A child in Aizawl had never approached the police, nor had the family. Yet digital signals generated by technology platforms eventually became part of a CBI investigation that helped identify and rescue the child. The case illustrates an international detection infrastructure through which material appearing online can be recognised, reported and routed to Indian law-enforcement agencies.
Internationally, such material is commonly called CSAM, or child sexual abuse material. In India, the Supreme Court has endorsed the term CSEAM, child sexual exploitation and abuse material, rejecting “child pornography” as terminology that can trivialise the seriousness of abuse. India has formally received NCMEC CyberTipline reports since an NCRB-NCMEC agreement in 2019. In 2025, NCMEC associated about 1.93 million CyberTipline reports with India. These are reports, not 1.93 million victims, offenders or criminal cases.
Technology companies broadly detect such material in three ways: recognising previously identified content, flagging potentially new material, and analysing suspicious account or network behaviour. Known material can be detected using hashes, effectively mathematical fingerprints. Microsoft’s PhotoDNA, for example, allows participating services to match images against fingerprints of previously identified material without identifying the person depicted or reconstructing the original image.
Previously unseen material requires a different approach. Google says artificial intelligence can flag images displaying characteristics similar to confirmed CSAM, after which trained personnel review them. Meta also uses AI and machine learning alongside matching technologies and behavioural signals. Human review remains important because platform enforcement and criminal proof are not the same. Microsoft reported that 15.08% of accounts it actioned in 2025 were subsequently reinstated following appeals and further review.
The scale of repeat circulation further complicates the numbers. Meta found that more than 90% of child exploitative material in one 2020 sample was identical or visually similar to previously reported material. NCMEC says imagery involving one child victim circulated for around two decades and appeared more than 1.4 million times in submissions. A single victim’s material can therefore generate enormous numbers of detections.
Ultimately, technology can identify a signal, but it cannot establish who created or uploaded material, who controlled an account, where the abuse occurred or whether a child remains at risk. Those questions require investigation, device examination, digital forensics and evidence capable of standing in court.
Read the full article on Chapter One Magazine.
