Efficiency, Surveillance, and the Law: Uttar Pradesh’s AI Turn in Policing

The AI Policing Stack: What Data YAKSH Actually Pulls Together

At its core, YAKSH is not a standalone mobile application but a state-level data integration layer for policing in Uttar Pradesh. The app consolidates multiple existing police datasets—many of which previously sat in separate systems or physical registers—into a single, searchable, beat-level interface.

According to officials, YAKSH draws on Crime and Criminal Tracking Network & Systems (CCTNS) data once a chargesheet is filed, ensuring that accused-level information automatically flows into the app without manual re-entry. This is supplemented by station-wise offender databases, history-sheeter records, gang charts, licensed firearm and ammunition registries, and beat-level verification reports prepared by local police personnel.

Crucially, YAKSH combines static data—such as criminal history, past FIRs, and court status—with dynamic inputs, including residence changes, movement alerts, and real-time verification updates. AI-enabled features like facial recognition, voice search, and gang-link analysis sit atop this consolidated dataset, allowing personnel to query across crime type, geography, associates, and patterns of repeat offending.

In practical terms, this means that what was earlier fragmented across paper beat books, separate databases, and supervisory memory is now algorithmically searchable and centrally visible, fundamentally altering how information circulates within the policing system.

Strengthening Beat Policing Through Data and Automation

Beat policing remains the backbone of law enforcement in Uttar Pradesh, where personnel are assigned specific localities and responsible for verifying offenders, tracking habitual criminals, and engaging with communities. YAKSH is designed to digitally reinforce this model, enabling beat personnel to instantly access criminal histories, monitor “history-sheeters” and wanted offenders, and prepare ranked lists of top offenders with greater transparency.

The app introduces automated alerts when an accused person changes residence, ensuring continuity of monitoring across jurisdictions. It also includes tools for licensed firearm and ammunition verification, colour-coded categorisation of accused persons, and a scoring mechanism based on the sensitivity of crimes, time of occurrence, and weapons used. Senior officials say this allows more focused deployment of resources and faster preventive action, especially against organised gangs and repeat offenders.

Predictive Policing by Another Name? Where YAKSH Sits on the Spectrum

Officials have described YAKSH primarily as a tool for identification, verification, and prevention, stopping short of calling it a predictive policing system. However, its functional design places it firmly along the predictive-policing spectrum, even if not at its most advanced end.

Predictive policing typically ranges from basic data aggregation to more advanced systems that generate risk scores, heat maps, and anticipatory alerts. YAKSH does not publicly claim to forecast crimes before they occur, but several of its features—crime sensitivity scoring, gang-link analysis, movement alerts, and categorisation of habitual offenders—enable forward-looking deployment decisions.

For example, when suspects are automatically scored based on crime severity, timing, and weapons used, or when gang networks are algorithmically mapped from FIRs and association data, policing shifts from reacting to individual incidents toward anticipating potential threats. Beat personnel can be directed to monitor specific individuals or areas not because of a new offence, but because the system flags elevated risk.

In this sense, YAKSH occupies a middle ground: it is more than a digital filing system, but short of fully autonomous prediction engines used in some global jurisdictions. Its significance lies in normalising algorithm-assisted suspicion within routine policing—laying the groundwork for more explicit predictive models should policy and technology evolve further.

A National Trend: AI Tools Spread Across Indian Law Enforcement

Uttar Pradesh’s move mirrors developments in other states and at the national level. As detailed by Indian Express, police forces are experimenting with AI systems that can analyse vast volumes of data—call records, CCTV footage, financial trails, and digital evidence—far faster than human investigators.

In Maharashtra, MahaCrimeOS AI, built with support from Microsoft, is being piloted across police stations to generate investigation plans, link cases, analyse digital evidence, and assist officers in complex crimes such as cyber fraud, narcotics, and financial offences. Delhi Police is expanding the use of AI-assisted facial recognition technology, while Bengaluru Police have deployed AI-enabled CCTV analytics to detect firecracker violations during festivals, reportedly flagging over 2,000 incidents during Diwali.

Legal Guardrails (or the Lack Thereof): How AI Policing Is Regulated in India

While YAKSH represents a significant technological leap in policing, its deployment sits within a largely under-specified legal framework governing the use of artificial intelligence by law-enforcement agencies in India. At present, there is no dedicated statute or binding regulation that specifically defines how AI tools may be used by the police, what safeguards must accompany them, or how individuals can challenge AI-assisted decisions.

India’s primary data-protection law, the Digital Personal Data Protection Act, 2023, provides broad exemptions for law-enforcement and public-order functions, allowing police agencies to process personal data without many of the consent, purpose-limitation, and notice requirements that apply to private entities. While these exemptions are designed to preserve operational effectiveness, they also mean that AI-driven systems like YAKSH operate with minimal statutory constraints on data use, retention, or secondary analysis.

Crucially, there is no explicit legal requirement for:

• Independent audits of AI systems used by the police
• Disclosure of how algorithms rank, score, or flag individuals
• Periodic accuracy testing or bias assessment
• Providing reasons or explanations when AI-assisted outputs influence policing actions

In practice, accountability is expected to flow through internal standard operating procedures (SOPs), executive instructions, and supervisory oversight, rather than enforceable rights available to citizens. If an individual is misidentified through facial recognition, incorrectly flagged through gang-link analysis, or subjected to enhanced monitoring based on algorithmic scoring, the primary remedies remain administrative complaints or judicial intervention, both of which can be slow and opaque.

The absence of clear legal guardrails becomes more consequential as AI tools move from post-incident investigationtoward preventive and anticipatory policing. Systems that influence patrol patterns, suspect prioritisation, or surveillance intensity can materially affect personal liberty, reputation, and freedom of movement—interests that ordinarily trigger heightened constitutional scrutiny.

In this context, YAKSH exemplifies a broader national pattern: technological adoption is racing ahead of legal governance. While courts have recognised privacy as a fundamental right and emphasised proportionality in state surveillance, these principles have yet to be translated into operational rules for AI-enabled policing. Until such frameworks emerge, systems like YAKSH will continue to rely on administrative restraint rather than legally enforceable safeguards to balance efficiency with civil liberties.

Benefits Promised — and Risks Flagged

Proponents argue that AI-enabled policing can improve efficiency, speed up investigations, and help overstretched forces respond to increasingly sophisticated crimes, especially online fraud and cybercrime. By automating pattern recognition and data correlation, such tools promise quicker decision-making and more consistent enforcement.

However, critics caution that AI systems trained on historical police data risk reinforcing existing biases, leading to disproportionate targeting of certain communities. Concerns also persist around accuracy, transparency, and accountability, particularly as India’s data-protection framework provides broad exemptions for law-enforcement agencies. The expanding use of facial recognition and predictive analytics in public spaces has raised questions about surveillance, due process, and citizens’ ability to challenge algorithmic decisions.

Why This Matters

The launch of YAKSH signals a decisive step in Uttar Pradesh’s push toward technology-driven policing, aligning the state with a national trend that sees AI as a force multiplier for law enforcement. Yet it also underscores a central tension confronting Indian policing: how to harness the speed and scale of artificial intelligence without undermining fairness, privacy, and public trust.

As more states deploy AI tools—from beat-level apps to predictive analytics—the challenge will lie not only in technological capability, but in clear legal safeguards, oversight, and transparency. How YAKSH is used on the ground—who it targets, how errors are corrected, and how accountability is enforced—will shape whether AI becomes a genuine aid to justice, or a new source of contention in India’s evolving policing landscape.