Meta Deploys New AI Tools to Combat Child Exploitation
Meta introduces advanced AI tools and LLM-based detection to stop child sexual exploitation on Facebook and Instagram, including proactive measures in India.
In a major push to protect minors online, Meta has rolled out advanced artificial intelligence tools and upgraded its advertising review systems to disrupt child sexual exploitation on Facebook and Instagram. This move comes as the company faces increasingly sophisticated tactics used by online predators to bypass traditional safety filters.
Tackling Sophisticated Predator Tactics
Meta revealed that predators are now frequently using advertising as a shield to evade detection. These bad actors often create advertisements that appear completely harmless on the surface, but are actually designed to "signpost" or covertly direct users to illegal and harmful material hosted on external websites outside of Meta's platforms.
To counter this, the company has introduced Large Language Model (LLM)-based detection. This technology allows Meta to identify subtle patterns in advertisements that suggest they are directing users toward violating activity. Furthermore, Meta has enhanced its ability to scan the actual destinations of advertisements, rather than just looking at the visible content, allowing the company to block illegal websites and take down the accounts responsible for linking to them.
Massive Proactive Detection in India and Globally
The scale of the challenge is immense, but Meta's proactive systems are showing significant results. According to data released by the company, between January and June 2026, the following actions were taken:
- In India: Meta actioned 5.3 million pieces of child sexual exploitation content on Facebook and Instagram. Notably, over 98 per cent of this content was identified and addressed proactively before any user reported it.
- Globally: Meta actioned 33.2 million pieces of such content, with more than 97 per cent being detected proactively through its automated systems.
New AI Defensive Layers
Beyond LLMs, Meta is deploying several other cutting-edge layers of defense to ensure predators cannot find gaps in the system:
- AI-Driven Ad Sweeps: New sweeps are being conducted to identify exploitation content that might have been missed by older, legacy systems.
- Red-Teaming AI Agent: Meta has deployed a specialized AI agent designed to act as an attacker, testing the company's own defenses to find and fix vulnerabilities before predators can exploit them.
- Repeat Offender Detection: Strengthened systems are now in place to identify and block individuals who attempt to create new accounts immediately after being banned.
- Multi-Signal Safety: The company continues to use behavioral signals, PhotoDNA, image and video-matching technologies, and automated classifiers to spot emerging patterns of abuse.
Collaboration with Indian Authorities
Recognizing the gravity of these crimes, Meta has committed to deeper cooperation with Indian law enforcement. In India, the company stated it will report child safety cases directly to the National Cyber Crime Reporting Portal, which is operated by the Indian Cybercrime Coordination Centre (I4C).
This initiative builds upon Meta's existing protocols for reporting child sexual abuse material (CSAM) to law enforcement, supported by the National Center for Missing & Exploited Children.