Hundreds of ordinary-looking apps on Google Play and the Apple App Store are hiding something ugly behind friendly icons. A WIRED investigation has pulled back the curtain on a network of nudify apps: disposable tools disguised as pet portrait makers, language tutors and workout coaches that quietly generate sexually abusive deepfake images of real people.
This is not one rogue developer or a handful of shady listings. It is a commercially run network of more than 400 connected apps and domains, pushed through thousands of ads, paid for with mainstream payment systems, and repeatedly cleared by platform review. Here is how the disguise works, who appears to be running it, and what has to change.
- Inside the Nudify Apps Network: Hiding in Plain Sight
- How Nudify Apps Disguise Themselves
- Scale and Speed: Hundreds of Domains, Dozens of Nudify Apps a Week
- The Advertising Engine Behind the Nudify Apps Network
- How the Nudify Apps Slipped Past App Store Review
- Following the Trail: KeLing Technology and the Operators
- Stopping Nudify Apps: What Has to Change
- References
Inside the Nudify Apps Network: Hiding in Plain Sight
Reporters from WIRED, working with researchers at AI Forensics and Infoblox, traced a network of nudify apps spanning more than 400 connected apps and domains. Many pose as everyday utilities. One self-proclaimed pet portrait app turns your cat into a person in a Hawaiian shirt for a few dollars in in-app coins. Under the right conditions, that same app quietly loads a hidden interface for generating explicit deepfake images and videos from photos users upload.
Ads, not app stores, started the trail
The nudify apps surfaced through links, not downloads. Researchers documented thousands of video ads on Meta’s Facebook and Instagram containing explicit deepfake content that led to the nudify apps. Earlier batches of those ads, as WIRED reported, contained AI-generated child sexual abuse imagery, including images of real children and a member of a European royal family. Regulators around the world opened investigations after the ads reached tens of thousands of people.
A commercial operation, not a hobby
AI Forensics lead researcher Paul Bouchaud describes an organized, commercially run operation: one operator behind hundreds of disposable nudify apps and domains, cloaked to pass app store review and pushed through tens of thousands of Meta ads under fake payer names. Their calculation is blunt. They do not expect enforcement to catch up, he says, and so far it has not. That single sentence doubles as an indictment of every gatekeeper between a developer account and a billion devices.
Why the pet-portrait disguise works so well
Cute-animal apps are the perfect cover because they attract exactly the scrutiny that abusive tools fear. They get family-friendly ratings, cheerful screenshots and gift-purchase coins that look harmless on a receipt. A reviewer sampling the app sees a hamster in beach shorts. Nothing in that experience hints that the binary on the same device can fetch a different product on demand. The disguise is not a side effect of the design; it is the design.
A familiar economy aimed at new targets
The monetization model is lifted straight from mainstream mobile games. Around four dollars buys roughly 130 coins, and every generation spends them, so a curious buyer barely notices the cost. The same pack mechanics that make puzzle games profitable make abusive generation frictionless: no review, no conversation, just a checkout. When payments feel routine, harm scales at the speed of impulse, and the same pipeline that turned celebrity AI deepfakes into a national conversation now reaches anyone with a photo.
How researchers count an invisible network
Measuring the nudify apps is itself a research problem. AI Forensics and Infoblox worked from different ends, one tracing ad infrastructure and reverse-engineering app code, the other counting linked domains through registration and hosting signals. Their numbers differ slightly, from four hundred to nearly five hundred domains, which is exactly what you would expect from a moving target. The overlap between two independent methods is the strongest evidence available that the network is real, coordinated, and commercially maintained.
How Nudify Apps Disguise Themselves
The disguise is the product. Download one of the nudify apps and you will often see exactly what was advertised: pet pictures, flashcards, fitness tips. That version exists to survive review and keep ratings clean. The real product appears only when the network decides the moment is right, and the switch is flipped from the server, not the app.
The server decides which version you see
On first launch, each install of the nudify apps phones home with details such as where the user came from, the device language and a unique identifier. The server replies with a decision: show the benign interface, or load the explicit web layer. Researchers found no clear public list of triggers, but arriving after clicking an associated ad on Meta’s platforms appears to be one. Reviewers get the postcard version; paying users get the other.
A hidden web layer doing the real work
Open one of the network’s websites and you often get an entirely black screen; the only visible change is the word “risk” appearing in the URL. That layer hosts dozens of explicit deepfake templates, sells “diamonds” to generate images, and even lists social accounts where users can send questions or complaints about the service. Buried deeper in the same code, a JavaScript configuration bundle leaks the whole playbook: the full list of domains and apps, package names, encryption keys and advertising trackers.
What the config bundle reveals
Maël Le Touz of security firm Infoblox, who identified almost 500 linked domains, says the bundle contained a giant configuration list of other domains running the same code, all referencing one another. It is the fingerprint of an assembly line: the same product reskinned dozens of times a week, with keys and trackers pasted into each copy. The configuration file behind the nudify apps is effectively a public directory of the operation, and it was never even obfuscated well.
Laid bare, the trick that lets nudify apps survive review is simple to describe and hard to police:
| Layer | What users see | What runs underneath |
|---|---|---|
| App store listing | Pet portraits, language tips, workout plans | A shell built to pass review |
| First launch | Exactly the advertised feature set | A call home with referrer, language and device data |
| Server check | Nothing visible to the user | Server chooses the benign UI or the explicit layer |
| Payment | Coins or diamonds for “images” | Revenue for abusive generation |
Scale and Speed: Hundreds of Domains, Dozens of Nudify Apps a Week
Domain-registration analysis shows the network accelerating toward the end of 2025 and spiking in the spring: at one point around 30 of the nudify apps and their websites were being registered each week. By September, 27 were live on Google Play, collectively downloaded more than 350,000 times. The volume is not a bug in the business; it is the business. Each week’s crop of nudify apps is priced, branded and listed before the previous week’s takedown notices arrive.
Disposable domains, endless supply
The redundancy is what keeps the network ahead of takedowns. Every app shares infrastructure patterns with its siblings, so removing one listing costs the operator minutes while a replacement takes hours to build and days to review. Store sweeps always arrive late because registrations, hosting accounts and payment profiles remain cheap and near-infinite across the internet. One detail captures the laziness inside the scale: apps kept their siblings’ package identifiers inside their own code, a shortcut that hands researchers the network’s family tree for free.
Rated for everyone
Many of the listings carried rating labels suitable for all age groups, putting them a search tap away from children. One of the most popular downloads, DiffAI, appeared in December 2025 and passed 100,000 installs on Google Play. A user review stated plainly that the app contained templates for creating abusive material and performed no age verification. It vanished from the store shortly before WIRED contacted Google about the network.
Takedowns are arithmetic, not strategy
Compare the snapshots and the trend becomes clear: hundreds of domains registered inside a year, a handful still live at any moment, and a steady stream of replacements waiting in review queues. A store that reports removing apps one by one is fighting a volume game with a headcount budget. Counting nudify apps removed tells you almost nothing; counting new domains registered each week tells you everything.
The Advertising Engine Behind the Nudify Apps Network
The apps never needed organic discovery because they bought it, thousands of times over. Researchers documented tens of thousands of explicit deepfake ads linked to the nudify apps network on Meta’s platforms, many reusing nearly identical advertising copy across a rotating cast of accounts and pages.
Ads that broke every rule
Katie Paul, director of the Tech Transparency Project, whose team first surfaced the ads, said they violated several of Meta’s policies simultaneously and left no question about what viewers were seeing. Earlier reporting found that hundreds of these ads contained AI-generated child sexual abuse imagery, including depictions of real children, prompting regulatory interest on multiple continents while the campaigns kept running. The transparency project’s count covered only a sample of the nudify apps ads then live, meaning the documented scale understated what reviewers had approved.
After WIRED shared the 400-plus domains, Meta reported actions across the network. The totals show how much of the machine was running at once:
The cat-and-mouse on ad platforms
Meta banned KeLing Technology Limited from advertising in 2025 after it failed to comply with a cease-and-desist letter, and says the entity continued trying to buy ads through other identities. Two-thirds of the linked ads Meta reviewed had already been removed before WIRED’s inquiry, the company said, describing criminals who hide behind fake names and run ads that look benign until a click reveals their intent.
Why the ad machine is the pressure point
Advertising is the one layer where the network must keep buying, because fresh installs only come from fresh clicks. Every other asset, from domains to store listings, can be regenerated quietly. That makes ad campaigns the choke point: verified payments, cross-account fingerprinting, copy similarity detection and rapid reporting pipelines all hit the operation where it is most exposed. The nudify apps chose a growth model that forces them to show themselves, over and over, to systems built to find them.
Reach bought, not earned
The advertising pattern matters beyond this one network. Fake payer names, pages with zero followers, machine-generated account names and near-identical copy are cheap to produce at scale, and ad systems tuned for growth will always accept them first and ask questions later. The same playbook that funded these campaigns could fund any abuse-as-a-service product waiting in the queue.
How the Nudify Apps Slipped Past App Store Review
Every major store bans nudification tools, and every major store shipped nudify apps anyway. Apple removed 19 listings, including the pet portrait app at the center of the reporting, after WIRED asked questions and shared the associated domains. Google had already removed the majority of the nudify apps before it was contacted, and it did not respond to a request for comment.
Approved first, swapped later
Apple’s explanation is precise. The apps complied with the rules when reviewers examined them; afterward, the developers changed the website the app pulled content from, replacing the approved experience with pornographic material promoted through pornographic ads on a third-party platform. Apple says it has suspended tens of thousands of apps for this bait-and-switch pattern over the last year and recently updated its rules to restrict nudify apps further.
Why pre-publication review misses this
The pattern exploits a structural gap. Reviewers approve a binary, which is a snapshot, while the app’s true behavior lives on a server the developer controls. Nothing in the listing changes when content flows turn explicit, so rating labels, store metadata and review history all stay clean. Static checks cannot catch a remote switch; only continuous monitoring can. The same gap exists on web-based versions of the nudify apps, where a single hosted interface can serve hundreds of front doors at once.
A quick side-by-side of where each gatekeeper’s defenses bent:
| Platform | The gap | What they said |
|---|---|---|
| Apple App Store | Apps pass review, then swap the content source | 19 apps removed; tens of thousands suspended for bait-and-switch; rules updated |
| Google Play | Same remote switch; many listings rated for everyone | Removed most apps before WIRED’s inquiry; no comment |
| Meta | Ad accounts and pages stitched the network together | 11,600 ad accounts and 15,000 pages actioned; KeLing banned from ads in 2025 |
Why the stores’ own numbers undercut their defenses
Apple’s disclosure that tens of thousands of apps have been suspended for bait-and-switch behavior is remarkable on its own. It confirms the tactic is endemic, not exceptional, and that the current review model produces a large, predictable leak. A store that expects thousands of apps to break the rules after approval is running a system designed to be gamed. The stores know the pattern exists; the question is whether they will fund detection at the scale the pattern demands.
What continuous review would require
Closing the gap means stores stop treating publication as the finish line. Pin apps to vetted content endpoints and re-crawl linked domains after updates. Diff remote configuration changes the way code changes are diffed, and require verified developer identity across every account in a family of listings. None of this is exotic security engineering; it is what payment networks and email providers already do, applied to the distribution channel that a billion people already trust.
Following the Trail: KeLing Technology and the Operators
Strip the branding and the network keeps a signature. A review of web domains and how the services operate points to Chinese developers and a Hong Kong-registered entity, KeLing Technology Limited. The company’s website is little more than a shell, yet its name appeared in the contact details of the first website registered in the network, and its brand name showed up inside the explicit deepfake template the apps loaded.
A shell at the center
Researchers also identified an administration login page linked to the explicit interface that mentions “Keling”, and email addresses on the company’s domain appeared on at least one app removed from stores. In late September, users of the hidden layer were told a new AI model was available: Veyra, pitched with the tagline “More real. More natural. More wild.” A website for Veyra AI listed a KeLing contact email, showing how willingly the model vendors behind nudify apps brand themselves while the operators stay hidden.
Scrubbing the trail after the call
Outreach produced silence, then cleanup. After WIRED contacted KeLing addresses and emails listed in app privacy policies, someone renamed associated social media channels, removed the company’s email addresses from the Veyra website, and pulled its registered address from the first site. The rest of the concealment kit is equally standard: generic privacy policies, free Gmail and Outlook addresses, obscure developer names and ad buyers using gibberish names that look machine-made.
The money questions nobody answers
Tech Transparency Project’s Paul says companies must answer where the money generated by nudify apps goes and how their moderation systems keep losing the race. Meta says it acted against tens of thousands of assets after WIRED shared domains, yet the cease-and-desist letter it issued to KeLing went unanswered, a sign that legal pressure alone does not move an operation built from disposable parts.
What the takedown map suggests
Registration timelines read like a weather report for enforcement. Bursts of new domains follow periods of quiet, and quiet follows every high-profile sweep, which tells researchers the operators watch the same news the public does. When one store tightens, another region’s registrations spike within days. The map suggests the network treats enforcement as predictable seasonality rather than a real threat, and plans capacity around it accordingly.
Stopping Nudify Apps: What Has to Change
The network runs on a single assumption: it will not be caught. Every week that bet pays out. Closing the gap means hard gates at every layer nudify apps already pass through, from submission and review to advertising, payments and model access. One weak layer undoes all the others, which is exactly how the current system fails.
Hard gates at the platform layer
Stores approve snapshots, so they must watch servers: pin apps to vetted content endpoints, re-crawl linked domains after updates, and treat any remote switch as a new submission. Apple’s recent rule update points the right way; continuous enforcement is the missing piece. Detection research is advancing too, as we covered in a report on a breakthrough in AI deepfake detection that spots manipulated video more reliably than older methods.
Payment and identity controls
The network monetizes through mainstream rails: coin packs, diamonds, and payments processed via Apple Pay and Google Pay. Regulators keep asking the follow-the-money questions, and the industry keeps not answering them. Payment providers can match card identifiers and business details across banned accounts, and ad platforms can require verified identities before campaigns run. That is routine banking practice applied to a frontier the platforms treated as special. Store payout schedules and developer account families are already searchable; what has been missing is the institutional will to connect them across rivals and follow the money through.
Regulation, age checks and shared lists
Laws are moving. Our coverage of xAI losing its court challenge shows a US state nudification ban surviving appeal, and our comparison of how the US, China and the EU regulate AI maps the wider ground these networks exploit. Cross-platform hash lists, mandatory age verification for any service that accepts face photos, and watermarks on generated media all shrink the space the tools operate in.
Users can lower their exposure too
Practical steps help while the platforms catch up. Check developer names and reviews before installing, and be suspicious of any “free” image app that sells coin packs. Do not upload faces you would not want redistributed, report listings that show hidden features, and never assume an age rating means the store tested the app’s server behavior. Awareness spreads faster than enforcement, and abuse only works when nobody looks. Start by treating surprise coin purchases and hidden web content as the red flags they are, and report what you find.
The bar just moved
The operators of these nudify apps bet that nobody connects the dots across stores, ad platforms and payments. Researchers and reporters did exactly that, and the map is now public. Every platform named in the reporting has the same view and the same tools. What happens next is a choice about will, and the people targeted by these products deserve better than another cycle of quiet takedowns.
References
Hundreds of Mundane AI Apps Are Hiding Sexually Abusive Deepfake Features
Nudify apps: how Meta and app stores became a shopfront
Meta Ran Hundreds of Paid Ads With Child Sexual Abuse Imagery
Meta Ran Ads That Contained AI-Generated Child Sexual Abuse Imagery
Porno disguised as AI tools: banned apps offer pornographic deepfakes
Apple Developer News: App Store rule updates on nudification apps
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.
