Delhi Police Tender Links Facial Recognition to Protest Routes

Five days after the Cockroach Janta Party called off its weeks-long sit-in at Jantar Mantar, Delhi Police quietly floated a ₹15.66 crore tender for 338 cameras fitted with facial recognition, built-in microphones and crowd-detection software. The timing, the siting, and the paperwork together raise a question that goes well beyond procurement: whether the capital's surveillance infrastructure is being built around the geography of dissent itself.

A tender that maps onto protest ground

The bid, placed on the Government e-Marketplace on 30 July 2026 and closed on 20 August, seeks 248 fixed cameras, 43 pan-tilt-zoom units and 47 automatic number-plate readers for Central and New Delhi districts. Officially, the justification is VIP and VVIP security for Republic Day and Independence Day routes. But the site list, reviewed in tender documents, reads like a map of the city's protest arteries: Rajghat, Delhi Gate, Turkman Gate, the India Gate hexagon, Mandi House, Patiala House, and - pointedly - the VVIP gate of Ramlila Maidan. On 11 October, the Commissioner of Police de-notified Jantar Mantar and designated Ramlila Maidan, along with Burari Ground and CBD Ground Shahdara, as the city's new sanctioned protest sites. Of the three, only Ramlila Maidan sits inside the zone this tender already covers.

Built for faces, not just traffic

Every camera class in the specification carries facial recognition as a mandatory feature, not an optional add-on. Fixed cameras must run intrusion detection, tripwire alerts, people-counting and face recognition simultaneously. The pan-tilt-zoom units must support face recognition outright, and even the number-plate readers are required to perform face detection. The video-management software must embed a facial recognition system under a "perpetual license," while a mobile application is meant to stream live footage from up to 500 cameras at once. Clause 1.7(f) instructs that cameras be positioned "to identify and capture the face of a particular person roaming around at critical/vulnerable points" - language that sits uneasily beside a public assembly ground. Built-in microphones on every camera type, with no stated rule against recording sound, add an audio dimension the tender never explains or justifies.

Gaps where the law should be

What the 76-page draft does not contain is as telling as what it does. There is no mention of the Ministry of Home Affairs' Standing Order governing such deployments, no privacy impact assessment, no identified database against which faces will be matched, no similarity threshold, and no correction mechanism for false matches. The only retention figure given - a minimum 30-day storage window - describes disk capacity, not a data-protection policy. This arrives against the backdrop of Delhi Police telling the Supreme Court that facial recognition run on Jantar Mantar crowds in July 2026 produced 2,873 "matches," a figure independent reporting has already called unreliable. A Supreme Court committee under Justice R. Subhash Reddy is separately examining the legality of such surveillance, even as police issue fresh standing orders and procurement continues on parallel tracks.

A tender riddled with errors

The document itself undermines confidence in how carefully it was assembled. Its technical annexure appears lifted from an unrelated 194-camera tender for North District, complete with that bid's page numbers still visible. The bill of quantities transposes pan-tilt-zoom and number-plate camera counts. Software specifications echo a single vendor's product brochure closely enough to name its proprietary features outright, a practice that GeM rules are meant to prohibit. One camera's coordinates place it roughly 40 kilometres outside Delhi. The integrity pact names anti-corruption monitors whose terms expired years ago, ignoring a replacement appointed by the Ministry in 2025. These are not isolated slips; researchers have logged 28 such defects across the document - a strikingly loose foundation for a system meant to make permanent, searchable records of who gathers where, and why.