Meta adds new AI tools to detect child exploitation in ads

Meta has introduced additional artificial intelligence (AI) tools to its ad review process as it looks to identify advertising that may be used to direct people towards child sexual exploitation material hosted outside its platforms.

The changes come as the company observes bad actors shifting their tactics to evade detection, including through ads that may appear harmless on their own but contain signals designed to covertly direct users towards illegal material elsewhere.

After identifying the tactic, Meta expanded its investigation to look for related activity across its platforms, disabling accounts found to be violating its policies and blocking links associated with the activity.

Don't miss: India orders Meta to take down child sexual abuse ads 

Among the new measures is large language model (LLM) detection for what Meta describes as "signposting". This refers to seemingly benign advertising content suspected of directing users towards child sexual exploitation material or other harmful activity outside its platforms.

Its systems have also been updated to consider where an ad ultimately directs users, alongside the content appearing within the ad itself. The move is intended to help identify violating destinations and take action against accounts associated with them.

Meta has also rolled out additional AI-driven sweeps of advertising content to surface material that previous systems may not have detected. A red-teaming AI agent has been introduced to probe its detection systems for weaknesses and identify emerging tactics, while its systems for detecting previously removed users attempting to create new accounts have also been strengthened.

Beyond advertising, Meta uses a mix of behavioural signals, AI models and photo and video-matching technologies, including PhotoDNA, to identify suspected CSAM, while sharing newly identified image and video hashes with participating companies through the Tech Coalition's "Lantern programme".

When links to external websites associated with violating CSAM content are blocked, Meta searches for other ads, posts or comments containing the same links. It also takes steps to prevent blocked links from being posted across Facebook, Instagram and Threads, while ads containing them are rejected at upload.

Between January and June 2026, Meta actioned 33.2 million pieces of child sexual exploitation content across Facebook and Instagram globally. According to the company, more than 97% was identified and addressed before being reported by users.

In India, 5.3 million pieces of child sexual exploitation content were actioned over the same period, with Meta reporting that more than 98% was identified before receiving a user report.

The company also conducts investigations into networks of accounts suspected of predatory activity, with its team including former FBI investigators. Potential child sexual abuse or exploitative content detected or brought to its attention is investigated, with violating content removed and accounts potentially disabled where malicious sharing is identified.

The measures come amid recent scrutiny of how child sexual abuse material can make its way through Meta's advertising systems. An investigation by the Tech Transparency Project (TTP) found that Meta ran 332 paid ads containing CSAM on Facebook and Instagram this year, with most allegedly using AI-manipulated images of children to depict sexual activity. TTP also identified images of real children taken from social media posts and websites that were subsequently used in the ads.

Responding to the findings at the time, a Meta spokesperson told MARKETING-INTERACTIVE that the company does not tolerate child exploitation, whether involving real or AI-generated material. The spokesperson added that many of the ads flagged had already been removed, most received fewer than 200 impressions and total ad spend was below US$5,000, while pointing to the changing tactics used by those attempting to evade its detection systems.

Related articles:   
Why Indonesia's child-safety push puts eCommerce's growth model under scrutiny  
Meta gives parents visibility into teens’ AI use as scrutiny grows  
X, TikTok under watch as SG joins global push to protect young social media users