Hundreds of Innocent-Looking AI Apps Secretly Hid Sexual Deepfake Tools, Researchers Find as Apple Pulls 19

SAN FRANCISCO — A network of more than 400 apps and websites posing as ordinary AI tools secretly offered features for creating sexually explicit and violent deepfakes, according to an investigation that has prompted Apple, Google and Meta to take action.

The apps, which appeared in Apple's App Store and Google Play, presented themselves as everyday services such as pet-photo editors, language tutors and AI workout coaches. But under certain conditions, they loaded a hidden interface that allowed users to generate explicit images and videos from photos they uploaded.

The findings were published Monday by Wired, based on research by the European nonprofit AI Forensics.

How the apps worked

One example was Viefur, an app that appeared to let users dress up photos of their pets in human clothing.

But when accessed under certain conditions, apps in the network stealthily loaded a web interface for generating sexual deepfakes, the investigation found. The apps could display different content depending on where a person had downloaded them, the language they used and a unique identifier assigned to their device.

One trigger appeared to be installing an app after tapping an advertisement for the deepfake features on Meta's platforms, such as Facebook and Instagram.

Users could then access dozens of explicit photo and video templates and buy credits to generate deepfakes from uploaded images. Payments were processed through Apple and Google's systems.

Researchers said some templates depicted sexual violence and threats, including scenes in which a man holds a knife to a woman's neck.

A large network

Paul Bouchaud, lead researcher at AI Forensics, reverse-engineered the apps and found code containing a list of related domains and apps, along with encryption keys and Facebook advertising trackers.

AI Forensics said it first spotted the operation in early September. At one point, about 30 new app websites were being registered each week.

Researchers found 27 of the apps live on Google Play in September, with more than 350,000 combined downloads. A separate researcher at cybersecurity firm Infoblox identified nearly 500 linked domains.

The network appears to have been publishing the apps for about a year. Researchers linked it to Chinese-language developers and to a company registered in Hong Kong.

Tech companies respond

Apple removed 19 apps after being contacted by Wired. The company said the apps changed after passing its review process.

Google had already removed most of the apps before Wired contacted it. Google Play's policies ban AI-generated nonconsensual sexual content and require developers of generative AI apps to prevent prohibited material.

Meta said it took action against 11,600 ad accounts and 15,000 pages linked to the network.

"We have strict rules against promoting nudify apps," a Meta spokesperson said, adding that the company works aggressively to keep such material off its services.

Meta has previously sued a Hong Kong-based operator over ads promoting nudify services.

A growing problem

The investigation highlights the challenge tech platforms face in policing so-called nudify apps, which use AI to create fake explicit images of real people without their consent.

Such images can be used to humiliate, harass or extort victims, and researchers say women and girls are most often targeted.

App stores rely on review processes to catch harmful apps, but the network's ability to change its behavior after approval allowed it to slip past those checks.

Governments have also stepped up pressure. San Francisco has moved to force Apple and Google to remove dozens of nudify apps it says create nonconsensual intimate deepfakes.

What's next

Despite the removals, researchers warned that related ads and apps have continued to emerge, suggesting the operators may try to evade enforcement by launching new apps.

Experts say stronger coordination among app stores, advertising platforms and regulators will be needed to stop such networks.