Meta Ran Ads for AI ‘Nudify’ App Featuring Female Politicians

Meta allowed advertisements for an AI image app that promoted the creation of sexual deepfakes using the likenesses of real people, including an ad featuring a woman closely resembling a prominent female U.S. politician.

The ads promoted an app called Kromix, which presented itself publicly as an AI image-styling tool. Reporting on the campaign found that some of its capabilities went much further, allowing users to generate sexualized synthetic images and videos based on photographs of real people.

One advertisement used the likeness of a well-known female political figure to demonstrate what the tool could produce. The campaign included 32 ads and was targeted at male users, according to reporting based on Meta’s advertising records.

The episode is particularly notable because Meta says advertisements are reviewed for compliance with its Advertising Standards before they begin running.

Meta explains its official advertising review process here

Meta removed the ads after questions were raised about the campaign. Apple also removed Kromix from the App Store after being contacted about the app’s capabilities.

The incident adds to growing scrutiny over whether large technology platforms are doing enough to stop AI tools designed for nonconsensual sexual imagery from reaching users through their own advertising and app-distribution systems.

Kromix Was Marketed as an AI Image Tool

At first glance, Kromix could appear similar to the growing number of consumer apps that use generative AI to restyle photographs.

Its marketing described it as an AI image styler and offered users different ways to transform uploaded pictures.

But material connected to the service showed that it could also be used to create sexualized synthetic content involving real people without their consent.

That distinction is critical.

An ordinary AI photo editor may change clothing colours, backgrounds, lighting or artistic styles. A nudify service instead uses generative technology to fabricate intimate imagery that makes it appear as though a real person was photographed or recorded in a situation that never occurred.

The person whose likeness is used does not need to participate, provide an original intimate photograph or give permission.

That capability has turned nudify applications into one of the most controversial uses of consumer generative AI.

One Ad Used the Likeness of a Female U.S. Politician

One of the Kromix ads demonstrated the technology by featuring an AI-generated woman who closely resembled a prominent U.S. female politician.

The advertisement first showed the political likeness in a conventional setting before presenting a fabricated, sexualized version.

The significance goes beyond the identity of one public figure.

Politicians, journalists, celebrities and ordinary users all have large numbers of publicly available images online. Those photographs can become source material for synthetic imagery without the subject ever interacting with the AI service.

Advances in image and video generation have also reduced the technical knowledge needed to produce convincing manipulations.

What once required specialized editing software can increasingly be generated using consumer applications that ask users to upload a photograph and select an effect.

Meta’s Own System Reviews Ads Before Publication

The ads raise questions because Meta does not describe its advertising platform as an unmoderated system in which advertisers can automatically publish anything.

According to Meta’s official advertising review documentation, advertisements are reviewed against the company’s policies before they go live.

Meta says that process can examine an advertisement’s images, video, text, targeting information and destination website or landing page.

The company says its ad-review system relies primarily on automation, although human reviewers are also used to improve the systems and manually review some advertisements.

Meta can also review ads again after they have begun running.

That makes the Kromix campaign an enforcement issue rather than simply a question of whether Meta has a written policy covering sexual material.

The relevant question is how advertisements promoting such a service were able to pass through the review process.

The Problem Is Larger Than One App

Kromix is not the first nudify service discovered advertising through Meta.

In July, the Tech Transparency Project reported that thousands of Facebook and Instagram ads promoted apps capable of digitally removing clothing from images of women or generating sexualized videos.

The watchdog linked many of those advertisements to GatherOne, a China-based advertising company that has served as an authorized Meta advertising partner.

The organization said it identified more than 7,500 ads promoting nudify services during its investigation.

That report was followed by further findings involving advertisements linked to exploitative AI-generated material.

The issue has now attracted attention in Washington.

U.S. Senator Demands Answers From Meta

On Aug. 18, U.S. Sen. Mark Warner of Virginia sent Meta CEO Mark Zuckerberg a letter demanding information about advertisements involving nonconsensual intimate imagery and child sexual exploitation.

In his official Senate statement and letter to Meta, Warner cited research showing that paid ads promoting nudify services had appeared across Facebook, Instagram, Messenger and Threads.

The senator questioned how such advertisements could pass Meta’s systems, given that the company reviews paid advertising before publication.

Warner also asked Meta to disclose how many impressions the identified ads received, how much advertisers paid Meta, whether the company retained that revenue and what corrective steps it had taken.

The senator gave Meta an Aug. 26 deadline to respond to a series of questions.

Warner’s letter also asks for information about ads linked to Meta’s advertising partners in China that were removed for violating policies related to nudify services.

AI Deepfakes Now Face a New Federal Legal Framework

The controversy is unfolding after the United States enacted the TAKE IT DOWN Act in May 2025.

The federal law addresses both authentic nonconsensual intimate imagery and certain digitally generated imitations.

Congress specifically defined a digital forgery to include intimate imagery created using software, machine learning, artificial intelligence, or other technological means that appears to depict an identifiable individual.

Read the full TAKE IT DOWN Act on Congress. gov.v

The law prohibits certain intentional publication of nonconsensual intimate depictions and establishes removal requirements for covered online platforms.

Covered platforms must provide a process through which an identifiable person can request removal of qualifying nonconsensual intimate imagery.

When a valid request is received, the law requires covered platforms to remove the material as soon as possible, and in any case no later than 48 hours after receiving the request.

Congress.gov provides the official status and summary of the TAKE IT DOWN Act here.

The law became Public Law 119-12 on May 19, 2025.

Its inclusion of AI-created digital forgeries is particularly relevant as synthetic images become harder to distinguish from authentic photographs.

Public Figures Are Not Immune From Deepfake Abuse

The use of a political likeness in the Kromix advertising campaign demonstrates a broader problem with generative AI.

A photograph being publicly available does not mean its subject consented to having it transformed into fabricated intimate imagery.

Public figures naturally have more photographs and videos available online than most people, which can provide AI tools with abundant source material.

But the technology can be applied just as easily to private individuals.

A profile picture, school photograph, workplace headshot or social media post may provide enough visual information for some generative systems to create convincing synthetic material.

That means the problem is not limited to celebrity deepfakes or political misinformation.

It is increasingly a personal privacy and online abuse issue.

Apple Removed Kromix From the App Store

Kromix had also passed another major technology platform’s review process.

The application was available through Apple’s App Store before reporting raised questions about its capabilities.

Apple subsequently removed the app.

According to reporting on the removal, Apple determined that the app violated its rules. The company indicated that prohibited capabilities may have been added after the application initially passed App Store review.

That possibility points to another moderation challenge.

Reviewing an application once at launch may not be sufficient if developers can later change server-side functionality, prompts, content libraries or subscription features without submitting an entirely new application.

The same challenge applies to advertising.

A harmless-looking landing page or advertisement can potentially lead users toward functionality that becomes clear only after registration or payment.

AI Moderation Is Competing With AI-Generated Abuse

Meta’s advertising platform operates at enormous scale, making automated review unavoidable.

But generative AI is also making it easier for advertisers to quickly create new versions of prohibited material.

An enforcement system may block one image, phrase or advertiser while another version appears with slightly different creative material.

Some problematic ads also use ordinary-looking preview images before directing users to more explicit services.

That creates an adversarial cycle in which platforms use AI to detect policy violations, while advertisers can use AI to alter the content they submit continually.

Meta says advertisers that repeatedly violate its policies or attempt to evade review can face restrictions on their advertising accounts and associated business assets.

Meta describes those enforcement measures in its official advertising review guide.

The Kromix episode nevertheless shows that having automated review technology and written standards does not guarantee that every prohibited campaign will be stopped before reaching users.

The Bigger Question Is Platform Accountability

The controversy is no longer only about whether nudify applications should exist.

It is also about the infrastructure that allows them to reach customers.

An AI developer may build the service, but app stores distribute it, payment providers monetize it, and advertising platforms can introduce it to large audiences.

That creates several potential points at which abusive services could be detected.

Sen. Warner’s questions to Meta focus heavily on that distinction. These were not simply user posts appearing somewhere in a social network. They were paid advertisements that entered Meta’s commercial advertising system.

The company was therefore not only hosting the platform where the ads appeared. It was also operating the system through which the advertisements were submitted, reviewed, targeted and distributed.

As AI-generated images and video become more realistic, that responsibility will become harder for large platforms to separate from the broader debate over deepfake abuse.

Kromix has been removed from Apple’s App Store and the identified Meta advertisements are no longer running, but the underlying technology is widely available.

The harder problem is preventing the next service offering the same capability under a different name from passing through the same systems again.