Meta runs ads for Kromix app that creates sexual deepfakes

Summary

Meta allowed 32 ads for the Kromix AI app, which is used to generate sexually explicit deepfake videos of political women, to run on its platforms. This violates the company's explicit policies prohibiting such content. The ads were targeted exclusively at male users and were removed only after an investigation by WIRED. This incident occurs amid a wave of increased scrutiny regarding non-consensual intimate imagery generated by AI. xAI's Grok app also generated approximately 3 million sexually explicit images in just 11 days.

Why it matters

This situation highlights the gap between Meta's stated ad review policies and the content that actually passes through its systems, undermining trust in the platform's content moderation claims. Since nearly 98% of Meta's revenue comes from advertising, the failure to block a campaign that clearly violates policy has direct financial and legal consequences. This incident coincides with regulators in California and San Francisco increasing sanctions against app stores and platforms hosting undressing tools, thereby raising risks for all major tech companies.

Highlights

• The ads were published 32 times (each running for 5 to 46 hours) and were removed only after WIRED contacted Meta; they were not detected during routine review processes.
• The ad page was created on August 3, days before the campaign was launched. This suggests a deliberate strategy of short-term activation.
• One of the ads featured a prominent female politician next to an image resembling the U.S. President, accompanied by a pornographic headline. The ad was targeted solely at male users.

Critical take

The article relies heavily on a single WIRED report and the CCDH/NYT investigation into Grok. However, Meta's internal ad review records have not been independently verified. There is also no evidence provided regarding whether the Kromix ads were approved by automated systems or human moderators. Consequently, the fundamental question of *how* the policy violation occurred remains largely unanswered. The article also conflates the glitch in Meta's ad platform with xAI's separate deepfake scandal. While contextually relevant, this risks blurring accountability for the specific incident.