ECI’s Questionable Tech, Rogue AI And The Aadhaar Paradox

Last week, the Election Commission was in the news for what was happening inside it.

An Indian Express investigation reported that two Election Commissioners had raised 14 objections over 10 months to changes around voter registration, including concerns about the centralisation of the electoral roll database. 

At the centre of some of these concerns was ECINet, the ECI’s new technology backbone, which brings its electoral systems under one roof. The investigation found instances where software controlled from Delhi could restrict what Election Registration Officers on the ground were able to do, even though the law assigns them responsibility for the electoral roll.

Another The Reporters’ Collective investigation found a “Mark VIP” option in the ECI’s BLOApp during Delhi’s SIR, with informal WhatsApp messages asking election officers to ensure these voters were not deleted, to avoid “inconvenience or embarrassment” to the ECI.

The ECI has disputed the characterisation, saying its decisions were unanimous and part of institutional deliberation.

But beneath the disagreement over who said what was a larger question: what happens when technology stops being a tool for implementing decisions and starts shaping who gets to make them?

That question also came up at a consultation on AI, rights and accountability at the National Law School of India University, which I was part of. The two-day consultation looked at AI and digital systems across welfare delivery, land and the Special Intensive Revision of electoral rolls.

Rajendran Narayanan, an Assistant Professor at Azim Premji University who was part of the consultation, made an important connection about the systems: what we are seeing with SIR, he said, is already familiar from welfare systems.

A person’s identity gets tied to a number. Consent becomes secondary to getting a system to work at scale. The burden of proof moves from the state to the citizen. Accuracy becomes less important than getting the numbers onto a dashboard. And when the system fails, it is often the person at the other end who has to prove that they are eligible.

“Only when elites are affected, it makes it to the national news,” Narayanan said.

There was another pertinent idea that came through strongly in the room: technology itself is not the problem. The question is what we choose to make it do.

If an algorithm should not decide whether someone gets food, a pension, land or a place on the electoral roll, technology can still be used to track exclusions, expose patterns, audit systems and make failures visible. It can help citizens and civil society scrutinise the systems that govern them, rather than quietly becoming part of those systems.

On Decode This Week, we look at some of those fault lines, where technology can either make power harder to question or make it harder to hide.

Here are the stories that caught my attention.

Meta’s New Smart Glasses Cannot See, But Still Listen

Meta’s latest Ray-Ban smart glasses have dropped the camera but added six microphones for audio and voice commands. The move comes as glasses with tiny cameras are being dubbed as ‘pervert’ glasses and face growing privacy concerns. That means the privacy question now includes not just what these glasses can record visually, but what they can hear, and what happens to people around the wearer who never chose to interact with them.

An AI Agent Accessed An Australian Government Portal

An AI agent by OpenAI accessed public and non-public files on an Australian government Medicare portal during a security test. After encountering “repeated blocks”, it found ways around them, prompting OpenAI to say it took actions it was not intended to take. Australia is now considering whether there could be legal consequences. As AI agents move from generating answers to taking actions, who is accountable when they cross a line?

Your LPG Subsidy Now Comes With An Aadhaar Check 

Consumers seeking subsidised LPG cylinders will have to complete Aadhaar authentication from October 1. The government says 89.9% of active domestic LPG consumers had completed the process by September 19. Those who do not authenticate can continue to buy LPG at the market price, after registering their choice. Another everyday service has now acquired another digital checkpoint.

Karnataka Says Aadhaar Isn’t Proof Of Age For Marriage 

Karnataka has tightened its marriage registration rules, directing officials not to accept Aadhaar as proof of age for brides and grooms. Applicants will instead need documents such as birth or school certificates, with a medical certificate permitted where these are unavailable. This comes as another reminder of the limits of Aadhaar as proof, even as the same ID continues to become a gateway to welfare and public services. 

Truecaller And TRAI Are At Odds Again 

Telecom Regulatory Authority of India’s new rules require caller-ID and call-management apps to share user reports of spam calls with telecom operators, feeding into the industry’s anti-spam system. Truecaller has called the requirement a “one-way exchange” and “anti-competitive”, arguing that it hands over commercially valuable data to telecom companies. The earlier tussle over who gets to decide which calls are spam, and who gets the data to do it, remains. 

In Case You Missed It

At Decode Internet this week, we followed stories of AI, people, and the internet.

Ranu Mondal Is Still Going Viral. She Doesn't Know It.
The singer who once recorded with Himesh Reshammiya now lives alone, barely eats, and has no idea the internet is still making money off her. (Read here)

Can AI Really Kill All Humans, Or Are We Looking At The Wrong Risk?
While AI agents can behave in unexpected ways, scenarios of AI independently wiping out humanity are far less plausible than they sound. (Read here)

Decode This Week is your weekly guide to the technology changing how we live, work and navigate the world around us. From AI and algorithms to scams, surveillance, platforms and digital policy, we look beyond the hype to ask a simple question: what does this mean for you?

Until next week :)