This year, AI has rapidly moved from tech labs into police stations, courtrooms and crowded streets across India. Several states embraced AI to address long-standing problems, including slow investigations, missing offenders, procedural lapses and overstretched personnel.
From facial-recognition systems that claim to solve robberies in hours, to biometric databases mapping criminal histories, to chatbots explaining legal procedures, these tools promise efficiency and precision. But every deployment comes with trade-offs. Critics warn of mass surveillance, misidentification, uninformed decision-making and weak safeguards.
Here’s a list of six instances in 2025 in which Indian states leveraged AI in everyday policing.
1. Delhi
Delhi Police updated their advanced facial recognition system (FRS) this year to address robbery and burglary cases. By late February, reports indicated its effective use, with notable successes including the solving of an ₹80-lakh robbery in Chandni Chowk within 24 hours, attributed to Israeli/Corsight technology. The system analyses CCTV footage and links it to existing databases, significantly expediting investigations and resulting in arrests and the recovery of stolen goods.
However, criticisms emerged regarding privacy concerns, wrongful arrests and the potential for bias, highlighted by a Pulitzer Center report from July.
2. Maharashtra
Maharashtra CID’s AI-based biometric data collection unit in Pune is part of the state’s larger Maharashtra Research and Vigilance for Enhanced Law Enforcement (MARVEL) initiative, which aims to modernise policing through analytics and digital evidence. Launched at Pune Rural Police headquarters, the unit captures multi-angle facial photographs, fingerprints and iris scans of accused individuals. The AI system standardises data and links it to the Crime and Criminal Tracking Network and Systems (CCTNS), enabling the police to track suspects across districts and recognise repeat offenders even if they alter their appearance.
MARVEL’s integration promise is ambitious, but its success depends on accuracy, governance, and transparency, not just technology. As of now, the system’s benefits are mostly projected, while debates over privacy and oversight remain unresolved.
3. West Bengal
West Bengal Police’s AI legal-assistant bot, introduced in December, is designed to reduce procedural errors in investigations and case files. Rolled out to roughly 400 investigators across eight police units, the tool was developed jointly by Birbhum Police and a Pune-based firm. It contains over 50,000 pages of legal material, including case law, NHRC and MHA guidelines, training manuals and victim-support procedures. Investigators enter a brief case description, and the bot recommends relevant IPC/BNS sections, correct procedures and documentation requirements.
The project’s main aim is to cut “legal slips” that can weaken charges or lead to acquittals, offering timely legal references to younger or overstretched officers. However, senior officials caution against over-reliance: AI cannot replace field judgement. Critics also highlight opacity, potential bias in the training data and unclear accountability if wrongful actions result from AI suggestions.
4. Uttar Pradesh
Uttar Pradesh Police and IIT-Kanpur’s RAG-based chatbot, unveiled in mid-2025, aims to make police procedures clearer for both officers and citizens. It indexes more than 1,000 Hindi police circulars, guidelines, and departmental instructions. Users can type everyday questions, such as how to report a vehicle theft, steps in passport verification or election-duty protocols, and the chatbot provides procedural answers in plain language.
For police personnel, this reduces dependence on senior officers or manual document searches, helping avoid procedural delays and misinformation at stations. For citizens, it counters the typical “come tomorrow” response by showing official rules directly. However, outdated or misinterpreted answers could mislead complainants, and query logs may expose sensitive personal information if data retention and privacy safeguards are not clearly defined.
5. Odisha
Odisha Police’s AI-enabled Integrated Command and Control Centre (CCC) in Puri was heavily deployed for the 2025 Rath Yatra to assist with crowd management using AI-processed CCTV feeds, drone footage and density analytics. The system was supposed to detect choke points, alert officials in real time and coordinate warnings via LED panels. However, during the June 29 stampede, which killed three people and injured several others, the technology failed to deliver actionable alerts.
Investigations found that only 123 of 275 cameras were functional, feeds were inconsistent and drones were under-utilised. Authorities recommended blacklisting the vendor and disciplinary action was initiated against seven senior officers for negligence.
6. Telangana
Hyderabad police expanded AI use in November across surveillance, crime analysis and cyber investigations. The city leverages AI-enhanced CCTV systems, facial recognition tools and video analytics to identify suspects, missing persons and risky behaviours in crowded spaces. Thousands of cameras feed data into command centres, where algorithms flag anomalies and help trace movement patterns.
Alongside this, police publicly discussed AI-driven tools for cybercrime, including blockchain analysis and social media monitoring to detect fraud networks and online threats. These systems could provide faster case turnaround and improved tracking in dense urban areas. As deployment scales, the city risks trading public safety for privacy, transparency, algorithmic bias and democratic accountability.
Country-Wide
India proposed the use of AI facial recognition at major railway stations, which emerged in mid-2025, when the Union government informed the Supreme Court that it planned to deploy FRS at seven high-traffic stations to monitor convicted and repeat sex offenders. The system would scan live CCTV feeds, compare faces against law-enforcement databases and alert authorities if a match appears.
However, the plan effectively creates mass surveillance zones that scan millions of passengers without their consent. It is worth noting that railway CCTV often has poor angles, lighting and image quality, which increases the risk of misidentification. It can also be argued that India lacks dedicated legislation governing facial recognition, including use, retention and appeal rights, raising questions about proportionality, privacy and accountability before deployment.
The post 6 Indian States Integrating AI Into Everyday Policing appeared first on Analytics India Magazine.
