India’s Kerala Police Launch AI Platform to Combat Online Child Abuse on the Dark Web

Kerala Police’s Counter Child Sexual Exploitation team stumbled onto an encrypted dark web forum in 2025 called “CHEEZE PIZZA,” where criminals traded links to child sexual abuse material and funneled victims into private Telegram groups for purchase and access. A months-long digital manhunt followed next, ending with a child saved and a busted perpetrator. 

Investigators spotted a “Keralite” child while combing through the forum. The user had posted multiple photos and videos of the same child. Officers went deeper. They built an online rapport with the suspect, gained access to Facebook and Instagram details, and traced a trail that led to a woman in Bengaluru whose account pointed to Thiruvananthapuram.

A Facebook connection from the same city cracked the case open further. Officers then pulled up the suspect’s Telegram profile and compared photographs he had posted publicly. The house visible in one of those profile photos matched the address they had already identified as his location in Thiruvananthapuram. Police arrived at the door and found the child. She was his niece. Officers rescued and counselled her, and the offender faced arrest and charges.

The investigation demanded months of persistence and cross-platform tracking. But without AI filtering through terabytes of data spread across multiple platforms, the breakthrough would have taken considerably longer. Manual analysis alone could not have moved fast enough.

India Puts AI at the Center of Its First Child Exploitation Unit

Kerala Police made history by becoming the first law enforcement agency in India to integrate artificial intelligence into child sexual exploitation investigations at its Counter Child Sexual Exploitation Centre (CCSEC). The platform fueling this shift is Katalyst, a masterpiece of New Zealand-based Kindred Tech. An upgrade is already running to allow this platform to run on MongoDB’s infrastructure for advanced data storage and processing.

Ankit Asokan, Superintendent of Police for the Cyber Crime Division, explained the core challenge that made AI essential. “When you are a small team, you simply cannot go through everything by hand,” he said. “Technology has made this worse, and we need to use technology to address this problem.” He also flagged a second barrier unique to these cases: “No victims will come to us ever.” That reality makes proactive, AI-driven detection not just useful but necessary.

Kindred Tech CEO Bree Atkinson said Katalyst exists to help detectives surface key information faster and prioritize each child’s case during investigations. The platform cuts down the time officers spend on data searches, freeing them to focus more energy directly on child protection work.

Over an 18-month pilot beginning in 2024, the partnership between Kindred Tech and Kerala Police produced 96 arrests, 20 rescued children, and 18 international referrals, according to data from both Kindred Tech and MongoDB. The results validated what the CCSEC team had argued from the start: that fighting a technology-driven crime requires a technology-powered response.

Similar outcomes are being achieved in the United States, where courts have handed down significant sentences in two child exploitation cases, demonstrating that successful investigations lead to real consequences for offenders.

The Kerala Police did not outsource this work to external vendors. Officers built and operated the tools themselves, in collaboration with New Zealand technologists through the Cyber Dome initiative, including hackathons where both teams worked side by side to develop solutions.

“No outside vendors contracted the investigations,” Asokan clarified. “The members of the police have technology know-how.” This initiative also earned the Technology Sabha Excellence Award at the state level, recognising the unit’s innovative approach to digital law enforcement.

Kerala Police’s broader operational record underscores the scale of the work. In February 2022, Operation P-Hunt resulted in the arrest of four suspects facing charges related to the exploitation of a girl. Officers seized 51 digital devices in the Ernakulam area and conducted raids at 82 locations across the state, including 36 in urban areas and 46 in rural areas, according to The Hindu newspaper. The operation produced substantial evidence supporting charges against all four suspects.

The Scale of the Crisis Driving the Push

Child Sexual Abuse Material (CSAM) covers any image, video, or online content depicting the abuse or exploitation of children. Asokan broke down how the threat has evolved beyond its traditional definition. “Previously, we referred to it as online child sexual abuse. Now it encompasses a broader range of cases.

In addition to involving both a child and technology, there are various methods of producing the content that can be used for abuse. Before now, coercion or extortion from children and those from individuals with authority over a child remained the two basic types. A third category is emerging; using AI to generate or digitally manipulate materials with identifiable traits of real children.”

Distribution moves at an alarming speed once creators produce the material. “Once there’s creation, the starting point of an ecosystem forms,” Asokan stated, describing how content gets transmitted, sold, viewed, and shared across platforms at an exponential rate. The moment material enters circulation, stopping its spread becomes nearly impossible through conventional means.

The numbers reflect the true scale of the crisis. The National Center for Missing and Exploited Children (NCMEC) received 2.2 million cyber tips related to online child sexual abuse in 2024 alone. The Internet Watch Foundation recorded 291,273 reports of CSAM in the EU during the same year.

ECPAT International, in a submission to the United Nations Office on Drugs and Crime, reported that NCMEC received 29.3 million reports in 2021, covering 39 million child sexual exploitation images and 44.8 million videos. Of those, only about 4,260 potential new victims were flagged to law enforcement specifically for children who may have faced online sexual exploitation.

Asokan put the broader trajectory plainly: “It’s going to be difficult to get accurate statistics, but the reality is that the increase in children using the internet, the growth of social media platforms, and the meteoric rise of synthetic media (such as computer-generated imagery) have all led to the proliferation of this type of crime worldwide.”

He also highlighted a structural problem that separates CSAM cases from most other crimes. In conventional offenses, victims typically seek justice through law enforcement or other channels. In online child sexual exploitation, that almost never happens. Victims cannot identify themselves, cannot report the abuse, and in many cases do not even know the material exists. That absence of a reporting victim is exactly why reactive policing fails here. Officers must go looking.

How Katalyst Turns Terabytes Into Leads

Manual evidence review in CSAM cases is punishing work. A single investigation can involve 200 to 250 GB of data or more, spanning seized devices, cloud storage accounts, and other digital evidence sources. Cyber-tip reports alone run 35 pages, packed with metadata, detailed chat logs, platform IDs, and file hashes. Reviewing all of that by hand makes investigations slow and places an enormous burden on already stretched teams.

Katalyst automates the ingestion and prioritization of all of it. The platform offers advanced case management tools, a secure media library, AI-driven categorization of sensitive material, and automated forensic support. “Katalyst stores AI findings alongside the raw data and helps investigators search for similar patterns to spot repeated abuse,” according to Peter Pilley of Auckland-based Pathfinder Labs.

One of its most powerful features is object tracing. A single abuse series might span months or even years, filmed repeatedly in the same location. AI platforms such as Katalyst can take apart a recurring visual detail from within an image (a wall mural, a paint color, a lamp, a curtain pattern) and use it as an investigative anchor across hundreds of files.

Asokan described how this works in practice: “Suppose a child is in someone’s guardianship and is being captively abused. There might be 100 videos of the entire room taken over the course of one year. If an investigator observes the same curtain in several videos, he can begin to trace it. If you find where the curtain was made or its design’s origin, even a block print, police can analyze the pattern for clues. We can triage it.”

In the Thiruvananthapuram case specifically, AI analysis of social media metadata and the comparison of digital images across platforms helped investigators pin down the perpetrator’s location and confirm his identity as a suspect. That kind of cross-platform correlation, done manually, would have taken investigators far longer to complete.

Katalyst does not replace investigative judgment. It accelerates it. Officers still make the calls, build the cases, and knock on the doors. But the AI gets them to the right door faster, and in cases involving child exploitation, speed is everything.

Asokan closed with the philosophy driving the entire initiative: “Technology created this problem, so we have to use technology to fight it. Victims won’t come forward, so we must take action ourselves.”