Meta disclosed on Wednesday that it addressed 33.2 million instances of child sexual exploitation material across Facebook and Instagram during the first half of 2026. Concurrently, the company is deploying advanced AI tools to identify advertisements and user accounts potentially facilitating the covert redirection of individuals to illicit child abuse content.
According to Meta, over 97% of the content it addressed was detected by its internal systems prior to any user reports. Specifically in India, Meta took action on 5.3 million pieces of child sexual exploitation content within the same timeframe, with more than 98% of these being identified proactively before user flagging.
Meta stated that its newest AI technologies are designed to target individuals who employ seemingly innocuous advertisements to channel users towards illegal material external to Meta’s platforms.
A new large language model (LLM) system has been launched by the company to identify “signposting,” a term it uses for advertisements that appear legitimate but are suspected of guiding users to illicit content or other detrimental activities on other online venues. Meta notes that this tactic has recently been adopted by malicious actors who continuously evolve their methods to evade detection. Even if these ads do not directly feature illegal material, they can serve as entry points to websites containing child sexual abuse material.
Consequently, Meta indicated that its focus now extends beyond the ad’s content to also include its destination. This approach enables the company to block websites or other targets that violate its policies and to penalize the associated accounts.
Furthermore, Meta is employing extra AI-powered scans to uncover child exploitation content that might have escaped previous detection systems. The company affirmed its commitment to integrating new indicators as it gains further insight into the operational methods of these networks.
An additional new utility is a “red-teaming AI agent” designed to evaluate Meta’s internal safety protocols. This agent identifies vulnerabilities that malicious actors might exploit to bypass the company’s safeguards. Meta believes this could aid in discovering novel abuse tactics before they become widespread.
In addition, the company is enhancing its systems aimed at identifying individuals who rejoin its platforms with fresh accounts subsequent to the removal of their prior ones.
These latest initiatives are being implemented while Meta remains under scrutiny concerning child safety across its platforms. The company has encountered legal challenges and legislative critique regarding the potential hazards its social media offerings present to young individuals. Last August, Meta agreed to a settlement of up to $18 billion in a child safety lawsuit initiated by 29 U.S. states, intending to resolve issues related to online child protection.
Throughout this year, Meta has rolled out multiple child-safety functionalities, such as parental controls for Meta AI, the introduction of preteen accounts on WhatsApp, and alerts for guardians when adolescents search for self-harm content on Instagram.
By September, WhatsApp incorporated additional parental controls, allowing guardians to regulate teenagers’ Channel usage, manage visibility of status updates, and determine who can invite their children into groups. The option to receive notifications regarding specific group activities was also provided.