Image source: Getty Images
Law enforcement agencies (LEAs) are increasingly using artificial intelligence (AI) to enhance capabilities, particularly in predictive policing capabilities.
Globally, law enforcement agencies are increasingly adopting new technologies. In the US (US), the New York Police Department employs tools such as Patternizr for pattern analysis and executive deployment. Similarly, in China, the government uses robots for crowd control and uses drones and detention cameras to monitor suspicious activity. Scientists are developing virtual reality models in Shanghai, including offices and family housing, to provide real-time assistance to police and rescue services. Both the US and Australia also focus on child protection using AI. Initiatives like the US use of ClearView AI and Australian Centres allow for faster threat detection and prevention in the case of child exploitation to counter child exploitation. Meanwhile, South Korea has introduced car patrol vehicles to ensure road security by integrating voice recognition, video analytics and real-time data processing. These examples highlight the global trend of integrating technologies to address policing challenges.
Scientists are developing virtual reality models in Shanghai, including offices and family housing, to provide real-time assistance to police and rescue services.
The global market size for forecast policing alone is estimated to increase to US$157 billion by 2034, with a CAGR of 46.7% for 2025-34. The prospect of consolidating a huge criminal dataset for a rapid investigation process was a major attraction for governments, including India. The police-population ratio in India is 153 per 100,000 people. That's below the UN-advised 222 per 100,000 people. This reduction in gaps and increased resource distribution efficiency are part of the motivation for including technology in law enforcement.
The application of AI-led law enforcement agencies ranges from counter-terrorism to crowd control. In Uttar Pradesh use, the use of AI-powered drones and CCTVs was convenient for tracking people and managing traffic at large gatherings like Kumbh Mela. This illustrates the widespread adoption of technologies that increase the accuracy of criminal forensic medicine and are reflected in unique applications such as fingerprint storage and digitization. Additionally, modern tools developed by central agencies such as the Police Research and Development Agency (BPR&D) delve into spaces like depth and dark webs to measure sentiment and provide LEA with reliable intelligence input.
India has also sought to curb new age cybercrime, such as online money laundering. The Enforcement Bureau uses the Financial Intelligence Unit (FIU) advanced analytical AI/ML tools to detect suspicious financial patterns. Data analysis of FIU's Mule accounts helps prevent the routing of unissued funds in the form of virtual digital assets.
However, AI systems can be overwhelming and offer suboptimal performance. Despite having 275 AI CCTVs, Rath Yatra of Puri witnessed the end of three pilgrims. Technical contradictions such as false positives (targeting people with dark skin tones) raise additional concerns in diverse countries like India, as documented in countries like the United States. Such instances raise important questions regarding accountability and governance considerations for technology service providers.
Governance considerations in AI deployments
The state's efforts to equip and modernize police were enhanced by the Centre via “Supporting the State and UT for modernizing police” (ASUMP) – expenditure of 4,846 crore over five years from 2021 to 26. Delhi and Tamil Nadu are said to have adopted “Innsight.” It is an AI tool for data analysis developed by Innefu Labs, which has been subject to cyberattacks and data breaches due to its weak security structure. Such cases suggest the need for a private company due diligence framework in securing contracts, strengthening the need for due process and strengthening the need for proper testing prior to deployment. AI tools promise efficiency, but their deployment must involve explanatory mechanisms to ensure accountability to counter the occurrence of required feedback loops and opaque behavior.
The governance framework for AI deployment in law enforcement must explain the positive potential of bias, discrimination, falsehood, and raise issues of responsibility and accountability. This requires adjusting the passing test and operational use of legality, necessity and proportionality, particularly within the context of fragmented regulation situations of biometric data. Without this framework, it would be difficult for governments to gain the trust of the public, especially in the age of social profiling.
LEAS can use this to quickly scan large amounts of data, build better prevention mechanisms for new age crime such as cyberattacks, and efficiently deal with resource allocations, leading to aggressive policing.
However, protection and regulatory frameworks must complement skills and capabilities on the ground. Cases of the use of AI generated to write police reports in the United States highlight how reports generated to AI missed the contextual specificity of a particular jurisdiction and legal nuances of police practices.
India's current legal and operational framework must carefully incorporate protections and standards to close these institutional and technological divisions.
The integration of technology in law enforcement has great potential to increase executive productivity and streamline the process of arresting criminals. LEAS can use this to quickly scan large amounts of data, build better prevention mechanisms for new age crime such as cyberattacks, and efficiently deal with resource allocations, leading to aggressive policing. However, liability for accidents during the deployment of technology in law enforcement is primarily based on the Indian governing body.
Evolution and evaluation
To keep up with the pace of technology growth and use AI responsibly, LEAS must embrace the evaluation mechanisms that come with technology evolution.
Companies must undergo regular external algorithm audits and receive compliance certification from the auditing body. Additionally, to qualify for procurement through the LEA, you need to communicate clear compliance mechanisms to your company. The role of the Artificial Intelligence Safety Institute in the development of robust safety and ethical testing standards suitable for the Indian context will be an important policy initiative to further implement it.
Pilot programs must be mandatory to determine the actual impact of AI and assess the same for risk parameters within a particular context. The proposal to establish an “incident database” that creates collections of risks and builds harm reduction mechanisms associated with the deployment of AI tools can help us recognize the evolving nature of harm. All of these proposed mechanisms require deep collaboration between public bodies and private institutions, particularly the developers of these new technologies.
The role of the Artificial Intelligence Safety Institute in the development of robust safety and ethical testing standards suitable for the Indian context will be an important policy initiative to further implement it.
Additionally, sensitization efforts are required within LEAS to train field officials regarding the responsible use of AI, including knowledge of potential risks and harms. Some AI surveillance systems can detect inconsistencies in police activities by identifying bias and disproportionate patterns of force among police officers and supporting internal performance assessments. BPR&D can set up dedicated AI and technical training modules for police verticals and play an important role in this context.
Thoughtful regulations are the cornerstone of developing effective governance frameworks to ensure the development of safe and reliable systems in diverse countries like India, along with human surveillance.
Srijan Jha is a research intern at the Observer Research Foundation
The views expressed above belong to the author. ORF research and analysis is now available on Telegram! Click here to access curated content (blogs, longforms, interviews).
