Meta Platforms (NASDAQ: META) ran advertisements for an app called Kromix that generated non-consensual nude deepfakes of real people, including a prominent U.S. politician, before investigators forced the company to act.
Kromix marketed itself as “the AI that men actually use,” running 32 ads on Meta’s platforms under the tagline “Unleash Next-gen AI Magic,” targeting exclusively male users.
The ad account was created in early August with no followers, and individual ads ran for between five and 46 hours before being removed after WIRED made inquiries.
Most of the ads received fewer than 10 impressions, according to Meta’s own ad library, suggesting the campaign had limited reach but still cleared the platform’s review systems.
Kromix was also available on Apple’s App Store, where it allowed paid users to upload photos and place subjects into scenarios the app explicitly labeled, including “bedroom rape” and “Disney love.”
Apple removed the app after WIRED reporters Vittoria Elliott and Matt Burgess requested comment, with the company stating that the developer had added prohibited content after the app’s initial review.
Meta has long-standing policies against non-consensual intimate imagery, and Apple maintains explicit policies prohibiting nudification applications, yet neither platform identified Kromix until journalists intervened.
The incident reflects a structural problem for the technology industry, as generative AI has made it increasingly easy to produce convincing nude imagery from a single ordinary photograph.
Platform moderation systems were not designed to handle the volume or sophistication of content that modern generative AI tools can now produce at scale.
Both Meta and Apple employ thousands of trust and safety professionals and operate AI-based detection systems, yet those layers of review failed to flag Kromix before it ran any advertisements.
The fact that it required outside journalists to surface the problem points to one of three conclusions: detection systems are failing, human reviewers are missing violations, or both failures are occurring simultaneously.
Women in public life, including the unnamed U.S. politician whose likeness appeared in one of Kromix’s promotional videos, are disproportionately targeted in non-consensual intimate imagery incidents of this kind.
The episode will intensify ongoing pressure from regulators and lawmakers on major platforms to demonstrate that self-regulatory frameworks are adequate to handle AI-generated abuse at scale.
