🗞️ Why in News A government policy paper on AI integration in law enforcement has highlighted how machine learning, CCTV analytics, and predictive algorithms are transforming Indian policing from reactive crime response to evidence-based, predictive investigation. Pilot programmes are operational in Delhi (CCTV network with facial recognition) and Hyderabad (Hawk-Eye surveillance system and AI crime analytics). The policy paper calls for a national legal framework for AI-aided policing — balancing security effectiveness with privacy rights under the Digital Personal Data Protection (DPDP) Act, 2023 and due process guarantees under Articles 20, 21, and 22 of the Constitution.
The Case for AI in Policing
India’s Law Enforcement Challenges
| Challenge | Scale |
|---|---|
| Police-to-population ratio | ~145 per 1 lakh (vs. UN recommended 222) |
| FIR backlog | Millions of cases pending across states |
| Conviction rate | ~40-50% in IPC cases (lower for serious crimes) |
| CCTV coverage | Major cities have extensive networks but limited analytics capacity |
| Detective capacity | Shortage of trained investigators; forensic lab backlogs |
India’s police forces are understaffed, underfunded, and overwhelmed — making AI augmentation a systemic necessity rather than a luxury.
What AI Can Do in Policing
| Application | Function |
|---|---|
| Predictive crime mapping | ML algorithms analyse historical crime data to identify hotspots before crimes occur |
| Facial recognition | Match CCTV footage to criminal databases for suspect identification |
| Digital evidence management | AI-assisted processing of call records, financial transactions, social media |
| Automated FIR analysis | Pattern detection across FIRs to identify serial offenders/linked crimes |
| Traffic and crowd management | Real-time analytics for events, protests, disaster response |
| Cybercrime detection | Pattern recognition for financial fraud, phishing, and cyberstalking |
Current Deployments in India
Delhi — Integrated Command and Control Centre (ICCC)
Delhi Police operates an ICCC integrating:
- ~2.8 lakh CCTV cameras (target rollout across all 70 assembly constituencies) across the city
- Facial Recognition System (FRS) — used for identifying suspects, missing persons, and tracking
- AI-powered analytics for real-time traffic management and crowd density monitoring
Delhi’s FRS was controversially used during the 2020 Delhi riots for retroactive suspect identification — raising civil liberties concerns.
Hyderabad — Hawk-Eye System
Hyderabad City Police’s Hawk-Eye system:
- Integrates CCTV cameras across the city with centralised AI analytics
- Tracks vehicles via automatic number plate recognition (ANPR)
- Monitors crime hotspots in real-time
- Has a mobile application for beat officers to access real-time data
Hyderabad has also implemented SHe Teams — AI-assisted monitoring of public spaces for women’s safety — integrated into the surveillance network.
Other States
| State | AI/Tech Initiative |
|---|---|
| Tamil Nadu | AI-powered traffic management; CCTNS (Crime and Criminal Tracking Network System) |
| Maharashtra | CCTNS integration; drone surveillance for law and order |
| Uttar Pradesh | Dial 112 AI dispatch system; CCTV network in Lucknow, Noida |
| Kerala | Janamaithri Suraksha Project with digital policing |
National: CCTNS
The Crime and Criminal Tracking Network System (CCTNS) — a national initiative under Ministry of Home Affairs — digitises police records and FIRs across all police stations, creating a national database for AI analytics.
The Shift: Reactive to Predictive
Traditional (Reactive) Policing
Crime occurs → FIR filed → Investigation begins → Arrest (if successful)
- Resource-intensive
- Evidence-dependent
- High rates of cases going cold
AI-Augmented (Predictive) Policing
Data analysis → Hotspot identification → Preventive deployment → Incident prevention
AND
Crime occurs → AI pattern matching → Suspect identification → Faster arrest
Predictive Policing — How It Works
- Historical crime data ingestion — location, time, type, recidivism
- Socioeconomic overlay — unemployment, alcohol sales, weather patterns (controversial)
- Risk scoring — areas and individuals flagged for elevated risk
- Resource allocation — patrol deployment optimised to flagged areas
Legal and Ethical Concerns
Privacy — DPDP Act 2023
The Digital Personal Data Protection (DPDP) Act, 2023 — India’s primary data protection law — has implications for AI policing:
| DPDP Provision | Policing Implication |
|---|---|
| Consent for data processing | Police processing biometric/CCTV data without explicit consent |
| Data localisation | Centralised AI databases must comply |
| Data minimisation | Broad surveillance collection may violate this principle |
| Right to erasure | Citizens can request deletion — complex for criminal records |
Exemptions in DPDP Act: The Act explicitly exempts national security and law enforcement data processing from most consent requirements — giving police wide latitude.
Facial Recognition — The Accuracy Problem
Facial recognition systems have documented bias issues:
- Higher error rates for darker-skinned individuals, women, and elderly
- NIST (US) studies show some systems have 10-100x higher false positive rates for Black vs. White faces
- In India, concerns about accuracy for Dalit, tribal, and minority communities
- A false positive in policing = wrongful arrest and detention
Algorithmic Bias in Predictive Policing
Predictive algorithms trained on historical crime data can encode existing bias:
- If certain communities are historically over-policed, their areas appear in crime data more often
- Algorithm flags these areas → more policing → more arrests → data reinforces the bias
- Creates a feedback loop that perpetuates discriminatory policing
Constitutional Rights
| Article | Right | AI Policing Risk |
|---|---|---|
| Article 20 | Protection against self-incrimination | Compelled biometric data |
| Article 21 | Right to life and personal liberty | Pre-emptive detention based on algorithms |
| Article 22 | Protection against arbitrary arrest | Arrests on algorithmic suspicion |
The Supreme Court’s K.S. Puttaswamy v. Union of India (2017) judgment established Right to Privacy as a Fundamental Right — directly applicable to mass surveillance systems.
The Accountability Gap
Current AI policing deployments in India operate without a dedicated legal framework:
- No requirement to disclose when AI-generated intelligence was used for an arrest
- No audit mechanism for algorithm accuracy or bias
- No independent oversight body for law enforcement AI
- No right for suspects to challenge AI-generated evidence in court
The policy paper recommends:
- A Police AI Accountability Bill — transparency in AI deployment
- Algorithmic Impact Assessments before deployment
- Independent Ethics Committee for law enforcement AI
- Mandatory disclosure when AI evidence is used in prosecution
International Models
| Country | AI Policing Approach |
|---|---|
| UK | Facial recognition pilots; Information Commissioner oversight; court challenges |
| EU | AI Act (2024) prohibits real-time biometric surveillance in public (with narrow exceptions); post-hoc biometric systems classified as high-risk |
| USA | City-level bans (San Francisco, Boston) on facial recognition by police; federal regulation pending |
| China | Comprehensive AI surveillance; SCS (Social Credit System) integration |
The EU’s AI Act (2024) is the world’s most comprehensive AI regulation — it prohibits real-time biometric surveillance in public spaces except for narrow security exceptions.
UPSC Relevance
| Paper | Angle |
|---|---|
| GS3 — S&T | AI, machine learning, facial recognition, predictive analytics, surveillance technology |
| GS2 — Governance | Police reforms, CCTNS, MHA, law enforcement modernisation |
| GS2 — Polity | Articles 20, 21, 22; K.S. Puttaswamy judgment; privacy rights |
| GS4 — Ethics | Algorithmic bias, surveillance ethics, accountability in governance |
| Mains Keywords | Predictive policing, facial recognition, DPDP Act 2023, CCTNS, K.S. Puttaswamy, algorithmic bias, AI Act EU, surveillance |
Facts Corner
- India’s police ratio: ~145 per 1 lakh population (UN recommended: 222)
- CCTNS: Crime and Criminal Tracking Network System — national digital police records network; MHA
- Delhi ICCC: ~2 lakh CCTV cameras; Facial Recognition System (FRS) deployed
- Hyderabad Hawk-Eye: Integrated surveillance + ANPR + crime analytics
- DPDP Act 2023: India’s data protection law; exempts national security and law enforcement from most consent requirements
- K.S. Puttaswamy v. Union of India (2017): SC 9-judge bench; Right to Privacy = Fundamental Right under Article 21
- Facial recognition bias: NIST studies show 10-100x higher false positive rates for darker-skinned individuals in some systems
- EU AI Act (2024): Prohibits real-time biometric surveillance in public spaces (with narrow security exceptions); post-hoc biometric analysis classified as high-risk
- Predictive policing feedback loop: Historical over-policing of communities → algorithm flags those communities → more policing → reinforced bias
- Article 20: Protection against self-incrimination; Article 21: Right to life and liberty; Article 22: Protection against arbitrary arrest
