Meta has introduced new artificial intelligence tools designed to detect and disrupt advertisements that may be used to facilitate child exploitation on its platforms. The company said the systems are intended to identify activity that can be difficult to detect when advertisements appear legitimate but direct users toward harmful or illegal material elsewhere online.
AI Targets Hidden Signals
One of the new systems uses a large language model to identify what Meta describes as signposting. This involves seemingly ordinary advertising content that contains signals potentially directing users toward illegal activity or prohibited material outside Meta’s platforms.
Meta is also examining where advertisements direct users, rather than evaluating only the content displayed in the advertisement. The company can use information about destinations and associated accounts to identify violating activity and take enforcement action.
Additional AI Detection Systems
Meta said it has expanded automated sweeps of advertising content to identify material that earlier detection systems may have missed. The company is also using an AI red-teaming agent to test its safety systems and identify potential weaknesses that could be exploited to avoid detection.
Meta has also strengthened measures designed to identify people who attempt to return to its services through new accounts after previous accounts have been removed for violating its policies.
Millions of Pieces of Content Actioned
Meta reported that it actioned 33.2 million pieces of child sexual exploitation content on Facebook and Instagram globally between January and June 2026. The company said more than 97% of this content was detected and addressed before users reported it.
In India, Meta said it actioned 5.3 million pieces of child sexual exploitation content during the same period. More than 98% was detected and addressed proactively, according to the company.
How AI Supports Detection
Meta uses automated classifiers, behavioural signals, rules-based systems, AI models and image and video matching technologies as part of its broader enforcement systems. The company has also used technologies such as PhotoDNA to identify known child sexual exploitation material and prevent its distribution.
The platform can additionally block links associated with violating material and investigate other advertisements, posts and comments that contain the same destinations or related signals.
Focus on Evolving Tactics
Meta said people involved in child exploitation have increasingly attempted to use advertisements and other indirect methods to avoid detection. An advertisement may not contain illegal material itself but could be designed to direct users toward prohibited content hosted elsewhere.
By examining advertising content, external destinations, account behaviour and connections between different pieces of activity, Meta aims to identify broader patterns rather than treating each violating item as an isolated case.
Importance of Online Child Safety
The new measures highlight the growing role of AI in online safety enforcement. Platforms are increasingly using AI not only to identify prohibited material but also to analyse behaviour and detect attempts to circumvent existing safeguards.
Meta said it will continue updating its detection systems as the methods used to distribute child exploitation material evolve. The company is also combining automated detection with enforcement measures and investigations to strengthen its child-safety efforts.
Latest Status
Meta’s new AI systems form part of its broader efforts to prevent child exploitation across its platforms. The company says it will continue expanding detection signals, testing its safety defences and taking action against accounts, advertisements and external destinations connected to prohibited activity.