The Prayagraj Smart City Limited (PSCL) officials have begun experimental deployment of generative AI (GenAI) within its existing AI-enabled monitoring infrastructure at Magh Mela 2026, in a bid to strengthen surveillance and crowd-management mechanisms.
The new GenAI layer is being integrated to enhance real-time threat detection and improve response to emerging challenges associated with massive pilgrim gatherings.
The move marks a significant technological upgrade to the city’s security apparatus ahead of future large-scale religious congregations, including the 2031 Kumbh.
Out of the 400 CCTV cameras installed across the mela area and major entry and exit points, 197 are already equipped with AI capabilities.
According to Mani Shanker Tripathi, IT manager, Prayagraj Smart City Limited, specialised teams working under L&T—the master system integrator—are refining the existing AI software to deliver next-generation surveillance functions.
“The system is being fine-tuned to detect physical altercations, flag suspicious movement patterns such as a person repeatedly visiting the same location more than six times within 30 minutes, and incorporate elements of emotional intelligence for better troubleshooting and crowd security,” Tripathi said.
He explained that the GenAI presently being tested has two ends, one which picks up input based on criteria set in the software while the other assesses and either accepts or rejects the input. For example, if the AI spots a fight like situation. The input would be assessed by the discriminator which would either raise an alert after confirming or reject the input. Several features, one by one would be added to the software to assess human behaviour, as the testing progresses. This form of AI, unlike the traditional AI will enable deeper situational understanding,” he added.
{{/usCountry}}He explained that the GenAI presently being tested has two ends, one which picks up input based on criteria set in the software while the other assesses and either accepts or rejects the input. For example, if the AI spots a fight like situation. The input would be assessed by the discriminator which would either raise an alert after confirming or reject the input. Several features, one by one would be added to the software to assess human behaviour, as the testing progresses. This form of AI, unlike the traditional AI will enable deeper situational understanding,” he added.
{{/usCountry}}The official stated that the deployment remains in the testing phase and will undergo improvements through upcoming Magh Melas before being fully fine tuned for the 2031 Kumbh.
Beyond security, the enhanced AI system aims to significantly improve the accuracy of pilgrim flow monitoring, a key requirement for preventing stampedes during high-density events at the Sangam. The upgraded architecture offers real-time detection of pilgrims and vehicles, optimised to function reliably in crowded environments. Its object-tracking algorithm prevents double-counting by maintaining identity continuity across frames, while pre-processing tools adjust for fluctuating lighting and visual noise.
Early performance metrics from the field trials show promising results, with officials reporting a 95% accuracy in tracking.
“The introduction of GenAI marks a crucial step towards creating a safer, more responsive, and technologically resilient monitoring ecosystem for one of the world’s largest annual religious gatherings,” said the IT expert.