that can load a hidden web interface for generating sexual deepfakes from uploaded photos.
  • The apps ask a server on each download whether to show the benign or the explicit version; one trigger appears to be installing after clicking a Meta ad. Bouchaud's research counts tens of thousands of explicit deepfake ads tied to the network this year.
  • In September, 27 network apps were live on Google Play with more than 350,000 downloads collectively; DiffAI alone passed 100,000 downloads.
  • Apple removed 19 apps after WIRED asked, Google had removed most before being contacted, and Meta says it acted against 11,600 ad accounts and blocked 209 apps from advertising.
  • The network appears to be run by Chinese developers and a Hong Kong entity, KeLing Technology Limited; similar ads and apps were still appearing when WIRED finished reporting.

Why it matters

The story shows a way around app store review: the apps pass inspection as harmless utilities, then switch to explicit deepfake generation by loading content from a website. Apple's own account fits this: the apps complied when reviewed, and developers changed the website content after approval. The scale is large. The network spans more than 400 domains and apps (Infoblox counted almost 500 domains), and at its peak around 30 nudify app-websites were registered each week. The content includes tools for sexual deepfakes of real people, and the ads previously reported by WIRED included AI-generated child sexual abuse material and have prompted investigations from regulators around the world.

Who it affects

Most directly, the people whose photos users can upload into sexual scenes, a harm that falls on people who never used the apps. The ads reached tens of thousands of people, and the earlier WIRED reporting involved images of real children and a member of a European royal family. Users who download what looks like a pet or language app may be shown something else. Apple, Google and Meta are affected as platform operators: they host the apps and ads, and take payments via Apple and Google Pay. Katie Paul of the Tech Transparency Project says these companies need to answer where the money goes and how the operators keep bypassing moderation.

How to use it

This is an investigation, not a product, so there is nothing to adopt. For researchers, moderators and app reviewers, the article lists concrete traits of the network: a startup call to a server that decides which version to show, websites that load as a black screen with "risk" added to the URL, a JavaScript bundle listing sibling domains, encryption keys and Facebook trackers, generic privacy policies, free Gmail or Outlook contact addresses, and ads under apparently false names from pages with no followers. Le Touz's work shows the domains reference one another and run the same code, which is how almost 500 were mapped.

How solid is it

The reporting is WIRED's own investigation, built on technical work by named researchers: Bouchaud (AI Forensics) reverse engineered the code, and Le Touz (Infoblox) mapped the domains. Download counts and registration timing come from Bouchaud's research. Apple and Meta responded on the record with specific figures; Google did not respond to WIRED's request for comment, and the claim that it removed most apps is WIRED's. Attribution to KeLing Technology Limited and Chinese developers is hedged: the apps and domains "appear to be run" by them, based on a review of web domains and how the services operate. Neither KeLing nor the other linked email addresses responded.

Risks and caveats

The operators are not conclusively identified, and the KeLing link is stated as appearing, not proven. KeLing did not respond, and Meta says it banned the company from advertising in 2025 but that it keeps trying other routes. Removals are partial snapshots: Apple pulled 19 apps, and Google removed most, but WIRED says similar ads and apps were still being published when it finished reporting. Bouchaud's view is that the operators do not expect enforcement to catch up and that so far it has not. The article names no regulator and reports no outcome of regulators' investigations; it mentions no arrests, legal action or fines. Apple's statement that the apps were compliant at review is the company's account.

“This is an organized, commercially run operation: 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”

— Paul Bouchaud, lead researcher at the nonprofit AI Forensics