Meta Adds AI Detection for Harmful Ads on Its Platforms

Meta says Facebook and Instagram removed or restricted 33.2 million items involving child sexual exploitation during the first half of 2026. Its automated systems identified more than 97% before users submitted reports. In India, the company acted on 5.3 million items, with its systems finding more than 98% first.

The company is now deploying an AI system designed to recognize “signposting.” In this tactic, an apparently ordinary advertisement serves as a path to illegal abuse material or other harmful activity hosted beyond Meta’s services. The system evaluates an ad’s destination as well as its visible contents. Meta can then restrict destinations that violate its policies and act against the associated accounts.

Other AI scans will search for exploitation material missed by earlier detection systems. A red-team AI agent will probe Meta’s safeguards for weaknesses that abusers might exploit, while improved systems will seek people who create replacement accounts after removal. Meta plans to incorporate additional signals as it learns how these networks work.

The changes may affect what Simi Valley parents and other local users encounter across Meta’s services. They arrive amid lawsuits and scrutiny from lawmakers concerning risks to younger users. In August, Meta agreed to pay as much as $18 billion to resolve litigation brought by 29 U.S. states over online protections for children. This year’s other measures include Meta AI parental controls, WhatsApp accounts for preteens, and Instagram alerts to parents when children search for self-harm material.