Did AI fail to prevent fatal stampede at India's Kumbh festival?
A man helps a relative cross over a barricade, after a deadly stampede before the second "Shahi Snan" (royal bath), at the "Maha Kumbh Mela" or the Great Pitcher Festival in Prayagraj, India January 29, 2025. REUTERS/Adnan Abidi
What’s the context?
Branded the "Digital Maha Kumbh", organisers use tech to control crowds, authorities say AI not enough to prevent stampedes
Organisers of India's Maha Kumbh festival made much of their use of artificial intelligence (AI) to manage the crowds of millions who gathered for the Hindu festival, but that did not stop nearly 40 people being killed in a pre-dawn stampede this week.
Held every 12 years, more than 400 million people attend the six-week Maha Kumbh Mela or Great Pitcher Festival, in Prayagraj, a city in the northern state of Uttar Pradesh.
India has a grim track record for stampedes at religious festivals - 36 were killed at the last event in 2013.
Critics have blamed organisers for failing to prevent the latest crush despite extensive tech surveillance and security, but authorities say they were helped by AI and other factors were to blame.
"The AI technology-enabled cameras were sending us timely alerts on crowd surge from every junction, but there were not enough policemen on the ground to prevent people from creating a jam ... where the stampede happened," a senior administrative official told Context on condition of anonymity as he was not authorised to speak to the media.
A police official, also requesting anonymity for the same reason, said while policemen rushed to the confluence of the rivers where the crush happened, another stampede was sparked on the other side of the river with limited forces to control it.
'AI also learning'
Authorities deployed AI software-based technology at the event spread across 4,000 hectares (9,900 acres) - equivalent to 7,500 football fields - on the banks of rivers Ganga, Yamuna and the mythical, invisible Saraswati.
They employed some 2,760 close circuit television cameras, drones and tethered drones to monitor the festival area as well as anti-drone craft to find and remove unauthorised drones.
Authorities also used underwater drones that can operate to a depth of 100 metres (3,330 feet), equipped with advanced technology to function in low-light conditions.
The AI tech not only calculates crowd numbers, but also alerts authorities when there is a surge in any section of the festival city, a fire, or if people cross barricades. The alerts are sent to personnel on the ground to take corrective action.
Police agreed that using AI has had its limitations at the festival, which began on Jan. 13 and ends on Feb. 26.
"The AI is also learning. This is the first time AI is being used for such a large gathering. Once the festival is over, we will see where it was right and where it was wrong," said Amit Kumar, a senior police officer heading the control centre.
"We are not totally dependent on AI and we do have human intelligence in place to keep an eye and ear on the ground."
(Reporting by Saurabh Sharma; Editing by Annie Banerji and Jonathan Hemming.)
Context is powered by the Thomson Reuters Foundation Newsroom.
Our Standards: Thomson Reuters Trust Principles
Tags
- Tech solutions