In a scene from the Hollywood film Minority Report, in April, the UK’s Metropolitan Police began investigating hundreds of police officers thanks to an AI tool. The tool, called Nectar and created by US spy-tech company Palantir, aims to root out rogue cops through internal police force data.
AI software runs algorithms on existing internal data to detect patterns, anomalies, and red flags. It looks for what doesn’t fit and it’s very good at it.
Within a week of being deployed, the tool flagged hundreds of UK officials for violations – ranging from abuse and misconduct when working from home to serious cases like corruption and even rape. The AI tool exposed officials who abused roster shifts for personal or financial gain, or who were accused of misconduct in public office. So far, three officers have been arrested while 98 officers are being investigated.
The software deployment is part of the UK Police’s AI hub announced earlier this year. With a budget of £115 million (approx.) ₹1480 crore), the AI Center seeks to adopt AI across 43 forces in England and Wales to fight crime, centralize innovation and strengthen testing. Speaking about the software, Alex Murray, director of the UK’s National Crime Agency, said in a press release: “Police are facing increasing demands, complex digital crime and continuing financial pressures.”
Britain is not alone in applying AI to fight crime. As Murray said, policing is becoming increasingly complex. Digital and financial crimes are abusing porous borders and becoming sophisticated with AI, while militaries face budget pressures. Policing systems around the world, from India to the US to South America, are trying out different ways AI can help them become more efficient and root out more crime.
Getting Super-Powerful Eyes
The first phase of AI adoption by police in the West and India was through Facial Recognition Technologies (FRT). These technologies run on top of CCTV feeds to recognize faces and objects in the crowd. A policeman can ask AI to detect a person in an orange kurta or a blue car in several weeks of CCTV footage, reducing days of human intervention.
The purpose of FRT, which is still being implemented and improving due to AI advancements, is to respond rapidly to an evolving situation or during crowd management. For Maha Kumbh 2025, Uttar Pradesh Police used AI analytical software by Gurgaon-based startup Stack Technologies Pvt Ltd to run video analytics on top of existing CCTV to detect live incidents – crowded areas, fires, brawls – so they could respond faster. Nashik, which is hosting the Simhastha Kumbh Mela in 2027, held an AI strategy workshop with MIT and companies like Meta, Google, Microsoft and Indian startups to build a Kumbh AI stack that can simultaneously process live feeds from 5000+ CCTV cameras, IoT sensors, surveillance drones and mobile networks to keep track of 12 crore people visiting its city.
solves the problem of abundance
The rapid, even forced, adoption of AI sounds a bit like competition or FOMO, but it is a necessary requirement for policing modern life. Thanks to the proliferation of recording devices in public places such as CCTV, drones, bodycams, footage, forensic logs and digital evidence such as social media, the problem for police is not a lack of evidence. This is too much.
It is not humanly possible to scan more than 5000 CCTV cameras installed during Kumbh in Nashik. However, AI is adept at sifting through mountains of data.
Increasingly, police departments are handing over all aspects of forensic analysis to AI to generate patterns, a method called predictive policing that identifies promising targets, crime hotspots for police intervention, even solving old crimes.
India’s fragmented, state-specific policing is a challenge for this massive database of big data, which is why the Union Home Ministry is launching an Integrated Crime and Criminal Tracking Network and System (CCTNS 2.0). CCTNS 2.0, which is already being implemented, connects 17,000 police stations across India on a centralized online platform and is part of a larger initiative called Inter-Operable Criminal Justice System (ICJS) that connects police, courts, jails and forensics into a single data pool.
Run AI tools on this huge data, and the Home Ministry expects greater efficiency in predictive policing – find crime hotspots or high-risk individuals or manage live emergencies.
Can AI be a good spy?
Countries that have implemented predictive policing are already seeing amazing results in AI in finding new clues and suspects in previously unsolved cases. In Anchorage, Alaska, US, the police system adopted the technology of Closer, a startup that searches large datasets for evidence and solves old cases. Its AI agent combed through police jail calls, interviews, social media, photos and old case files to flag new moments and important messages, even in different languages.
“There are some cases where you have detectives listening to over 1,000 hours of jail call data to try to find a word, a phrase, a name, a threat, things like that,” said Sean Case, police chief at the local assembly at the Anchorage Police Department earlier this month. “When we tested the software, one of the things we primarily used it for was to throw out jail call data.”
Over the past few years, there has been a proliferation of AI startups that seek to automate the labor-intensive tasks of real-time monitoring and analyzing massive police datasets. Clojure is one, LongEye is another that offers something similar to Seattle’s police department. AI is the new spy that police forces have – but only if you have a huge dataset on your citizens.
empowering specific departments
A few months ago, the Counter Child Sexual Exploitation (CCSE) team of Kerala Police found a forum named ‘Cheese Pizza’ on the dark web, which was linked to child sexual exploitation material. He saw several photographs of a child whom he suspected to be from Kerala. Using Catalyst, an AI tool developed by New Zealand-based non-profit Kindred Tech, they examined a large number of photos and data points scattered across social media platforms like Facebook, Telegram and Instagram to catch the potential culprit. Within a few weeks, they arrested a woman in Thiruvananthapuram who was posting pictures of her niece. “This problem has been aggravated by technology. We should use technology to fight it,” Ankit Ashokan, SP Cyber Crime, Kerala Police, said during a press interaction. He said a small team like his would not be able to manually review or examine all this data.
combating financial fraud
Another place where AI is proving very useful is financial crime. Interpol launched Operation Shadow Storm in March which uses AI-powered data analytics to track money trails of global “scam hubs”. Within seconds after a transaction, AI Case predicts where the stolen funds will flow next and freezes these accounts across borders.
India’s Home Ministry is working with IIT Bombay and RBI to use AI to identify suspicious mule accounts to combat financial fraud. The ministry wants to develop a model that works in real time on financial transactions so that banks can identify and stop potentially fraudulent transactions. The tools developed by it identify suspicious mobile connections through pattern recognition, real-time tracking of cyber criminals, identify mule accounts and also monitor the dark web and social media.
bad, weird and ugly
While AI can help streamline processes, even finding small details in piles of data, using it in policing can also have serious consequences on civil liberties and rights – if done wrong. The Government of Maharashtra in collaboration with IIT Bombay is currently building a ₹3 crore AI tools that will help law-enforcement agencies identify suspected illegal Bangladeshi citizens in the state by analyzing their speech patterns, intonation and linguistic usage. Will this tool made in Mumbai be able to differentiate between the pronunciation of dialects of West Bengal and dialects of Bangladesh? Or will this turn out to be another case of poorly implemented software like Minority Report?
In Australia, police used CCTV to automatically detect whether a person is wearing a seat belt or using their phone while driving. Based on this, the police issued a fine of Rs 300 per day. Sixty percent of these AI-detected fines were overturned – particularly those flagged by AI for using a wallet, glass case or battery pack as a phone.
Potential misdirection due to limitations of technology is one thing, surveillance and bias against citizens is another. In Minority Report, the protagonist is happily using a predictive policing method called Pre-Crime (which uses mutants to spot crimes before they happen), to arrest citizens who might commit crimes. He sees how wrong this is when the predictable police pull his name.
AI is great, when used well, but poorly created AI agents based on faulty datasets, automation bias or even human biases can lead to false positives and misuse.
The author tracks the evolving relationship between science, technology and modern society. She also works as a philanthropy researcher and consultant.
