Public Safety Technology — Surveillance, Predictive Policing & Emergency Response in Rio

Updated March 2026

Public safety in a metropolitan area of 6.7 million residents demands technology that operates at scale, responds in seconds, and integrates data from disparate sources into actionable intelligence. Rio de Janeiro’s public safety technology ecosystem encompasses 10,000 surveillance cameras (40 percent equipped with facial recognition), the CIVITAS AI system with 900 AI radars and 50 license-plate recognition cameras, the largest video wall in Latin America spanning 104 square meters on 125 screens, and a network of 9,000 georeferenced sensors that provide environmental awareness for disaster preparedness. All of these systems converge at the Centro de Operacoes e Resiliencia (COR), where 500 professionals from 50 integrated government agencies coordinate responses to approximately 1,200 occurrences per month across the city.

The Camera Network: From 600 to 10,000

The evolution of Rio’s surveillance camera network tracks the city’s broader smart city maturation. COR launched in December 2010 with 600 cameras, a number sufficient to monitor key intersections, government buildings, and a handful of public spaces but far from comprehensive urban coverage. By 2015, the network had grown to 1,000 cameras integrated with 15,000 sensors, providing meaningful coverage of major corridors and central business areas.

The transformation came with the 2022-2024 expansion, funded through the Luz Maravilha PPP for public lighting, which set a target of 10,000 cameras. This expansion did not simply add more of the same cameras — it introduced qualitatively different capabilities. Approximately 4,000 of the new cameras (40 percent of the total) are equipped with facial recognition technology, enabling real-time identification of persons of interest against watchlist databases. The remaining cameras provide high-definition video surveillance at resolutions sufficient for forensic analysis — identifying vehicle makes and models, reading signage, and capturing behavioral details that support investigation after incidents.

The camera placement strategy reflects a risk-based approach. The highest density of cameras, including the majority of facial recognition units, is concentrated in areas with the highest security profiles: Copacabana and Ipanema beaches (which attract millions of visitors annually), the Centro business district, major transit hubs including Central do Brasil railway station and Rodoviaria Novo Rio bus terminal, and the approaches to critical infrastructure including hospitals, power substations, and government buildings. Secondary density coverage extends to residential commercial corridors in the Zona Norte and Zona Oeste, while basic coverage reaches suburban areas and major roadways.

Facial Recognition: Capabilities and Controversies

The deployment of approximately 4,000 facial recognition cameras across Rio represents one of the largest such deployments in Latin America. The technology works by comparing faces captured by cameras against databases of persons of interest — individuals with outstanding warrants, missing persons, and persons associated with active investigations. When the system identifies a potential match, an alert is generated on COR’s video wall for operator verification before any response is initiated.

The operational workflow is designed with a human-in-the-loop verification step: the AI system identifies potential matches, but a trained COR operator reviews each match before dispatch decisions are made. This design reduces the risk of false positive responses while maintaining the speed advantage that automated identification provides over manual monitoring of thousands of camera feeds. The 125-screen video wall at COR provides the display capacity needed for operators to simultaneously monitor automated alerts while maintaining manual oversight of cameras in high-priority locations.

The facial recognition deployment operates within the framework of Brazil’s General Data Protection Law (LGPD), which classifies biometric data including facial imagery as sensitive personal data subject to enhanced protections. The city has implemented data retention limits that automatically purge facial recognition data after defined periods unless it is associated with an active investigation, access controls that restrict who can query the facial recognition system and under what circumstances, and audit logging that creates a record of every facial recognition query for compliance review.

Civil liberties advocates have raised legitimate concerns about the potential for facial recognition technology to disproportionately impact certain communities, to generate false positives at different rates across demographic groups, and to enable surveillance beyond the security use cases that justify the technology. These concerns have prompted ongoing dialogue between the city government, civil society organizations, and technology providers about appropriate governance frameworks for biometric surveillance in public spaces.

CIVITAS: AI-Powered Vehicle Security

The CIVITAS AI traffic system contributes critical capabilities to Rio’s public safety infrastructure through its 900 AI radars and 50 license-plate recognition cameras. While CIVITAS serves dual purposes in both traffic management and security, its AI-powered mapping of stolen vehicle routes in real time represents a public safety application that has direct, measurable impact on crime reduction.

When a vehicle is reported stolen, its license plate is flagged in the CIVITAS database. The 50 LPR cameras at strategic chokepoints scan plates continuously, and upon detecting a match, the system triggers an automated alert cascade. What makes CIVITAS distinctive is the AI prediction layer: rather than simply alerting when a stolen vehicle passes a camera, the system predicts the vehicle’s likely route based on historical traffic data, criminal movement patterns, and real-time road conditions. This predictive capability allows patrol units to be positioned for interception rather than pursuit, a safer approach for both officers and bystanders.

The 900 AI radars supplement the LPR cameras by providing continuous monitoring of traffic behavior patterns that may indicate criminal activity. Vehicles driving erratically, making unusual U-turns near police checkpoints, or exhibiting speeds inconsistent with normal traffic flow generate behavioral flags that COR operators can investigate through camera feeds. This behavioral analytics capability extends the security network beyond the database-driven approach of LPR matching to pattern-based detection that can identify suspicious activity even when specific vehicle identities are not flagged.

Emergency Response Coordination

COR’s role as an emergency response coordination center extends beyond routine security incidents to encompass natural disasters, public health emergencies, major accidents, and crowd management events. The facility’s integration of 50 government agencies — including fire departments, police forces, ambulance services, civil defense, environmental agencies, and public utility companies — enables unified responses that avoid the communication failures and jurisdictional conflicts that characterized emergency management before COR’s creation.

The emergency response workflow follows a structured protocol. COR’s sensors, cameras, citizen reports through 1746, and social media monitoring detect an emerging situation. Operators assess the situation using camera feeds and sensor data, categorize it by type and severity, and activate the appropriate response plan. Relevant agencies receive simultaneous notifications through COR’s communication systems, with response coordination managed from the situation room (expanded by 30 percent during the 2022-2024 renovation) where agency representatives can work side by side.

Flood emergencies, which pose the greatest natural disaster risk in Rio, benefit from the most developed response protocols. The 9,000 sensor network provides continuous monitoring of rainfall, drainage system capacity, and hillside stability. When sensor readings cross alert thresholds, COR activates a graduated response: initial warnings to residents in affected areas through social media and SMS, pre-positioning of pumping equipment and emergency vehicles, road closures on flood-prone routes, and activation of emergency shelters if displacement is anticipated. The 30-percent reduction in emergency response times that COR has achieved translates most directly into lives saved during these flood events, where minutes of warning time can mean the difference between safe evacuation and dangerous exposure to rising water.

Predictive Policing and Data-Driven Security

Beyond real-time surveillance and response, Rio’s public safety technology enables data-driven approaches to crime prevention. Historical data from COR’s archives — years of camera footage metadata, incident reports, CIVITAS vehicle tracking data, and 1746 citizen reports — feeds into analytical models that identify crime patterns, hot spots, and temporal trends.

Geospatial crime analysis maps incidents by type, location, and time of day, revealing concentrations that may indicate organized criminal activity or environmental conditions that facilitate crime. A cluster of vehicle thefts near a particular intersection might indicate a surveillance gap that can be addressed by repositioning cameras. A pattern of robberies along a specific bus route during certain hours might justify increased police presence during those times. A correlation between street lighting outages (detectable through the smart energy grid monitoring systems) and crime incidents might prioritize lighting repairs as a crime prevention measure.

The COR.Lab innovation laboratory supports research into more sophisticated predictive policing approaches, including machine learning models that forecast crime probability by location and time period based on historical patterns, weather conditions, event schedules, and economic indicators. These models are developed in partnership with university researchers from UFRJ and PUC-Rio, ensuring that they incorporate academic rigor and ethical considerations alongside operational utility.

The use of predictive policing technology is subject to ongoing ethical debate globally, and Rio’s implementation reflects this. The city’s approach emphasizes using predictive models to inform resource deployment decisions — where to station patrol units, where to conduct community engagement — rather than to target specific individuals or communities. The human-in-the-loop principle that governs facial recognition also applies to predictive policing: model outputs inform human decision-makers rather than triggering automated responses.

The COR Video Wall: Situational Awareness at Scale

The 125-screen video wall at COR, measuring 104 square meters in total display area, is not merely an impressive piece of technology — it is a functional tool that enables the kind of simultaneous multi-source monitoring that effective urban security requires. The video wall can display any combination of live camera feeds, data visualizations, maps with sensor overlays, weather radar, social media dashboards, and statistical charts. Operators can quickly reconfigure the display to focus on a developing situation, pulling up relevant camera feeds alongside sensor data and map overlays to build a comprehensive picture of the event.

During major events like Carnival, the video wall becomes the primary command tool for managing crowd safety across multiple celebration venues simultaneously. Different sections of the wall display camera feeds from each venue, real-time crowd density estimates, ambulance and fire department positioning, traffic conditions on access and egress routes, and weather forecasts that could affect outdoor celebrations. The ability to see all of this information simultaneously, rather than switching between screens or dashboards, gives COR operators the kind of holistic situational awareness that allows proactive management rather than reactive response.

The physical design of the operations room — 446 square meters with tiered seating facing the video wall — follows the design principles of military command centers, where operators need unobstructed sight lines to shared displays while maintaining access to individual workstation screens for detailed analysis. This design enables a two-level workflow: operators monitor their individual assigned feeds and data sources on personal screens, while supervisors scan the video wall for cross-domain patterns that might not be visible to operators focused on specific data streams.

Integration With Broader Smart City Systems

Public safety technology does not exist in isolation from other smart city systems — it depends on and contributes to the broader ecosystem. The 5G infrastructure pilots provide the connectivity bandwidth needed for 4K camera feeds and real-time AI analytics at the network edge. The IoT sensor network provides environmental context — a surge in water level sensors during a rainstorm informs the emergency response context for any incidents that occur during the event. The smart mobility systems provide transportation awareness that supports both police patrol routing and emergency vehicle dispatch.

The DATA.RIO open data portal publishes aggregate safety data that supports public awareness and academic research. Crime statistics by neighborhood, emergency response time metrics, and flood warning effectiveness data are available through the portal’s API, enabling journalists, researchers, and advocacy organizations to analyze public safety trends and hold the city government accountable for security outcomes.

The planned Rio AI City hyperscale data center campus will provide the computing resources needed for next-generation public safety AI. Current AI capabilities — CIVITAS vehicle tracking, facial recognition matching, behavioral analytics — operate within the constraints of COR’s 84-server data center. The orders-of-magnitude increase in computing capacity available through Elea’s 3.2 GW campus will enable more sophisticated models: real-time analysis of thousands of camera feeds using computer vision AI, natural language processing of social media streams to detect emerging threats, and simulation-based scenario planning for major events and disaster preparedness.

Governance and Accountability

The concentration of surveillance capabilities — 10,000 cameras, facial recognition, license plate tracking, predictive analytics — in a single operations center creates a surveillance infrastructure that requires robust governance to maintain public trust and prevent misuse. Rio has implemented several accountability mechanisms: the ABNT standardization guidelines published in June 2024 include governance provisions for operations centers, LGPD compliance requirements apply to all personal data processing including biometric data, and COR’s integration of 50 government agencies creates inherent checks through multi-agency oversight.

The ongoing challenge is ensuring that governance mechanisms keep pace with technological capabilities. As camera resolution increases, AI becomes more capable, and data integration becomes more comprehensive, the potential for both beneficial use and misuse grows. The dialogue between the city government, civil society, academic institutions, and the public about appropriate boundaries for surveillance technology will continue to evolve as the technology itself advances.

External Resources

• COR Official — Operations and Safety — Real-time operations data and public safety alerts
• Hexagon — Rio de Janeiro Case Study — Technology platform details