Apart from Facial Recognition System (FRS)-enabled CCTV cameras, the Delhi Police is set to use AI-powered smart glasses to identify “wanted individuals”, referring to probable accused, convicted criminals, or fugitives, at the Republic Day parade, held every year on January 26 at Kartavya Path (previously known as Raj Path) in New Delhi.
“This time, we are also using specially developed cameras and wearable glasses. We have fed different algorithms into the database of these standalone systems,” said Delhi’s Additional Commissioner of Police, Devesh Kumar Mahala, at a press conference while speaking to news agency ANI.
How does the Delhi Police’s AI Glasses Work?
Without specifying the particular datasets used, the Delhi Police official stated that the database will be deployed as a ZIP file on each individual device linked to the AI glasses.
“The FRS cameras installed in place will be connected to our standalone system, which is not connected to the internet. Additionally, we will load the database onto individual mobile phones for the wearable glasses,” said Mahala.
“We have fed different algorithms into the database of these standalone systems. If wanted individuals are detected or spotted, even if they are wearing makeup, a mask, a cap, or have a beard, our algorithm will identify them,” he added.
Explaining the algorithm further, he said, “This database has been trained using different algorithms so that it can identify a person regardless of how they look, whether it is a photo from 20 years ago or a recent one, whether they have a beard or long hair, or even a cut on their face.” He further claimed that their AI systems can detect a person “even if they are wearing a face mask or if it is just a photo.”
Who is behind Delhi Police’s AI Glasses?
According to the Delhi Police, the AI-powered smart glasses are developed by Maharashtra-based Ajna Tech. Founded in 2015, Ajna Lens claims it “aims to bring this future to everyone with our [Extended Reality] XR solutions powered by artificial intelligence,” according to its mission statement.
IDBI-owned trust Maharashtra Defence and Aerospace Venture Fund, Paytm founder and CEO Vijay Shekhar Sharma, and others have invested in the company.
The company’s homepage claims that government entities, including the Ministry of Defence, the Indian Army, and the Indian Navy, trust its services. Other clients include the Aeronautical Development Agency and private companies such as mining and metal processing firms Vedanta and Tata Steel, as well as construction and technology manufacturing company Larsen & Toubro (L&T).
Why a Private Company in Surveillance Tech is a Privacy Concern?
A government-appointed NITI Aayog paper has outlined several privacy and trust-related issues arising from reliance on facial recognition technology (FRT), including:
- Technical inaccuracy: Facial recognition can fail due to lighting, camera angles, image quality, masks, ageing, facial hair, or physical changes, making identification in real-world conditions unreliable..
- Training bias: Systems trained on skewed datasets perform poorly on underrepresented groups, with documented error gaps across skin tone, gender, and region—especially in a diverse population like India.
- Over-reliance by humans: Poorly trained operators may treat algorithmic matches as absolute truth, amplifying errors and leading to wrongful identification or action.
- Algorithmic vulnerabilities: Minor, imperceptible image changes can cause systems to glitch or hallucinate, making FRT vulnerable to manipulation, sabotage, or false matches.
- Lack of independent scrutiny: Closed training datasets and proprietary code prevent independent evaluation, making bias, discrimination, and inaccuracies difficult to detect or prove.
- Data security risks: Centralised biometric databases are high-value targets for breaches or unauthorised access, particularly in outsourced or third-party deployments involving private companies.
However, the Delhi Police has not explained how it will secure the databases deployed on individual devices in ZIP format.
Similarly, in 2022, Ameen Hauhar and Jai Vipra, both senior resident fellows at the Vidhi Centre for Legal Policy, stated in a paper that there is a lack of transparency regarding the roles played by private corporations in government deployments of FRT.
“From a legal liability perspective, there are three main concerns, first, the liability of the private corporation or developer of the FRT algorithm; second, the liability of the state for deploying a flawed algorithm; and third, the potential of holding the algorithm liable,” the paper stated.
The authors also argued that secrecy surrounding FRT operations grants both the government and private companies a de facto immunity from legal liability, as it becomes nearly impossible to build a proper legal case. They also took cognisance of the difficulty of piecing together information on how state governments engage with private corporations for such deployments.
Questions to Ask:
- Under which specific law, rule, or executive order is the Delhi Police deploying AI-powered smart glasses at a public event like the Republic Day parade?
- Is the use of AI-powered smart glasses limited strictly to the Republic Day parade, or will the Delhi Police retain and reuse them for routine policing activities?
- Which specific databases does the Delhi Police deploy as ZIP files on each smartphone linked to the AI glasses?
- Will the ZIP file database be encrypted? If so, who holds the encryption keys? If not, why not?
- Does the dataset link to Aadhaar or the Crime and Criminal Tracking Network and Systems (CCTNS)?
- What due process does the Delhi Police follow to add new entries to the existing database?
- Which datasets does the Delhi Police use to train and populate the facial recognition database loaded onto the devices?
- What due process is followed if the system falsely identifies an individual as someone on the wanted list?
- Will images of non-matched individuals be stored, cached, or permanently deleted immediately? If stored, what is the data retention period?
- What is the tested accuracy rate of these AI glasses in real-world conditions, particularly in crowded public spaces?
- Who is liable for false positives, the company that developed the algorithm, the algorithm itself, or police personnel who relied on the AI output?
- Is there an independent auditing mechanism to check the algorithms for potential bias? If so, why has the Delhi Police not made it public?
MediaNama has filed an RTI with the Ministry of Home Affairs seeking clarity on the tender process, if any, and the auditing mechanisms behind the deployment of these systems.
Curated by Shiv Shakti Mishra
