Maharashtra’s MARVEL: India’s First AI‑Backed Law‑Enforcement Entity 

Maharashtra’s MARVEL: India’s First AI-Backed Law-Enforcement Entity

In March 2024, the Government of Maharashtra formally approved the creation of a Special Purpose Vehicle (SPV), officially named Maharashtra Advanced Research and Vigilance for Enhanced Law Enforcement (MARVEL). This marks India’s first state-level, dedicated public authority to deploy artificial intelligence (AI) for law-enforcement support.

Formed by a tripartite agreement between the Maharashtra government, IIM Nagpur, and Pinaka Technologies Pvt Ltd, MARVEL is wholly government-owned. It was registered as a private limited company under the Companies Act, 2013. The Superintendent of Police, Nagpur (Rural), Harssh Poddar, serves as the ex-officio CEO and director, while the Director of IIM Nagpur is also an ex-officio board member.

The state cabinet allocated ₹23 crore as seed funding, with annual support of ₹4.2 crore guaranteed over the first five years, making MARVEL operational as an independent entity under the state Home Department.

Objectives, Mandate & Structure

The primary mission of MARVEL is to augment law enforcement capabilities through AI—specifically in data analytics, pattern recognition, predictive modelling, and automated investigative support—not to replace human officers. It is not an investigative agency per se; instead, it develops technological modules that various units of the Maharashtra Police can request for case-specific support.

Key objectives:

• Predicting crime hotspots and trends using large datasets
• Accelerating investigations in complex cases (e.g. missing persons, organised crime, cyber-fraud, hit-and-runs)
• Automating routine investigative tasks—such as writing request letters, scanning bank data, identifying suspicious transactions
• Enhancing facial/object recognition from CCTV, license plate/OCR reading, telecom record analysis, and relationship mapping

Under its structure:

• The SPV serves all state police units upon receipt of problem statements
• Teams at MARVEL design customised toolkits (like Crime-OS, facial recognition, object detection, etc.), deliver them, and train police personnel to operate them

Early pilot deployments involve about 25 experts, one police sub-inspector and one head constable per station trained on operating the tools.

MARVEL in Action: Tools and Use Cases

1. Crime-OS: AI Co-Pilot for Investigations

One of MARVEL’s flagship AI platforms is Crime-OS. Launched by May 2025, it works like an AI assistant: when an FIR is uploaded, Crime-OS automatically parses it, generates official letters (for example to banks), scans transactional data, highlights suspicious elements, and summarizes findings for investigative officers.

This drastically reduces manual workload in fraud, narcotics, and complex financial investigations, allowing human officers to focus on action and strategy rather than data sifting.

2. Face Recognition & Object Detection

MARVEL has also developed an AI model that is capable of real-time identification of persons of interest, missing persons, or suspects by scanning CCTV frames. Object detection AI can identify and flag dangerous items such as firearms or knives in public footage. These tools can detect over 600 object categories and immediately alert authorities when any are identified.

3. Vehicle Plate & Document OCR

Automatic number-plate recognition aids investigations in hit-and-run or stolen vehicle cases. Similarly, optical character recognition (OCR) tools parse documents—FIRs, statements, registries—cutting investigation time significantly.

4. Telecommunication & Call-Record Chatbot

MARVEL’s telecom-analysis tool rapidly processes CDR, IPDR, and mobile tower records, building timelines and communication maps. This chatbot interface simplifies otherwise laborious manual telecom analysis.

5. Predictive Analytics & Pattern Recognition

MARVEL modules analyze historical and ongoing crime data to predict likely crime scenes, identify gang structures, map organised networks, and detect patterns that human analysts might not spot on their own. These insights guide resource allocation and preventive policing.

Governance, Oversight & Scale

Governance Model

MARVEL is accountable to the Maharashtra Home Department, with formal administrative approvals dated 16 March 2024. Oversight is provided through a board comprising government-nominated directors, including ex-officio members from IIM Nagpur and the state police.

Its SPV structure legally enables collaboration with academia and private technology, while retaining full capital ownership in government hands. Module development is legally and operationally managed through this vehicle.

Ethical, Privacy & Legal Safeguards

Although formal data governance policies are still in development, MARVEL emphasises human oversight: AI modules augment, not replace, decision-making. Interpretive authority remains with trained officers inside the police organisation.

Given evolving concerns about misuse, further legal and procedural measures will likely be needed in coming years to ensure transparency, data-security, individual privacy, and judicial admissibility of AI-generated evidence.

Expansion of Scope

By mid-2025, MARVEL began adapting its tools beyond law enforcement. In July 2025, Maharashtra signed an MoU to deploy 3,150 AI-enabled cameras near tiger-reserve buffer zones (e.g. Tadoba, Pench, Navegaon-Nagzira). These create “AI walls”: detection poles alerting villages to approaching wildlife through sirens, mobile alerts, lights or sound—mitigating human-animal conflicts without harming animals.

MARVEL is also tasked with securing government data across offices through AI-based confidentiality enforcement, showing its pivot into broader public governance roles.

Legal & Scholarly Implications

From a legal services and policy standpoint, MARVEL opens several interesting avenues:

1. Legal Frameworks for AI in Policing

MARVEL’s tools operate at the intersection of criminal procedure, digital privacy, and public administration law. New rules under the Maharashtra Police Act, the IT Act, or state data-protection policies may be required to regulate use of AI-based surveillance, evidence-generation, and automated decisions.

2. Admissibility of AI-Generated Evidence

Courts may need to decide how to handle output from MARVEL tools—e.g. predictive risk zones, object detection captures, pattern-analysis visualizations—and whether they satisfy standards of reliability, chain of custody, expert testimony, or algorithmic transparency.

3. Openness and Reduction of Bias

It is essential to ensure algorithmic transparency and bias mitigation because biased datasets may cause the AI module to generate unfair profiling that disproportionately impacts particular communities. The system should incorporate algorithmic accountability, audits, and legal protections.

4. Litigation in the Public Interest (PIL)

PIL petitions pertaining to surveillance, data storage, improper use of CCTV and AI, and privacy rights may be drawn to MARVEL. In order to balance due process, individual rights, and enforcement, Maharashtra courts may be asked to establish the legal parameters for AI-driven investigative tools.

Evaluating Impact: Early Outcomes & Challenges

Positive Signals

Initial deployment in Nagpur Rural Police indicates tangible efficiency gains: Crime-OS automating FIR analysis, CDR scanning, letter drafting; facial/object recognition cutting time on suspect identification; predictive modules improving patrol planning. Training for ground-level inspectors is ongoing.

In the wildlife-conflict domain, the AI-camera system has already begun saving lives and calming tensions in forest-edge villages by providing advance alerts to residents and deterring wildlife safely.

Operational Challenges

• Data quality & integration: Maharashtra’s surveillance and crime database infrastructure varies widely—rural areas may lack consistent digital records.
• Capacity building: Training local officers in AI tools is resource-intensive; scalability across ~7000+ police stations is a major task.
• Legal ambiguity: Absence of clear rules on data retention, consent, third-party data access, algorithmic fairness, and accountability.
• Public trust: Transparency and communication will be critical to win public confidence in MARVEL’s methods and safeguards.

Recommendations & Path Forward

The following should be taken into account by academics, decision-makers, and legal professionals in light of the legal and public governance stakes:

• SOPs, or draft model rules: Central authorities or Maharashtra should establish regulations governing:
  • Limits on data collection, storage, access, and retention
  • Use of CCTV and telecom data in MARVEL tools
  • Committees or auditing systems for algorithms
• Algorithmic Audits and Oversight Boards: Independent review boards could check MARVEL tools for fairness, accuracy, and bias. They could then report back to the legislature or courts on a regular basis.
• Building Capacity in the Police and Judiciary: Academic institutions (like IIM Nagpur) and legal service authorities must work together to create training programs that teach officers how to read AI outputs and how courts interpret machine-generated evidence.
• Public Consultation and Transparency: Establishing credibility will be aided by unambiguous public documentation of MARVEL’s operations, data usage, and privacy protections. Public oversight could be facilitated by the publication of impact assessments and summary statistics.
• Replication & National Policy Input: Other states could use Maharashtra’s MARVEL as a model. The Ministry of Home Affairs or the National Legal Services Authority (NALSA) may take into account national pilot projects or model guidelines.

Conclusion

MARVEL represents a historic shift in Indian criminal justice: for the first time, a state has dedicated a government-owned, AI-enabled SPV specifically to support policing with predictive, analytical, and investigative tools. Its modular, case-driven model ensures adaptability, while academic partnership provides technical rigor. Initial deployments—Crime-OS and wildlife-surveillance pilots—point to real benefits.

However, legal and ethical frameworks must evolve rapidly to govern data privacy, algorithmic transparency, and evidence admissibility. Legal Service India readers and practitioners may find fertile ground in advising on or litigating around these emerging domains: from drafting rules and forensics guidelines, to courtroom uses of AI output, to challenging or defending algorithmic policing in PILs.

This initiative offers a pioneering case study: blending governance, technology, and criminal procedure. Maharashtra’s experience with MARVEL may well shape India’s approach to AI-augmented law enforcement—and the legal infrastructure needed to safeguard democratic values while leveraging technological efficiency. References:

• To fight crime using AI, Maharashtra Police create MARVEL, Indian Express (Jul. 18, 2024), https://indianexpress.com/article/cities/mumbai/to-fight-crime-using-ai-maharashtra-police-create-marvel-9460472 The Indian Express.
• India’s 1st AI Police Tech: Meet IPS Officer Behind MARVEL, Solving Crimes with Crime-OS, The Better India (Aug. 26, 2024), https://www.thebetterindia.com/362886/maharashtra-police-ai-tool-marvel-government-home-department-ips-officer-harssh-poddar The Better India.
• Maharashtra to receive AI support through ‘MARVEL’ to expeditiously solve crimes, AP7AM (July 6, 2024), https://www.ap7am.com/en/82798/maharashtra-to-receive-ai-support-through-marvel-to-expeditiously-solve-crimes .
• State allows government company to use AI for analyzing police data, Times of India (Aug. 8, 2024), https://timesofindia.indiatimes.com/city/mumbai/state-allows-government-company-to-use-ai-for-analyzing-police-data/articleshow/112357629.cms timesofindia.indiatimes.com.
• MARVEL, Maharashtra Home Department (Govt. of Maharashtra) press release/introduction (2025), https://home.maharashtra.gov.in/en/marvel/ home.maharashtra.gov.in.