AI cameras at stations
Railways has rolled out AI‑enabled video surveillance across 1,874 stations to detect incidents automatically, a major step toward hands‑free crowd and safety monitoring. (prokerala.com) That system should help operational teams spot overcrowding, unattended baggage and fights faster — though it also raises questions about how alerts will be actioned in busy hubs. (prokerala.com)
Indian Railways says it has now expanded its video surveillance system to 1,874 stations, adding AI-based analytics that can automatically flag events like intrusion and loitering, plus facial recognition for real-time identification and monitoring. The announcement came in a Ministry of Railways release on April 7, 2026, as part of a wider push to modernize the network’s telecom and passenger information systems, not as a small pilot tucked away in one city (pib.gov.in). That scale matters because Indian Railways is not a niche transit system. It is one of the country’s basic operating systems, carrying billions of passengers a year across a network of more than 7,400 stations and more than 13,500 passenger trains a day. A surveillance upgrade at this size is not really about cameras. It is about turning stations into machine-watched spaces by default (rediff.com, ceicdata.com, ibef.org). The technology itself is more ambitious than the April 7 release first suggests. Railway projects described over the past year point to IP-based camera networks with centralized video management, remote access for officials, AI analytics for crowd management, unattended baggage alerts, intrusion detection, and face recognition systems built to Railway Design and Standards Organisation specifications. In other words, the cameras are not just recording footage for later. They are meant to generate live prompts that push staff to act in the moment (etvbharat.com, thehindu.com). That helps explain why the railways keep pairing surveillance with network upgrades. The same April 7 release says IP-MPLS telecom infrastructure has been commissioned at 1,396 stations to support centralized access to video feeds and other mission-critical systems. An AI alert is only useful if it can move quickly from camera to control room to someone on the platform who can do something about it (pib.gov.in). The government has also been steadily widening the purpose of these systems. In July 2025, the Union government told India’s Supreme Court that it planned AI-based facial recognition at seven major railway stations, including Mumbai CST and New Delhi, as part of efforts to curb crimes against women and track people listed in the National Database on Sexual Offenders, which it said had crossed 2 million entries. That is a different use case from spotting a crowd surge or an abandoned bag. It moves the system from behavior detection into identity matching (timesofindia.indiatimes.com, hindustantimes.com). And once that line is crossed, the hard question is no longer whether AI can spot something unusual on a platform. It is who gets flagged, who gets stopped, and what happens when the system is wrong. Indian Railways has offered big numbers and broad claims, but not much public detail on error rates, retention rules, audit trails, or how facial recognition matches will be verified before action is taken. The official language is confident. The operational details are still mostly offstage (pib.gov.in, rediff.com). What is visible is the direction of travel. Southern Railway alone awarded a project covering 484 stations with AI-enabled video systems, including 441 smaller category stations and upgrades at 43 larger ones. South Western Railway has been building integrated command and control centers to watch feeds from thousands of cameras. The April 7 announcement simply reveals how far this buildout has already spread: 1,874 stations, with the cameras watching for loitering before most passengers even know the system is there (etvbharat.com, thehindu.com, pib.gov.in).
