Bengaluru’s emergency response system to get AI boost

Bengaluru: Bengaluru police are on track to integrate an Artificial Intelligence (AI) model into their Emergency Response Support System (ERSS), with testing yielding promising results, multiple sources aware of the development have told DH. 

Once fully operational, it will minimise human intervention and streamline the emergency response process, saving crucial time in dispatching first responders, according to officials. 

Currently, the ERSS has an average response time of 10-11 minutes—from the time a call is made to patrolmen being dispatched. 

"The aim is to make the 112 emergency system self-sufficient in the long term. The AI model is being fed with historical data to learn patterns, types of emergencies (murder, theft, accident, assault, fire, etc) and issue adequate replies to callers," a top source said. 

However, this will not fully minimise human intervention, at least during the early rollout stages, as the models have a long way to go to function by themselves. "The need for human intervention will still exist since the model may take time to learn to respond to sensitive matters," the source added. 

A senior police officer aware of the development said the AI model is also being trained to respond to callers in their selected or native languages. The model will also be equipped with the state police database. So in case of an inquiry call, where the person would need details of their police stations and such, the response will be swift without any delay, he said. 

"The AI model, which will also include a voice assistant, will help the caller in the language they speak during emergencies. The model is being trained to be equipped with the majority of the Indian languages and some commonly spoken foreign ones like French, Spanish and Arabic," the officer said.  

The tender has been floated and testing is going on track. The rollout will happen sooner than expected, the officer added. 

Solution to crank calls?

According to the Bengaluru police’s ERSS data, 26.92 lakh calls were made to the helpline between January and December 2024. Of these, around 58% were blank or crank calls. However, officials believe that human intervention will be crucial to identify the crank calls at least in the initial stages. 

"Crank calls often turn abusive and vulgar, because currently the majority of the 112 operators are women," a senior Bengaluru police officer explained.

"An automated system may see fewer such calls and also will handle them better compared to its human counterparts, since such calls can take a toll on the operator. Although, at least for now, human intervention is extremely necessary in distinguishing these types of calls."