Hyderabad: In a first-of-its-kind initiative in the country, Telangana police have deployed a sovereign artificial intelligence platform to identify child sexual abuse offenders, analyse massive volumes of digital evidence and accelerate investigations into crimes against children.Telangana Cyber Security Bureau (TGCSB) will analyse lakhs of CyberTipline reports linked to child sexual exploitation and abuse material (CSEAM), dramatically accelerating investigations and helping police identify suspects, register FIRs and rescue children.The bureau’s Child Protection Unit (CPU) has developed C-SIGHT, which it describes as India’s first sovereign AI platform designed specifically to process large volumes of child abuse tipline data while ensuring that sensitive information never leaves police systems.Developed by TGCSB in partnership with technology firm Vatins Systems, the platform automates the examination of images and videos, estimates the probable age of victims, identifies the nature of abuse and categorises digital material based on content and risk indicators.TGCSB director general Shikha Goel said the platform’s defining feature was its sovereign AI architecture, under which all analysis is conducted offline within the police network. “Not a single byte of sensitive investigation information is sent to external or regular AI tools,” she told TOI.The CPU, established on Feb 18, 2025, in collaboration with Indian Child Protection, New Delhi, uses C-SIGHT to manage CyberTipline reports and supervise investigations while retaining complete control over sensitive police and citizen data. CyberTipline is a centralised reporting mechanism managed by the National Center for Missing and Exploited Children (NCMEC) which collects data on online sexual exploitation and abuse of children.The bureau said its work has also resulted in the rescue of three children based on online tipline evidence, underscoring how the combination of AI-assisted analysis and investigator oversight is transforming the fight against online child exploitation. A 13-year-old girl studying in class VIII was rescued on Sept 24, 2024. She was from a poor economic background. A four-year-old girl from a middle class family was rescued on Jan 8 this year. In the third case, a five-year-old girl from a poor economic background was rescued on March 30.The AI-enabled, self-hosted platform automatically extracts information from Tipline PDF files and converts it into structured datasets, significantly reducing the time required to organise and examine reports. It analyses images and videos and segregates them into categories depending on whether they contain child sexual exploitation material.According to TGCSB, each tipline can now be processed in 20 minutes or less. The system has achieved 98% accuracy in identifying CSEAM content and 100% accuracy in identifying suspect information, while maintaining 98% data completeness.The platform also provides real-time dashboards showing the number of tiplines received, processed, pending or forwarded to other police units and states. Access to sensitive material is controlled through role-based permissions.Goel said the offline architecture was intended to protect both investigations and victims. “Sensitive investigation details and citizen data remain within the organisation’s network. Data sovereignty and privacy cannot be compromised while handling material involving children,” she said.The AI system helps investigators prioritise cases by examining indicators such as the probable age of a child, the nature of interaction, whether it appears playful or forceful, inappropriate gestures, attire and the surroundings visible in images or videos. Cases involving children aged up to five years receive particular attention.The impact has been significant. Before the CPU was established, the volume of reports far exceeded the capacity for manual scrutiny. In 2023, police received 2,75,912 CyberTipline reports but processed only 21,817, resulting in 11 FIRs and five arrests.Around 28,000 reports were processed in 2024. After the establishment of the CPU and integration of AI into investigations, the number of processed tiplines rose sharply to 1,20,563 in 2025. Based on 925 specific tiplines, police registered 866 FIRs and arrested 421 people.The momentum continued in 2026, with 40,755 tiplines processed and 192 suspects arrested in the first five months alone. The bureau’s analysis found that Instagram accounted for 84% of reported material, followed by Facebook at 10%, Google at 1.7%, WhatsApp at 1.6%, and Snapchat at 1.1%.The CPU also identified clusters of repeat offenders. In one group, 52 individuals were linked to 151 tiplines, while another 41 individuals were connected to 142 reports. A statewide operation targeting repeat offenders led to 28 arrests.How C-SIGHT works: The technology behind T’s child protection driveHyderabad: TGCSB said combination of automated tipline processing, risk-based prioritisation, proactive cyber patrolling and investigator supervision enables the CPU to handle a much larger volume of reports while keeping sensitive evidence within the police network.Bureau director general Shikha Goel said proactive detection allowed investigators to intervene before such material spread further. “The objective is early detection and faster action against those circulating or selling child sexual exploitation material on messaging platforms, peer-to-peer networks and hidden parts of the internet,” she said.While the arrests and rescues have drawn attention, the real innovation behind Telangana Police’s anti-child abuse initiative lies in the design of C-SIGHT itself.Unlike most artificial intelligence applications that depend on cloud computing, C-SIGHT operates entirely within a secure, offline government environment. This makes it India’s first sovereign AI platform built for child sexual exploitation investigations.The system was developed to process the enormous volume of CyberTipline reports received from the National Center for Missing and Exploited Children (NCMEC), which alerts law enforcement agencies worldwide about suspected online child abuse material and exploitation.Traditionally, investigators had to manually examine each report, verify digital identifiers and establish links between suspects and offences — a time-consuming process that often delayed action. C-SIGHT automates much of this preliminary analysis. It categorises reports, identifies patterns, flags high-priority cases and generates investigative leads, allowing officers to focus on field inquiries and victim protection.Because all data processing takes place within State-controlled infrastructure, sensitive information relating to children and ongoing investigations remains protected from external access. Officials say the platform is intended to assist, not replace, investigators. By reducing the time spent on repetitive analytical tasks, it enables faster decision-making in cases where timely intervention can make the difference between continued abuse and a child’s rescue.
