This plan includes not just facial recognition technology (FRT), but also automatic number plate recognition and predictive analytics
Starting June, the Delhi Police will roll out a citywide artificial intelligence-based surveillance system, marking a major shift from localised experiments to a centralised facial recognition network. The move is part of a broader transformation of the city’s policing infrastructure, aimed at boosting both security and crime prevention capabilities.
This plan includes not just facial recognition technology (FRT), but also automatic number plate recognition and predictive analytics. At the heart of this expansion lies the new Integrated Command, Control, Communication & Computer Centre—commonly referred to as the C4I. This facility, being developed by the Centre for Development of Advanced Computing (C-DAC), is envisioned as the nerve centre of Delhi Police’s AI-powered operations.
The C4I will consolidate surveillance data from across the city, receiving real-time feeds from over 10,000 high-resolution CCTV cameras. It will also integrate legacy video networks maintained by municipal bodies, Residents’ Welfare Associations (RWAs), and the Public Works Department (PWD). The footage, once analysed by advanced AI models, will allow officers to detect suspicious activities more swiftly and even track down individuals using facial recognition, including those partially visible or in disguise.
Presently, facial recognition is already in use in certain parts of the city. North and North-West Delhi police districts operate mobile surveillance vans equipped with an Israeli facial recognition software. Initially procured in 2018 for tracing lost and found children, this software has since been deployed during several sensitive events and investigations, including the 2020 North-East Delhi riots and the 2022 Jahangirpuri violence. It was also used during major national events such as Republic Day and Independence Day parades, and more recently, during the G20 summit in 2023.
The upcoming system, however, aims to take the technology to a new scale. “We are aiming for match times under five seconds,” said B S Jaiswal, who was Joint Commissioner (Tech and PI) and was recently transferred to his new post as Joint Commissioner (Central Range).
The new system will also incorporate features beyond facial recognition. Gunshot detection, crowd estimation, and automatic alerts for individuals who have collapsed or appear to be in distress will be part of the toolset available to officers. Up to 1,000 CCTV feeds can be monitored live at once from the C4I centre, and emergency operation centres are being designed to relay potential threats to the respective police stations and district headquarters in real time.
Supporting the technical backend of this initiative will be the Picture Intelligence Unit (PIU), a new division that will manage audit logs and access multiple national databases, including traffic e-Challan records and even telecom and banking data. The PIU will also help in training the AI software by tagging images from news reports, police raids, and public submissions to improve its accuracy over time.
Yet, the expansion of this powerful surveillance infrastructure is not without its critics. Apar Gupta, Executive Director of the Internet Freedom Foundation (IFF), called it “a quantum leap in the state’s ability to identify and monitor individuals.” He cautioned against the increasing potential for intrusive surveillance, adding, “Facial recognition is uniquely intrusive: real-time, automated identification at scale, erasing public anonymity.”
Privacy concerns, legal frameworks, and the risk of misidentification remain key challenges. The potential for bias in AI models—particularly when facial recognition systems struggle with diverse demographic data—has raised further questions around the reliability of such tools in law enforcement.
Jaiswal acknowledges these concerns, saying the technology is not without its flaws. “From poor camera angles to weather interference, and from demographic bias in AI models to privacy-compliant training datasets,” he said, “there are challenges.” However, he remains optimistic: “Like beat cops learn to spot anomalous situations, our machines will learn too.”
As Delhi prepares to step into a new era of digital surveillance, the balance between public safety and individual rights will be closely watched—not just by the citizens, but also by digital rights organisations and policymakers across the country.