Deepfake apps are back in the news, and this time they were hiding in plain sight. A WIRED investigation published on November 2, 2026 found a network of more than 400 linked apps and domains. They looked like pet photo tools, language tutors and AI workout coaches. Under certain conditions, they loaded a hidden web interface for making explicit deepfake images and videos.

I run delivery for app and web projects at Debugged Pro, so this story hit close to home. It is not only a platform problem. It shows how remote content, ad funnels and weak review notes can be abused. Here is what happened and what agencies in India and the USA should change in their own release process.

Photo illustration of women passing through bars that distort their image, used by WIRED for its deepfake apps investigation — Lead image from WIRED's report. Photo-Illustration: Jobanny Cabrera; Getty Images, via WIRED.

What WIRED Found Inside the Deepfake Apps Network

The investigation was written by Matt Burgess. Researchers found the network after tracing links in thousands of explicit deepfake video ads on Facebook and Instagram. Those ads pointed to apps in the Apple App Store and Google Play.

Here are the main facts reported:

Scale: more than 400 connected domains and apps, many posing as everyday services.
Speed: at one point, around 30 nudify apps and sites were registered each week.
Reach on Android: 27 apps were live on Google Play in September, with more than 350,000 combined downloads.
Age ratings: many of the apps were rated suitable for all age groups.
Payments: users could buy "diamonds" to generate images, paying through Apple Pay and Google Pay.
Operator: the apps appear to be run by Chinese developers and a Hong Kong registered entity called KeLing Technology Limited.

One app, DiffAI, was created in December 2025 and passed 100,000 downloads on the Play Store. It was removed shortly before WIRED contacted Google.

Chart of key numbers from the WIRED deepfake apps investigation including 400 plus domains and 350,000 downloads — Key figures from WIRED's November 2, 2026 report. Chart: Debugged Pro.

How the Bait-and-Switch Worked

The trick was simple and that is what makes it worrying. Each app showed normal content to most people, so it looked clean during review. The explicit layer lived on a server, not inside the app.

Paul Bouchaud, a lead researcher at the nonprofit AI Forensics, reverse engineered the apps. Bouchaud explained that each download triggers a call to a server. The call shares details such as where the user came from, their language and a unique identifier. "The server replies whether or not to show the explicit or the non-explicit version of it," he told WIRED.

One trigger seemed to be installing the app after clicking a related ad on Meta's platforms. Visiting the linked websites directly often showed only a black screen.

The setup also had a weak spot. A JavaScript bundle on the websites listed the whole network, including app package details, encryption keys and ad trackers. Maël Le Touz, a staff threat researcher at Infoblox, identified almost 500 linked domains this way.

What Apple, Google and Meta Said

Apple removed 19 apps, including the pet photo app Viefur. Its spokesperson said the apps "complied with our rules when we reviewed them." Developers later swapped the website the apps pulled content from. Apple says it suspended tens of thousands of apps for bait-and-switch tactics over the last year.
Google did not comment, but had removed most of the apps before WIRED reached out.
Meta said it banned KeLing Technology Limited from advertising in 2025. After WIRED shared the domains, Meta said it took action. That covered 11,600 ad accounts, more than 4,700 business pages and 15,000 pages, and blocked 209 apps from advertising.

WIRED also noted that similar ads and apps were still being published as it finished reporting.

Why Deepfake Apps Matter to Legitimate App Agencies

Most agencies will never build anything like this. So why care? Because the same tools the abusers used are standard parts of modern apps.

Remote config lets you change what an app shows without a new release.
Web views load live web content inside a native shell.
Ad attribution tells the app which campaign brought each user in.

Combine those three and you can show different users very different apps. That is exactly what reviewers will now look at more closely. When stores tighten checks, honest teams feel it too, through slower reviews, more questions and more rejections.

There is also a brand risk. If your client runs an AI photo or avatar feature, users and journalists will ask harder questions after a story like this. You want clear answers ready.

What the App Store Rules Actually Say

It helps to read the rules directly instead of relying on forum posts. Apple's App Review Guidelines, last updated October 16, 2026, include several clauses that apply here:

Guideline 2.3.1(a): all new features and product changes must be described with specificity in the Notes for Review. Generic descriptions will be rejected.
Guideline 2.5.2: apps may not download or execute code that introduces or changes features or functionality.
Guideline 2.5.19: no hidden, dormant or undocumented features. Hiding them through obfuscation can end your developer account.
Guideline 1.1.4: overtly sexual or pornographic material is not allowed.

Google is just as direct. Its AI-Generated Content policy for Google Play says AI content apps must include in-app reporting or flagging. Users should be able to report offensive content without leaving the app. Google's related guidance lists AI-generated non-consensual deepfake sexual material as a clear example of violating content.

A Deepfake Apps Safety Plan for Agencies in India and the USA

Here is the process we now recommend for any client app with generative AI, especially image or video features. It works for a Bengaluru startup and a New York retailer alike.

1. Write honest, specific review notes

List every AI feature, what it generates and which servers it calls. If the app loads remote content, say so and explain who controls it. Vague notes save ten minutes and can cost weeks.

2. Put remote content under change control

Treat remote config and web view content like code. Changes should need a pull request, a reviewer and a log. If a store asks what changed after approval, you should be able to answer in minutes.

3. Build reporting into the first version

Add a report or flag button next to every AI output. Route reports to a real queue with an owner and a response time. Google Play requires this for AI content generators, and it is good practice on iOS too.

4. Restrict edits of real people

If users can upload photos of people, add consent checks and content filters before generation. Block known abuse prompts and test them in QA, the same way you test login flows.

5. Match the age rating to the real product

Many apps in the WIRED network were rated for all ages. Answer rating questionnaires based on what the AI can produce, not only what the demo shows.

6. Audit the ad funnel

Review ad creatives, landing pages and deep links before launch. Make sure every campaign leads to the same reviewed experience. If a growth vendor runs ads for your client, ask for access to the ad library.

Six step AI app release checklist covering review notes, remote config, reporting, consent, age ratings and ad audits — A simple release checklist for AI apps. Graphic: Debugged Pro.

A Practical Example

Say a client wants an AI pet portrait feature. Users upload a pet photo and get a cartoon version.

Here is how we would ship it safely:

Detect and reject uploads that contain human faces, since the feature is for pets.
Run every output through a safety filter before showing it.
Add a one-tap "Report this image" option that sends the image ID to the support team.
Keep style templates in a versioned config file, reviewed before each change.
Describe all of this in the review notes, including the image model provider.

None of this needs a big budget. Most of it is standard engineering work that fits into a normal sprint, and it protects the client's developer account.

What This Means for the AI App Market

Expect stricter review for AI photo, avatar and video apps over the coming months. Apple says it recently updated its rules to further restrict nudify tools. Regulators are also investigating the ads, according to WIRED.

For honest builders, the best defence is a paper trail. Clear notes, logged changes and working report tools show reviewers you have nothing to hide. Tracking how Apple handles AI apps more broadly? Our post on Apple AI and third-party apps on the App Store covers the revenue side. For a related policy shift from an AI vendor, see how Claude Code now blocks game clones.

Need a second look at an AI feature before launch? Our software development consulting team can run a compliance and safety check on your build.

Key Takeaways

Deepfake apps hid behind harmless fronts like pet photo and fitness tools, and passed store review.
The switch happened on the server, based on user details and ad clicks, after approval.
Apple removed 19 apps, Google removed most of the Play listings, and Meta acted on 11,600 ad accounts.
Store rules already ban hidden features, undocumented changes and non-consensual deepfake content.
Agencies should document AI features, control remote content, add in-app reporting and audit ad funnels.

About the Author

Nitin Kumar is the Lead Project Coordinator at Debugged Pro. He manages web and mobile app delivery for clients in India and the USA, with a focus on release planning, QA and platform compliance. He writes about how AI news affects the way agencies build and ship software.