Hundreds of Deceptive AI Applications Discovered Hiding Explicit Deepfake Features

An international investigation has uncovered a sprawling network of approximately 400 applications and websites that masquerade as everyday digital services while secretly harboring tools to generate explicit and violent AI deepfakes [1]. The operation, which bypassed safety reviews on major app stores, used targeted advertising on social media to direct users toward illicit content [1].

Disguised as Everyday Services

Researchers discovered a network of disposable apps and domains that initially present themselves as mundane utilities, such as pet portrait generators, language tuition tools, or fitness coaches [1]. For instance, an application titled Viefur presents as an AI pet portrait creator where users purchase digital coins to place animals into humorous clothing and settings [1]. However, researchers found that these interfaces serve as a cover for a hidden backend that generates sexually abusive deepfake imagery and videos [1].

According to security analyses, the applications utilize conditional loading upon download [1]. By sending data such as user location, language, and identifiers to a central server, the application determines whether to display the benign interface or load a hidden web layer containing explicit video and image templates [1]. One trigger for the explicit interface appears to be downloading the app via promotional links hosted on third-party advertising platforms [1].

Scale and Operational Structure

Investigations by threat researchers from AI Forensics and Infoblox revealed that the network comprises nearly 500 linked domains sharing common codebases, encryption keys, and configuration lists [1]. Analysis of registration dates indicates that the apps began appearing in late 2025, accelerating significantly during the spring with peaks of roughly 30 registrations per week [1]. By September, dozens of these applications were live on mainstream digital marketplaces, accumulating hundreds of thousands of downloads collectively [1].

One popular application named DiffAI, launched in December 2025, amassed over 100,000 downloads on the Google Play Store before its removal [1]. Security experts note that the underlying code, hosting infrastructure, and video editing watermarks point toward Chinese-language developers, with a Hong Kong-registered entity named KeLing Technology Limited identified in early domain registration records [1]. Developers used obscure names, generic privacy policies, and free email services to mask their connections [1].

Platform Distribution and Moderation Challenges

Propagation of the network relied heavily on tens of thousands of video advertisements published across Meta platforms, including Facebook and Instagram [1]. These promotional clips frequently utilized explicit deepfake depictions—including imagery involving minors—to drive traffic to the applications [1]. Users were able to process payments for generation credits using standard payment gateways integrated within Apple and Google app marketplaces [1].

Following inquiries, platform operators took enforcement actions [1]. Apple removed 19 applications from its App Store, including Viefur, stating that the developers utilized bait-and-switch tactics by altering backend website content after initial review [1]. Google similarly removed the majority of the associated applications ahead of formal outreach [1]. Representatives from civil society organizations emphasized that the scale of the operation underscores ongoing vulnerabilities in digital platform moderation and safety verification pipelines [1].

The proliferation of generative artificial intelligence has led to a surge in unregulated applications capable of generating non-consensual explicit imagery and deepfakes. Major technology platforms, including Apple, Google, and Meta, maintain strict policies prohibiting harmful deepfake content, sexually explicit material, and child sexual abuse material on their respective marketplaces and ad networks. Despite these safeguards, malicious developers frequently employ stealth techniques—such as post-approval server updates and disguised landing pages—to evade automated review processes.

The presence of deceptive applications on mainstream app stores highlights persistent security challenges in digital distribution channels. For technology users and regulatory bodies, the incident underscores the difficulty of policing automated content generation tools and preventing the monetization of illicit media through standard payment systems.

Technology platforms may face intensified regulatory pressure globally to strengthen pre-publication audits and post-download behavioral monitoring of mobile applications. Increased scrutiny could also target payment processors and advertising networks that inadvertently facilitate the monetization and promotion of unauthorized deepfake services.

Open questions remain regarding the total financial proceeds generated by the network and the exact methods used to consistently bypass automated content moderation filters across multiple tech ecosystems. Furthermore, the complete identity and full extent of the developer network behind KeLing Technology Limited remain unconfirmed.