Meta has announced additional measures to combat child sexual exploitation on its platforms, including new AI tools to improve the detection of ads that feature child sexual abuse material (CSAM) or direct people to violating activity off-platform. The company said the measures address tactics used by predators to evade detection across its platforms.
Between January and June 2026, Meta actioned 33.2 million pieces of child sexual exploitation content on Facebook and Instagram globally, with over 97% found and proactively addressed before anyone reported it. In India, Meta actioned 5.3 million pieces during the same period, with over 98% found and proactively addressed before anyone reported it.
Meta investigates potential child sexual abuse or exploitative content whenever it detects the material or becomes aware of it. If the content violates its policies, the company removes it and, when signals indicate malicious sharing by predators, disables the accounts behind it. Meta also reports the content to the National Center for Missing & Exploited Children (NCMEC), in line with applicable law.
The company also conducts strategic network disruptions related to child safety. Its team of experts, including former FBI investigators, conducts manual investigations to proactively uncover, remove and report networks of predatory accounts. Meta investigates how these networks operate so it can act at the network level rather than only against individual accounts.
New improvements to ad review
Meta said advertising is one of the surfaces where online predators have shifted their tactics. Bad actors may use content specifically designed to evade detection, including ads that appear harmless while covertly directing people to illegal material hosted off Meta’s platforms.
After identifying this malicious advertising scheme, Meta expanded its investigation beyond the activity reported to it and searched for related activity across its platforms. It disabled the violating accounts and blocked the off-platform links involved. The company then added new AI-based measures to improve detection of violating advertising.
The new ad-review measures include:
- LLM detection for “signposting” of CSE material: Identifies seemingly benign ad content strongly suspected of covertly directing people to illegal content or harmful activity off Meta’s platforms.
- Ad destination analysis: Examines where an ad leads, in addition to what it shows, to block violating destinations and enforce against the responsible accounts.
- AI-driven ad sweeps: Surfaces CSE material that earlier systems did not catch using signals that Meta continues to expand as its investigation progresses.
- Red-teaming AI agent: Probes Meta’s defences for weaknesses to identify new adversarial CSE tactics before they scale.
- Recidivism detection: Identifies bad actors who attempt to create new accounts after removal.
Technology behind the detection systems
Meta uses a combination of AI, automated systems and matching technologies to detect CSAM and related child safety violations. These systems cover behavioural activity, known CSAM and emerging violation patterns.
The detection systems include:
- Behavioural signals: Analyse combinations of metadata and behavioural signals to detect accounts engaged in suspicious patterns.
- PhotoDNA and other matching technology: Used across all Meta apps since 2011 to detect identical or near-identical copies of known CSAM and remove them.
- Image and video hash database: Meta maintains one of the industry’s largest databases of image and video hashes, contributes newly identified hashes and shares them with industry through the Tech Coalition’s Lantern program so participating companies can also remove the content.
- Off-platform link-preview detection: Uses the same photo and video matching technology to identify CSAM in off-platform link previews. Meta reports all CSAM identified through these matching technologies to Law Enforcement Authorities with NCMEC assistance, in compliance with applicable law.
- External link blocking: Blocks links to external websites that host or create violating CSAM content. After blocking a link, Meta searches for and deletes ads, posts and comments containing it.
- Posting and advertising controls: Takes steps to prevent people from posting content containing blocked links on Facebook, Instagram and Threads, while ads containing blocked links are rejected at upload.
- Heuristics-based rules: Combine automated classifiers, AI models and other signals to proactively identify and act on new or emerging violation patterns at scale, including in ads.
Working with law enforcement
Meta said it continues to work with authorities on child safety cases. In September, the company announced that it would report child safety cases directly to the National Cyber Crime Reporting Portal (cybercrime.gov.in), run by the Indian Cybercrime Coordination Centre (I4C).
This builds on Meta’s existing reporting process, under which it reports child sexual abuse material to Law Enforcement through the National Cyber Crime Reporting Portal with the assistance of NCMEC.