AINEWS
A Delhi High Court public interest petition argues there is no statutory basis for live facial recognition at a protest, and that an 80% match threshold is set by police themselves.
When Srishti stepped near a Delhi police control van on July 28, her own face appeared on the monitor inside: a green rectangle, a name, a match score. She was covering the Cockroach Janta Party (CJP) sit-in, an activist protest at Jantar Mantar, Delhi's longstanding and legally recognized demonstration ground. A Reuters reporter on the same story saw the same thing on the same kind of monitor, his own face boxed and labeled before an officer pulled the van's window shut (TechTimes). The vans were running. The software was matching faces. The legal basis for doing either has not been made public.
Delhi Police deploy the vans as part of a fleet they call Ikshana, Hindi for "vision" or "eye." Each vehicle carries eight AI-enabled cameras on a telescopic mast with a 360-degree field of view, integrated with facial-recognition and video-analytics software that runs faces against a police database in real time (ThePrint). The Ikshana fleet was first inducted ahead of the G20 summit in New Delhi in 2023. A senior Delhi Police officer confirmed the system was running during the CJP protest (TechTimes, Big News Network).
The match threshold that turns a protester's face into a flagged "positive" hit is a Delhi Police internal benchmark, not a rule issued by any court or parliament. A Delhi Police report, cited by Deccan Herald, treats an 80% similarity score as a "positive match" against the police facial-recognition database. No Indian statute, regulation, or judicial precedent authorizes that threshold for use at a protest site. The question on the docket is whether the state can run facial recognition against a lawful assembly under a match score the operator sets for itself.
The petition that put that question on the docket was filed on July 17, 2026 by Aishe Ghosh, alleging "indiscriminate and continuous surveillance" of protesters and the installation of a permanent surveillance tower at the site (The Hindu). The Delhi High Court agreed to hear the matter on July 20. Per LiveLaw, the bench has not granted interim relief; instead, it has suggested a fresh plea after examining police standing operating procedures and the Digital Personal Data Protection Act. The legal-vacuum claim is a gap, not a holding. There is no ruling yet, but the question is now one the court has agreed to test.
In parallel, the Internet Freedom Foundation has written to the Commissioner of Delhi Police and filed Right to Information requests asking for the legal basis, the SOPs, and the data-handling rules behind live facial recognition at Jantar Mantar (IFF). The filings are the first public attempt to convert the absence of a cited statute into a record the court can weigh.
Reporters at the site documented an internet shutdown in the protest area, mobile-phone jamming, the logging of protesters' Instagram handles, and the alleged use of Meta smart glasses to photograph demonstrators (Newslaundry). Newslaundry also reported that Jantar Mantar reportage and reels are being taken down, and that social posts characterized as "offensive" are being targeted. These are press-freedom questions with their own standing. They share with the facial-recognition case the same underlying fact: the protest site is being treated as a surveillance perimeter, with the rules of that perimeter unpublished.
The next hearing will test whether the court takes the petition as a constitutional question or routes it through SOP review under the Data Protection Act. The IFF RTIs, the Newslaundry tracking, and the CJP's own video record remain the public record while the bench decides.