Meta Actioned 33.2M Child Exploitation Items in Six Months

Meta actioned 33.2 million pieces of child sexual exploitation content on Facebook and Instagram globally between January and June 2026, with over 97% detected proactively before any user report, according to figures the company published on its newsroom this week. In India alone, the company actioned 5.3 million pieces in the same period, with over 98% found before anyone reported them.

Alongside the numbers, Meta detailed a package of new ad-review measures. The company says predators have shifted toward "signposting" — ads that look harmless on their own but covertly direct people to illegal material hosted off Meta's platforms.

What did Meta change in ad review?

After uncovering the malicious ad scheme, Meta says it widened its investigation beyond reported ads, disabled the accounts behind them, and blocked the off-platform links involved. It then deployed new tools, including:

• A new Large Language Model (LLM) detector for CSE "signposting" — seemingly benign ad content strongly suspected of covertly directing users to illegal material off-platform.
• Improved analysis of where an ad leads, not just what it shows, so violating destinations can be blocked and the responsible accounts enforced against.
• Additional AI-driven sweeps of ad content to surface CSE material earlier systems missed.
• A red-teaming AI agent that probes Meta's own defenses to find weaknesses before adversarial tactics scale.
• Strengthened recidivism detection to catch bad actors who create new accounts after removal.

What technology sits behind the enforcement?

Meta's detection stack combines behavioural signal analysis of account metadata, PhotoDNA and other photo/video-matching technology used across its apps since 2011, and heuristics-based rules that combine automated classifiers and AI models to catch emerging violation patterns. The company says it maintains one of the industry's largest hash databases, contributes newly identified hashes, and shares them through the Tech Coalition's Lantern program so participating companies can remove the same content.

Blocked external links trigger a wider cleanup: Meta says it deletes other ads, posts and comments containing the link, prevents users from posting it on Facebook, Instagram and Threads, and rejects ads containing blocked links at upload. All CSAM identified via matching technologies is reported to law enforcement through NCMEC.

Who investigates the networks?

The enforcement goes beyond single accounts. Meta says a team of experts — including former FBI investigators — conducts manual investigations to proactively uncover, remove and report networks of predatory accounts, acting at a network level rather than account by account.

In India, Meta announced in September that it will report child safety cases directly to the National Cyber Crime Reporting Portal (cybercrime.gov.in) run by the Indian Cybercrime Coordination Centre (I4C), building on existing NCMEC-assisted reporting to law enforcement.

What does this mean for practitioners?

The enforcement tightening matters for legitimate advertisers too. Meta's improved destination analysis and upload-time link rejection mean ads pointing to pages whose content shifts after review face faster takedowns, and recidivism detection raises the cost of ban evasion. Advertisers running cloaked or redirect-heavy funnels should expect stricter automated scrutiny — the same systems built to catch signposting will flag destination mismatches broadly.

Meta has not disclosed detection error rates for the new LLM tools, so false-positive impact on legitimate ad delivery remains unmeasured. The company says its investigation into the ad scheme is ongoing and that it will continue expanding the signals feeding the AI sweeps.