WIRED investigation exposes hidden abusive deepfake tools in ordinary AI apps

What WIRED found

WIRED's November 2 investigation identifies more than 400 connected apps and web domains. Some presented themselves as everyday AI services, including pet portrait tools, but concealed features for generating sexually abusive imagery from uploaded photographs.

The reporting connects app distribution with advertising on Facebook and Instagram. Read Matt Burgess's original investigation for its evidence and the companies' responses.

Earlier research raised similar concerns

This follows months of scrutiny. In an April 15 report announcement, the Tech Transparency Project said searches surfaced 46 distinct apps in Apple's store and 49 in Google Play. Its testing found that roughly 40% could produce nude or scantily clad depictions of women.

TTP also examined discovery. Its findings described search suggestions and advertising that directed people toward these services. That puts the recommendation systems around an app under scrutiny, alongside the app itself.

Those figures belong to TTP's earlier sample. They should not be added to WIRED's count or treated as a separate count of confirmed operators.

Copyleaks published another investigation on August 18. Its researchers identified nearly 50 App Store apps claiming to generate explicit deepfakes over a six-month period. The report described understated store descriptions, subscription-driven access and other patterns.

These studies used different samples and methods. Together, they raise a practical question: how much can a shopper learn about an AI app from its public listing?

What Apple's rules already require

Apple's App Review Guidelines address both content and disclosure. Guideline 2.5.19 prohibits hidden, dormant or undocumented features. It says functionality should be clear to users and App Review, with removal and developer-account termination among the possible consequences for concealment.

Guideline 1.1.4 prohibits overtly sexual or pornographic material. Guideline 1.2 covers user-generated content and AI creation apps, requiring filtering, reporting mechanisms, the ability to block abusive users and published contact details.

Apple also describes a limited allowance for incidental mature material from web-based services, hidden by default and enabled through the website. That sits alongside its moderation requirements. It is not a general exemption for concealed functionality.

The policy question is therefore partly about enforcement. A written prohibition sets expectations; detecting a service that behaves differently after approval requires ongoing checks.

What this means when choosing an AI tool

For NaughtyPicks readers, our takeaway is to look beyond polished screenshots. Compare the product description with what the service actually asks you to upload, the permissions it requests and the charges it explains before payment.

An unexpected change in purpose deserves attention. If a service sold as a routine photo utility introduces a different set of features, pause before giving it more personal material. A familiar storefront or payment screen is only one part of assessing a product.

Adult fictional character creation and intimate edits of identifiable people also raise different consent questions. Someone making their own fictional character has a different relationship to the material than a person whose photograph has been taken and altered without permission.

Help if an intimate image is being shared

StopNCII.org explains a reporting tool that creates a digital fingerprint of intimate images or videos on your device. The fingerprint is shared with participating platforms; the original file stays on the device.

Those platforms look for matching material and apply their intimate-image abuse policies. The service cannot remove content from the entire internet. Check its eligibility information and participating-platform list before starting a case.