- Meta now checks where an ad sends people, as well as what it shows, to catch ads hiding links to child abuse material.
- A new language model hunts “signposting,” benign-looking ads suspected of steering people to illegal content off Meta’s platforms.
- Meta acted on 33.2 million pieces of child sexual exploitation content on Facebook and Instagram from January to June 2026.
Meta now looks at where an ad takes people and what it shows. The move, announced Wednesday with new AI tools, targets ads that quietly steer users to child sexual abuse material hosted off Facebook and Instagram.
Meta said predators have come into advertising. An ad can appear innocuous on its own and still lead people to illicit material on outside sites.
Meta didn’t stop at the ads reported to it. After working out the scheme, the company expanded its investigation across its platforms, disabling the accounts behind it and blocking the off-platform links involved.
Red-teaming agent joins PhotoDNA, in use since 2011
Meta’s new large language model looks for what it calls “signposting.” Meta describes it as ad content that appears harmless but is strongly suspected of secretly directing people to illegal content or harmful activity off Meta’s platforms.
More AI-powered scans of ad content look for exploitation material missed by previous systems. Meta says it continues to layer in signals as the investigation unfolds.
A red-teaming AI agent is probing Meta’s own defenses for weak spots, so new tactics come to light before they scale. Better recidivism detection is intended for bad actors creating new accounts after being removed.
Meta has been running PhotoDNA photo and video matching on its apps since 2011 and offers image and video hashes to other companies through the Tech Coalition’s Lantern program.
Meta looks for ads, posts, and comments containing a blocked link and erases them. Ads with blocked links will be rejected upon upload.
Proactive detection found over 97% of 33.2 million exploitation items
Meta acted on 33.2 million items of child sexual exploitation content from Facebook and Instagram globally between January and June 2026. More than 97% was detected before anyone reported it.
In India, the same six-month number was 5.3 million pieces, over 98% caught first. In September, Meta said India’s National Cyber Crime Reporting Portal, run by the Indian Cybercrime Coordination Center, would get its child safety cases directly.
The tools land while Meta faces child safety claims in court. In August, it settled with 29 US states for up to $18 billion, without admitting their claims.
The states alleged that Meta engineered Instagram and Facebook to addict children. Before Meta’s settlement, opening statements were underway in Oakland federal court, where Meta told the jury it didn’t design its apps to addict children, Cryptopolitan reported.
EU consumer group BEUC accused Meta, Google, and TikTok of failing to pull scam ads in May, Cryptopolitan reported.
Meta brushed aside the claims. It said it removed more than 159 million scam ads in 2025, and 92% were removed before any user reported them.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.
FAQs
What did Meta actually announce?
New AI tools for ad review, including a model that detects "signposting" ads pointing people to child abuse material off its platforms.
How much exploitation content did Meta remove in 2026?
Meta acted on 33.2 million pieces on Facebook and Instagram from January to June 2026, over 97% found before anyone reported them.
Why is Meta rolling this out now?
Predators moved into ads that look harmless but steer people to illegal material hosted off Meta's platforms, the company said.
Randa Moses
Randa Moses is an editor and reporter at Cryptopolitan covering tech, AI, robotics, crypto, scams, and hacks. She has worked in the crypto space since 2017. She held roles at Forward Protocol, AmaZix, and Cryptosomniac. Randa holds a degree in Electrical and Electronics Engineering from the University of Bradford.
