JupiteX Get the app
Social Issues8 Sep 2026 · about 6 min

Researchers say Meta ran 350+ ads with CSAM in 2026 so far; some featured morphed images of real children, and many linked to "nudification" apps from China (Matt Burgess/Wired)

The brief

Meta removed around 50 advertisements from Facebook, Instagram, and Threads early last month. The ads directly included child sexual abuse material, making the platforms’ role more than a simple hosting issue. They had carried abusive content into ordinary advertising feeds. Many removed advertisements also linked to “nudification” apps. These services claim to digitally remove clothing from photographs, often using artificial intelligence. In this case, the links could connect viewers with tools capable of creating sexualized images from children’s real photographs. The article says many of those apps were linked to China. The removals show that abusive imagery can enter mainstream ad systems, not only private groups or hidden websites. Researchers later identified more than 350 potentially abusive ads in 2026 so far above Meta’s reported removal of roughly 50. That gap raises questions about detection, enforcement, and how quickly harmful ads can return.

01

What did Meta remove from Facebook, Instagram, and Threads, and what was the connection to “nudification” apps?

Meta removed around 50 advertisements from Facebook, Instagram, and Threads early last month. The ads directly included child sexual abuse material, making the platforms’ role more than a simple hosting issue. They had carried abusive content into ordinary advertising feeds.

Many removed advertisements also linked to “nudification” apps. These services claim to digitally remove clothing from photographs, often using artificial intelligence. In this case, the links could connect viewers with tools capable of creating sexualized images from children’s real photographs. The article says many of those apps were linked to China.

The removals show that abusive imagery can enter mainstream ad systems, not only private groups or hidden websites. Researchers later identified more than 350 potentially abusive ads in 2026 so far above Meta’s reported removal of roughly 50. That gap raises questions about detection, enforcement, and how quickly harmful ads can return.

02

What is child sexual abuse material (CSAM), and why can an altered or AI-generated image of a real child still be treated as CSAM?

Child sexual abuse material, or CSAM, is sexual imagery involving children. It includes photographs, videos, and in some circumstances digitally created or altered depictions. The term focuses on the abusive sexualization of a child, not merely on whether a camera captured the original image.

A real child’s innocent photograph can be changed to appear nude or sexualized. Artificial intelligence can make that alteration look convincing. Even if the body or scene is fabricated, the child’s recognizable face or likeness remains part of the abuse. The child did not consent, and the image can still identify and target them.

Legal treatment varies by jurisdiction and by the image’s details, so authorities decide how specific cases are classified. But the article’s examples show why altered images matter: they exploit real children while enabling advertisers or app operators to distribute abusive content at scale. Digital manipulation does not erase the underlying harm.

03

How many potentially abusive ads did researchers identify in 2026, and how does that compare with the roughly 50 ads Meta said it deleted?

Researchers say Meta ran more than 350 ads containing CSAM or potentially abusive material during 2026 so far. This figure measures advertisements they identified, rather than only the smaller set Meta publicly said it removed. It suggests the problem extended beyond a single short-lived mistake.

Meta said it deleted around 50 ads posted across Facebook, Instagram, and Threads early last month. Comparing the figures gives a stark result: more than 350 is at least seven times 50. The numbers may not use identical methods, dates, or definitions, so the comparison is an indication of scale, not a perfect audit.

Still, the difference is important. It suggests researchers found substantially more potentially abusive advertising than Meta acknowledged removing. The gap points to possible weaknesses in detection, reporting, review, or enforcement. It also shows why independent monitoring matters when platforms assess their own advertising systems.

04

What do “nudification” apps do, and why are they especially dangerous when used to create sexualized images of children?

Nudification apps use image-manipulation technology to create the appearance that a clothed person is nude. They may rely on artificial intelligence to predict and generate parts of a body that were never photographed. The result can look realistic even though it is fabricated.

That capability becomes especially dangerous with children’s photographs. An ordinary school, family, or social-media image can become the basis for a sexualized fake. The child’s face and identity may remain visible. The app therefore turns an innocent image into material that can humiliate, threaten, exploit, or further target the child.

The article connects many of the reported advertisements to nudification apps, including services linked to China. Their promotion through major platforms can normalize abuse and make access easier. Removing advertisements is necessary, but reducing harm also requires app stores, payment providers, advertisers, and platforms to block abusive services and protect victims’ images.

05

What can happen to children whose real images are morphed, sexualized, or used in abusive advertisements without their consent?

Children whose real photographs are morphed into sexualized images can suffer serious privacy and emotional harm. They may feel fear, shame, anxiety, or loss of control over their identity. The image can make them recognizable even when the sexual content itself was fabricated.

Once distributed, an image can be copied, reposted, or attached to advertisements. Strangers may view, save, or share it. Abusers can also use it for harassment, coercion, or blackmail. A child may have to explain or defend an image showing something that never happened, while platforms struggle to remove every copy.

The article shows that advertising systems can help propel this material beyond a private act. Ads can connect abusive images with tools that produce more of them. Effective responses therefore need rapid removal, victim support, evidence preservation, and cooperation among platforms and investigators. Protecting children also means preventing their images from being reused in the first place.

06

Who is involved in creating, buying, approving, distributing, and reporting these ads, and what responsibilities does each party have?

The people and organizations involved can include image creators, nudification-app operators, advertisers, ad buyers, Meta, users, researchers, and law-enforcement agencies. Creators and buyers must not make or purchase sexualized images of children. App operators and advertisers must reject abusive content and avoid promoting tools for exploitation.

Meta controls the advertising systems on Facebook, Instagram, and Threads. It should screen ads, verify advertisers, detect repeat offenders, respond to reports, remove violations, preserve evidence, and notify authorities where required. Human reviewers and automated tools both have roles. Users and researchers can report suspicious ads and document patterns, while investigators can pursue criminal conduct.

Responsibility is shared, but it is not identical. Platforms have the strongest ability to stop paid distribution at scale. Advertisers and app companies control what they sell and promote. Reporters help expose failures. Clear rules, rapid action, transparency, and cooperation are needed because one group cannot reliably prevent this abuse alone.

07

How do online advertising systems review and target millions of ads at scale, and why can harmful ads evade automated and human moderation?

Online advertising systems must review enormous numbers of submissions quickly. Automated tools can scan text, images, landing pages, account behavior, and payment details. Systems also use rules and machine-learning models to decide whether an ad violates policy. Higher-risk or reported material can then be sent to human reviewers.

These defenses are imperfect. Abusers can disguise images, change wording, use new accounts, or direct ads to outside websites. Altered images may not match known abuse databases. Automated systems can miss context, while human reviewers face speed, language, and volume limits. A legitimate-looking advertisement may hide its purpose behind a link.

The article’s 350-plus identified ads, compared with Meta’s roughly 50 removals, shows why scale matters. Platforms need stronger pre-publication checks, advertiser verification, landing-page review, rapid reporting channels, and independent testing. No system can guarantee perfect detection, but layered controls can make abuse harder to buy and distribute.

This brief was written by AI from the original reporting and checked by other models. Names, figures and quotes come from the source; read it for full context.

Read more in the JupiteX app

Pulse is free. New stories every 4 hours, each one broken into the questions that explain it.

Or read more news on the web