Dhaka automates traffic fines with AI: real efficiency, but 38,000 unresolved penalties reveal the bottleneck isn't the camera

By Momentum IA · June 30, 2026.

The Dhaka Metropolitan Police (DMP) has expanded its smart surveillance network to 19 intersections in the Bangladeshi capital, the latest addition being Lake Road, a high-security corridor adjacent to the national Parliament. Commissioner Mosleh Uddin Ahmed announced that the system has already logged 1,500 automatically registered cases. The cameras, connected in real time to the DMP Command Center, detect speeding, dangerous lane changes, motorcyclists without helmets, illegal parking and traffic obstruction. When they detect a violation, they generate the electronic citation without any officer having to intercept the vehicle.

So far, this is the success story the announcement wants to tell. But there is one figure the commissioner slipped in almost in passing that deserves more attention: **38,000 violation citations remain pending resolution**. In other words, the detection machinery is operating at a pace the judicial and administrative apparatus cannot absorb. This is no minor problem: when the backlog of unresolved cases vastly exceeds processing capacity, the real deterrent effect —the very purpose of any enforcement system— is diluted. A fine that arrives months late, or that is never collected, is noise, not signal.

**Our take:** this case in Dhaka clearly illustrates a tension we will see repeated on a global scale as AI is deployed in public services across the Global South: automating *detection* is technically cheap and quick to install; reforming the *process* behind it —case management, notifications, collections, appeals— is slow, costly and politically cumbersome. Bangladesh is not alone in this. Cities in India, Kenya, Indonesia and Brazil are adopting similar cameras and will run into the same bottleneck: it is not the AI that fails, but the institutional infrastructure that has to turn that detection into real consequences.

There is also a dimension the article does not address and that should not be ignored: the cameras have been installed, among other reasons, in an area classified as a VVIP movement zone —that is, of senior state officials. In contexts where institutions have weaker accountability, ubiquitous surveillance systems can easily drift toward uses other than road safety. There is no accusation here —the source offers no indication of this— but it is the kind of infrastructure that demands clear legal frameworks on what data is retained, who can access it and for what purposes, something the official announcement omits entirely.

In terms of the long-term trend, automating traffic enforcement has genuine potential: it reduces officers' exposure to dangerous situations, eliminates arbitrary discretion —and the corruption that sometimes accompanies it— and can improve road safety in measurable ways. The commissioner himself noted that the layout of Lake Road encourages the speeding and dangerous lane changes that cause frequent accidents; any improvement in traffic flow and compliance has concrete humanitarian value. In the long run, well-governed systems like these save lives.

But that "well-governed" is the hard part. And the 38,000 unresolved cases are, for now, the most honest measure of how much remains to be built.