Meta Ran Ads Promising to Nudify Female Politicians

In what is becoming a quarterly tradition as reliable as pumpkin spice lattes or iPhone announcements, Meta’s advertising platform has once again demonstrated that its content moderation system operates with all the discernment of a goldfish with a concussion. This time, the company—whose motto may as well be “Move Fast and Break Things, Including Basic Human Decency”—approved and ran advertisements for an app explicitly promising to generate nonconsensual deepfake pornography of female politicians.

The app in question, charmingly named Kromix (presumably because “Digital Violation Emporium” tested poorly with focus groups), bills itself as an “AI image styler.” Its marketing pitch is refreshingly honest in its horribleness: “No restrictions,” the ads promise, “all the characters are real people,” and—here’s the pièce de résistance—”the AI that men actually use.”

Because nothing says “sophisticated artificial intelligence deployment” quite like helping lonely men weaponize democracy against the women participating in it.

The advertisement itself was a masterclass in subtlety: a woman bearing an uncanny resemblance to a prominent female US politician stands before American flags while the caption asks, “What if she moved?” She winks. The video then cuts to the same face performing in pornographic content. It’s the kind of seamless transition that would make a film school dropout weep with envy, if it weren’t, you know, completely abhorrent.

Meta, ever the responsible steward of the digital public square, ran 32 of these ads for between five and 46 hours each. Most received “10 or fewer impressions,” which is corporate speak for “we’re going to mention this number so you think it wasn’t a big deal, even though our platform actively promoted nonconsensual synthetic pornography of elected officials.” The ads were exclusively targeted to men, because apparently Meta’s algorithm has determined that the male demographic is just one poorly-targeted ad away from becoming amateur deepfake pornographers.

When contacted by WIRED for comment—because nothing motivates Meta to action quite like the looming threat of public embarrassment—the company removed the ads and issued the now-familiar statement about how they “work aggressively” to fight sexual exploitation. This is, of course, the same Meta that has removed over 340,000 ads for nudify apps in recent months, which raises the obvious question: If your automated systems are catching hundreds of thousands of violating ads, perhaps the problem isn’t a few bad actors but rather your entire business model?

Cindy Southworth, Meta’s head of women’s safety, explained that the ads used “adversarial tactics” like covering the video with benign images of beaches and waterfalls before revealing the pornographic content. This is apparently a sophisticated evasion technique that Meta’s AI moderators couldn’t detect, despite the fact that the account was created mere days before the campaign, had zero followers, and was explicitly selling a tool for generating nonconsensual intimate imagery. One imagines the AI moderation system squinting at the waterfall image and thinking, “Seems fine to me—wait, is that a nipple behind the cascading water? No, just a rock. Approved!”

Apple, not to be outdone in the race to the bottom of corporate responsibility, also hosted Kromix on the App Store until WIRED came knocking. The app invited users to upload photos for scenarios including “bedroom rape” and “Disney love,” which is either the most disturbing Venn diagram ever created or just another Tuesday in Silicon Valley. Apple, whose App Store guidelines are enforced with the same consistency as a weather forecast, removed the app and issued a statement about how they “aggressively work to identify and remove” such content—apparently between rounds of golf and counting their 30% commission.

This latest incident comes on the heels of Meta running ads containing AI-generated child sexual abuse material, which the company also removed only after journalists pointed it out. At this point, Meta’s ad review process appears to consist of an intern occasionally glancing at a dashboard between TikTok breaks, shrugging, and saying, “Eh, probably fine.”

Katie Paul of the Tech Transparency Project noted that Meta has increasingly relied on AI to vet advertisements while laying off human moderators. This is the Silicon Valley way: replace judgment with automation, consequences be damned, and then express shock—shock!—when the machines approve ads for apps that promise to strip the clothes off elected officials. “Ads are Meta’s primary product,” Paul observed, “and there appears to be zero quality control.” Which is unfair, really. There is quality control—it’s just calibrated to maximize engagement and ad revenue, not to prevent the platform from becoming a distribution network for digital sexual violence.

The Kromix affair is merely the latest chapter in Meta’s ongoing saga of pretending to care about platform safety while its automated systems continue to monetize human misery with the efficiency of a German train schedule. The company will issue statements, file lawsuits against a few developers, maybe roll out a new AI detection tool that will catch 99% of violations (and miss the 340,000 that slip through), and then wait for the next scandal to break.

After all, why fix the fundamental brokenness of your content moderation system when you can just apologize after the fact and keep cashing those checks? In the attention economy, there’s no such thing as bad publicity—only ad impressions. And Meta, apparently, is happy to sell them all.

Sources include: https://www.wired.com/story/meta-ran-ads-for-an-app-promising-to-nudify-female-politicians/