Meta is adding AI tools to detect accounts and advertisements that may guide users to illegal child abuse material outside Facebook and Instagram. The company says the systems will look beyond an ad’s content to consider its destination and will help identify tactics that may evade existing protections.
Looking beyond what an ad shows
Meta said 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% of the content was found by its systems before users reported it.
In India, the company acted on 5.3 million pieces of such content over the same period. Meta said more than 98% was detected before users reported it.
A new large language model system is intended to find what Meta calls “signposting.” The term describes ads that may appear ordinary but are suspected of directing people to illegal content or other harmful activity elsewhere online.
According to Meta, people using this tactic adapt their methods to avoid detection. An ad may not itself contain illegal material, but could still serve as a route to a website hosting child sexual abuse material.
Meta says it will assess where an ad sends people as well as what appears in the ad. That can help the company block destinations that violate its rules and take action against the accounts associated with them.
More scans and tests for weaknesses
The company is also adding AI-driven scans to find child exploitation content that earlier systems may have missed. Meta says it plans to add signals as it learns more about how these networks operate.
A separate “red-teaming AI agent” will test Meta’s own safety measures. It is designed to look for weaknesses that bad actors could exploit to get around protections. Meta says these tests may help it identify new forms of abuse before they become more common.
Another area of work is finding people who return to Meta platforms with new accounts after their previous accounts have been removed. The new measures therefore address several points in the process: content detection, routes that lead away from the platforms, system weaknesses, and repeat account creation.
Child safety remains under scrutiny
The tools arrive as Meta faces continuing pressure over the safety of children on its platforms. The company has faced lawsuits and criticism from lawmakers about the risks its social media services may pose to young users.
In August, Meta reached an agreement to pay up to $18 billion to settle a child safety lawsuit involving 29 U.S. states. The agreement addressed concerns about protecting children online.
Meta has also introduced other child-safety features this year. These include parental controls for Meta AI, pre-teen accounts on WhatsApp, and alerts for parents when children search Instagram for self-harm content.
In September, WhatsApp added more parental controls. Parents can limit how teenagers use Channels, manage who can see their status updates, choose who can add their children to groups, and opt to receive notifications about certain group activity.
What the new approach is meant to change
The new ad system reflects a broader challenge described by Meta: harmful activity may be connected to an ad even when the ad itself looks harmless. Checking the destination gives the company another point at which it can intervene.
The scans and red-team agent serve different roles. One searches for material that existing systems may have overlooked; the other probes Meta’s protections for weaknesses. Together with efforts to identify returning users, the tools expand how the company says it will look for abuse and evasion.
Meta has described the intended capabilities of these systems, while the reported figures show the scale of content it acted on in the first half of 2026. The company says it will keep adding detection signals as it learns how these networks change their methods.