Meta's Ad System Let Through Deepfake Porn Targeting Female Politicians

Meta's advertising platforms — Facebook and Instagram — ran paid promotions for an AI tool called Kromix that explicitly encouraged users to generate pornographic deepfake videos of real women, including a prominent female U.S. politician. One ad featured a woman closely resembling the politician standing before American flags, then cut to a pornographic scene featuring the same face. The ads were exclusively targeted at male users under the tagline 'Unleash Next-gen AI Magic,' and the Kromix app itself offered scenarios labeled 'bedroom rape' alongside other sexually explicit AI-generated content. The account running the campaign was created on August 3, had zero followers, and ran 32 separate ads for between five and 46 hours before being removed — only after journalists asked Meta about them.

The discovery was made by researchers at the Tech Transparency Project, which flagged the ads to journalists. Apple, which had hosted Kromix in its App Store, pulled the app after being contacted for comment, saying it appeared the developer had added prohibited features after the initial review. This follows San Francisco's city attorney sending Apple a cease-and-desist letter earlier this summer demanding removal of 13 specific nudify apps and accusing the company of enabling the distribution of explicit deepfake imagery.

Meta's own advertising standards require that each ad be reviewed — primarily through automated tools — before going live. The company says it removed over 344,000 violating nudify-related ads between November 2025 and January 2026 alone, has filed a lawsuit against at least one nudify app developer, and recently deployed new AI to catch violations. But critics note the review system failed here despite obvious red flags: the Kromix account was brand-new, had no followers, and ads used innocuous cover images — beaches, mountains, waterfalls — to disguise pornographic content. Meta acknowledged the tactic but said no system is perfect.

This incident is part of a wider pattern. Separately, the same research organization identified more than 50 Meta ads containing AI-generated child sexual abuse material, including face-swapped videos of minors in sexual situations. Those ads ran between November 2025 and August 2026 — some of them after Meta publicly announced new AI detection capabilities — and reached audiences across the U.S., UK, and at least a dozen European countries. Another app called MaskAI, linked to some of those ads, was also removed from the App Store after press inquiry.

The broader ecosystem enabling this content is rapidly democratizing. Researchers studying the problem note that formerly mundane material — official political portraits, press conference footage — now serves as raw material for AI tools requiring little technical skill. Female politicians have been disproportionate targets of nonconsensual deepfake content for years; in January, roughly 3 million sexualized AI-generated images were created via Grok, hitting officials in Sweden, England, and Northern Ireland. Critics argue Meta's increasing reliance on automated moderation — while reducing human reviewer headcount — is fundamentally incompatible with catching sophisticated adversarial ad campaigns in real time.