Meta details new measures to counter child sexual exploitation across Facebook and Instagram, including ad-focused AI tools

Meta says it has taken additional steps to strengthen how it fights child sexual exploitation on its platforms, emphasizing that predators increasingly adapt tactics to evade detection and can operate across multiple services.

In its October 7 update, the company described a sustained approach that combines automated systems with manual investigations. Meta says it investigates reports or detections of potential child sexual abuse or exploitative content, removes material that violates its policies, and in cases involving indicates of malicious sharing disables the accounts behind it. Meta also said it reports this content to the National Center for Missing & Exploited Children (NCMEC), following applicable law.

Meta also said it has conducted strategic network disruptions for years, using a team of experts that includes former FBI investigators who conduct manual investigations aimed at proactively uncovering and removing networks of predatory accounts. The company said it continues to study how these networks operate to act at a network level rather than only removing individual accounts.

A key focus of the update is the role advertising can play in adversarial schemes. Meta said online predators have shifted tactics, including exploiting ad surfaces, where content may appear harmless but is designed to direct people toward illegal material hosted off Meta platforms in ways intended to avoid detection.

Meta said once it understood the design of a malicious advertising pattern, it widened investigations to identify related activity across its platforms. It reported disabling violating accounts behind the activity and blocking the off-platform links involved.

To address these adversarial tactics, Meta said it deployed additional measures including new AI tools. The company said the tools are designed to improve the efficacy of systems that identify (1) ads that feature child sexual exploitation material and (2) ads that direct people to violating activity off-platform.

As part of its operational metrics, Meta provided figures for the first half of 2026. It said that globally between January and June 2026 it actioned 33.2 million pieces of child sexual exploitation content from Facebook and Instagram, with more than 97% found and proactively addressed before anyone reported it. For India over the same period, Meta said it actioned 5.3 million pieces of child sexual exploitation content on Facebook and Instagram, with more than 98% found and proactively addressed before anyone reported it.

Meta also tied its efforts to law-enforcement coordination and cross-border reporting. It said that in September it announced Meta will report child safety cases directly to the National Cyber Crime Reporting Portal run by the Indian Cybercrime Coordination Centre (I4C). Meta said this builds on existing reporting where it reports child sexual abuse material to law enforcement through the national portal with assistance of NCMEC.

What Meta did not specify in detail were the exact technical models behind the “new AI tools,” the thresholds used to determine when an ad is treated as directing users to violating off-platform material, or how often manual investigators override or augment automated determinations. The company also did not break out how many accounts were disabled specifically due to the advertising-focused scheme, or what portion of the 33.2 million and 5.3 million actions relate to ads versus other surfaces.

Going forward, Meta’s next disclosure points to watch are whether it provides further breakdowns by detection surface (for example, ads versus messaging or other formats), additional public metrics on proactive detection rates as the adversarial tactics evolve, and more clarity on how its network disruption work scales across new platform behaviors. Regulators and civil society groups will likely also look closely at transparency about categories of enforcement and the reported outcomes of AI-assisted investigations. (Meta did not describe enforcement outcomes beyond account disabling and removal in its update.)