00:00 IST 33°C

Opinion

When the camera is the judge: Bengaluru's automated fines punish first, verify never

Bengaluru's AI cameras issue 19,000 challans a day, and the official story is 99% accuracy. But when a guitar became a helmetless pillion rider — and the human reviewer signed off on it — the machine showed its real philosophy: accuse automatically, and make the citizen prove otherwise.

By The Justice news desk

· 2 min read

A surveillance camera on a pole
A surveillance camera installation. When Bengaluru's AI cameras flag 19,000 violations a day, who checks the checker? (Editorial illustration) Photo: Eyetechsecurities
The photograph is perfect: a helmeted man, a guitar, and a machine that saw a second human. The comedy ends where the bureaucracy begins. Bengaluru's AI cameras issue some 19,000 challans a day. When the system was adopted in December 2022, what was sold was efficiency; what is now defended is infallibility. The official story: 99% accuracy, only 0.13% challenged. Read it the other way around: the challenge rate does not measure accuracy; it measures the price of contesting. A Rs 500 fine buys silence more cheaply than justice. The state has discovered the perfect fine — one that costs less to pay than to fight. The human review exists, and it failed. A person looked at a photograph of a man with a guitar and approved a fine for a helmetless pillion rider. That is the honest part of this story: automation does not remove human error; it launders it. The officer's judgement becomes the machine's output, and nobody is responsible. Another officer has admitted the real accuracy may be closer to 90% once manually rectified cases are counted; at the end of 2023 it was 80% for certain violations. These numbers were not volunteered. A newspaper had to extract them. Consider what minimal human intervention actually meant in the guitar case. The system flagged, a human glanced, the fine went out. The review exists to legitimise the automation, not to check it — a rubber stamp with a lunch break. When both the machine and its minder fail at once, the design has failed, not the individuals. And consider who pays. The motorist who can spend a morning at a traffic office disputing Rs 500 is not the motorist who rides to a shift. Wrongful fines are a regressive tax on time, and time is the one currency the poor cannot spare. Smart policing, priced by the working day, punishes the worker first. The Periyarist frame fits here: power always finds a theory for its own innocence. Earlier it was the officer's discretion; now it is the algorithm's objectivity. The sanctity has moved from the uniform to the server, but the citizen still stands exactly where he stood — proving his innocence to the machinery that accused him. The demands are plain: publish the real error rates; submit to independent audit; make wrongful fines compensable, not merely revocable; and restore the presumption of innocence. The camera is a witness, not a judge.

Sources

Share this story

AMP

Related articles

A farmer working a field with a tractor in India (file photo)
Opinion

The drought is natural; the delay is political

When a state must beg Delhi for disaster money that is owed by right, federalism has become charity. Karnataka's drought letter exposes a system that makes farmers wait for files while the rain fails.

· 2 min read

A 'U-turn ahead' road sign — editorial illustration
Opinion

A government of reversals is a government without a rudder

Opinion: the TVK government's habit of announcing orders and withdrawing them within days — from the RTI exemption to student protest rules — is not responsiveness but trial-and-error governance, and the trials are being conducted on the public.

· 2 min read