Meta Expands AI Safety Tools After Acting on 33.2 Million Child Exploitation Content Items

TL;DR

• Meta says it took action against 33.2 million pieces of child sexual exploitation content in the first half of 2026.

• More than 97% of that content was detected by its systems before users reported it.

• A new language-model system targets seemingly harmless ads that direct users to illegal material elsewhere online.

• Meta is adding AI scans, automated safety testing, and stronger detection of users returning after account removal.

Meta has announced new AI tools to combat child exploitation, including systems designed to identify advertisements and accounts that direct users to illegal material outside its platforms.

The company said Wednesday that it took action against 33.2 million pieces of child sexual exploitation content on Facebook and Instagram during the first half of 2026.

More than 97% was identified by Meta’s systems before users submitted reports, according to the company. In India, Meta acted on 5.3 million pieces of content, with more than 98% detected proactively.

New AI System Examines Where Advertisements Lead

Meta has introduced a large language model system to detect what it calls “signposting”: apparently ordinary advertisements suspected of directing people to illegal content or harmful activity elsewhere online.

The company said it has recently observed this tactic as offenders adjust their methods to evade detection.

An advertisement may contain no illegal material itself while still directing users to a website hosting child sexual abuse material. Meta’s approach therefore examines the destination alongside the advertisement’s content.

The company said those findings can inform decisions to block destinations that violate its policies and take action against associated accounts.

Additional Scans and AI Testing Target Detection Gaps

Meta is also introducing further AI-driven scans to identify exploitation content that previous systems may have missed.

The company said it will incorporate additional signals as it learns more about how abusive networks operate.

A separate red-teaming AI agent will test Meta’s protections for weaknesses. The aim is to identify potential gaps and emerging abuse methods before they become more widespread.

Meta is also improving its ability to detect people who create new accounts after being removed from its services.

Together, the measures seek to address harmful content, links to external material, and attempts to return after enforcement action.

Child Safety Scrutiny Continues Despite New Measures

The announcements come as Meta faces ongoing criticism from lawmakers and lawsuits concerning risks to children using its services.

In August, the company reached an agreement to pay up to $18 billion to settle a child safety lawsuit involving 29 U.S. states, according to the supplied report.

Meta’s enforcement figures describe content its systems identified and acted upon. They do not establish how much harmful material remained undetected or measure the overall prevalence of exploitation across its platforms.

The new tools will therefore be judged on their ability to improve detection and enforcement alongside the protections already in place.

Parental Controls Expand Across Meta Services

Meta has introduced several child-safety features this year, including parental controls for Meta AI and parent-linked WhatsApp accounts for preteens.

Instagram also added alerts for parents when teenagers search for suicide or self-harm content.

In September, WhatsApp expanded parental controls for teen accounts. Parents can restrict Channel use, manage who sees status updates, and determine who can add their children to groups.

They can also opt to receive notifications about certain group activity.

The latest AI rollout adds another layer to those measures, focusing on exploitation networks and the accounts and advertisements used to direct people toward illegal content.