Meta's Ad Review Lets AI 'Nudify' App Target Female Politicians

Meta's advertising platform once again became a vector for nonconsensual intimate imagery, this time aimed at female politicians. The company ran ads on Facebook and Instagram for Kromix, an AI tool that bills itself as an "AI image styler," but which in practice allows users to upload photos and place real people—including prominent female U.S. politicians—into pornographic scenarios. The ads, exclusively targeted at male users, featured a voice-over promoting the tool as having "no restrictions," with one spot showing a woman resembling a well-known politician winking before cutting to an explicit scene with the same face.

The ads were removed only after WIRED inquired about them, and the Kromix app was pulled from Apple's App Store following a similar request for comment. Apple said the app "appeared to have added prohibited content and features after initial review," while Meta's spokesperson Cindy Southworth acknowledged the company is "continuing to investigate why these ads were able to bypass our systems." The account running the ads existed for about a week, had zero followers, and used benign cover photos—beaches, mountains, waterfalls—to obscure the explicit content. That this still slipped through points to a systemic fragility in how Meta reviews ads.

For developers and technical readers, Kromix is a reminder of how easily modern AI tooling can be weaponized. The app itself is not sophisticated—it leverages existing image-generation and face-swapping techniques to produce deepfakes that are convincing enough to harm. The real gap is in detection: Meta says it relies primarily on automated tools to review ads against its own policies, yet the platform has repeatedly failed to catch ads for nudify apps. The company removed 344,000 such ads between November and January, and has filed lawsuits against developers, but researchers at the Tech Transparency Project continue to find new examples, including ads containing AI-generated child sexual abuse material.

The episode also underscores a growing tension in platform governance: as Meta leans harder on AI to moderate content—while reportedly laying off human reviewers—bad actors adapt by obfuscating ads with benign first frames or by registering fresh accounts. For the developer community, this is not merely a policy story but a technical one. The same low-code tools that enable creative expression also lower the barrier for abuse, and the industry has yet to build reliable, scalable defenses. As Katie Paul of the Tech Transparency Project put it, "There appears to be zero quality control" on the platform's core product.

Politicians, overwhelmingly women, have long been targets of nonconsensual deepfakes, and official portraits are now raw material for abuse. This case shows that even with explicit policies, automated review, and legal action, the machinery of content moderation lags behind those who abuse it—a problem that will only intensify as generation and face-swapping tools become more accessible.