AI on the Rails: Transforming Mass Transit Security

In October 2025, two girls, ages 12 and 13, were killed while subway surfing—riding on top of train cars—in New York City. Their deaths sparked renewed attention to safety in urban transit systems. As agencies search for better ways to prevent such tragedies, many are looking to artificial intelligence to do what cameras, cops, and conductors alone cannot: detect danger before it turns fatal.

AI is beginning to transform how mass-transit networks operate and protect the public. From automated threat detection and crowd management to predictive maintenance and drone-enabled emergency response, the technology is redefining both the perimeter and purpose of transit security.

Now Boarding

Despite the hype, most transit authorities have barely left the station. “Everyone wants to use AI, but budget is an issue,” says Armando Leon, CPP, senior manager for security and infrastructure protection with the Metro Transit Police Department at WMATA. Leon notes that Washington Metro is consolidating its multiple camera and access-control systems to add analytics capable of recognizing suspicious packages, barrier crossings, and other anomalies.

“We’re still figuring out how best to integrate these tools,” he explains, echoing the reality that few U.S. systems have moved beyond pilots or proofs of concept. Many agencies need to unify legacy camera networks and update infrastructure before they can leverage intelligent analytics. Yet even in these early stages, AI is beginning to reshape what “security” means in public transportation.

AI as a Security Partner

The largest public transit agency in the United States is already exploring that future. At a public hearing in early 2025, Michael Kemper, the Metropolitan Transportation Authority's chief security officer, said the MTA is evaluating AI software to detect potentially dangerous behavior on subways in real time. The system would analyze data from the city’s thousands of cameras, now covering nearly every platform and train car, to sense aggression, trespassing, or risky movement on platforms. The goal is to alert police or guards before incidents escalate, improving response times and reducing danger for riders.

Kemper emphasized that facial recognition is not part of the plan, but behavioral pattern recognition could help spot subway surfers or others engaged in hazardous activity before tragedy strikes. His remarks underscore how transit agencies are beginning to view AI not as a futuristic luxury but as a practical safety tool that augments human vigilance.

From Cameras to Cognition

Transit networks already have extensive surveillance infrastructure. What’s changing is how that video is used. Instead of passively recording incidents for later review, AI platforms now scan every frame for behavioral or environmental cues, such as movement against the flow of the crowd, a bag left unattended, or a trespasser crossing tracks.

At WMATA, Leon’s team is exploring ways to track officer patrols across stations and trains without relying on QR codes or manual check-ins. “It’s about accountability, visibility, and safety,” he says.

Companies like Alpha Vision are pushing the envelope with “agentic” video analytics that can be tasked in plain language to detect specific behaviors such as running, climbing, or carrying oversized objects. “It’s the idea of having a security guard living inside the camera itself,” explains Brandon Krallis, a company vice president. Its systems can issue voice warnings in any tone or language, operate entirely on the edge, and integrate with access control or incident management systems, all while running 24/7 on any connected camera.

Meanwhile, Ambient.ai, partnering with Stone Security, helped a statewide U.S. transit agency cut incident response times by more than 50%. The platform flags falls, trespassing, and loitering that previously went unnoticed.

Infrastructure and Proactive Security

AI is moving beyond control rooms and into the infrastructure itself. The fusion of AI analytics with IoT sensors is transforming vehicles and stations into intelligent assets that monitor themselves in real time, including detecting faults, anomalies, and potential threats before they escalate.

In Quebec, exo, the regional transit agency serving Greater Montréal, uses predictive analytics and digital-twin modeling to identify service disruptions and risks before they occur. In Singapore, more than 5,500 buses are equipped with Streamax driver-assistance systems that detect fatigue, blind spots, and near-collisions. Not only do such tools reduce downtime, but they also make transit safer for passengers and workers alike.

The New York MTA, through its Transit Tech Lab, has deployed tunnel-intrusion and fare-evasion detection algorithms and integrated them into command operations. When commercial tools fall short, the MTA partners with Rutgers University to develop custom AI frameworks that identify trespassing and analyze behavioral patterns, turning raw footage into actionable safety intelligence.

Hayden AI has deployed bus-mounted computer-vision systems in New York and Washington, D.C., to identify bus-lane and parking violations in real time. The result has been faster service, reduced congestion, and safer passenger boarding zones—illustrating how AI enforcement also serves public safety.

In Latin America, the Lima Metro’s Line 2 project illustrates how AI and advanced analytics are being embedded into new transit infrastructure from the ground up. Herbert Calderón, Director de Seguridad Patrimonial at Consorcio Constructor Metro 2 Lima (CCM2L), is leading efforts to integrate AI-based monitoring, video analytics, and digital twin modeling into both construction and operational planning. The system combines feeds from cameras, ticketing, and environmental sensors to manage crowd density, detect irregular activity, and anticipate safety risks before they escalate. For a rapidly expanding network that will eventually serve more than half a million riders per day, this fusion of automation and situational awareness is already improving coordination between engineering, operations, and security teams.

Across these examples, AI’s role in infrastructure management is blurring the line between operations and security, creating a single, data-rich environment for resilience.

Weapons Detection and Threat Prevention

Perhaps the most visible frontier in AI security is weapons detection. The Regional Transportation Commission of Southern Nevada, which oversees Las Vegas transit, became the first U.S. system to implement network-wide AI scanning for visible firearms using ZeroEyes software. Its 400-plus buses are monitored in real time, with alerts verified by trained analysts and transmitted to dispatch within seconds. That’s critical for active-assailant events.

AI-driven acoustic sensors can complement visual detection. At Purdue University Northwest, researchers have found that acoustic algorithms identify gunfire with 99% accuracy, which is particularly valuable in enclosed or high-traffic transit hubs. Together, image- and sound-based systems are helping transit agencies move from reactive response to proactive risk sensing.

Airborne AI

Unmanned aerial systems are joining the security arsenal. Several agencies now deploy AI-enabled drones to autonomously scan rail lines, inspect bridges, or assess tunnel damage after an incident.

“AI-enabled drones represent a profound shift in how we protect and manage critical transportation networks,” says Peter Lambrinakos, O.O.M., CPP, distinguished fellow at the University of Ottawa and strategic advisor to Draganfly Inc. “They are not about replacing human responders but about augmenting their capabilities. This technology extends our reach, giving teams safer, faster, and smarter ways to assess risk, inspect infrastructure, and act decisively. The real opportunity lies in using this proactive intelligence responsibly, with strong governance, to enhance security and build lasting public trust.”

In the U.K., Network Rail has used AI-guided drones to inspect track corridors for structural defects or trespassing. At the same time, U.S. rail operator CSX employs autonomous drones to scan yards and switches for damage or tampering. Drones exemplify how AI is shifting security from static monitoring to dynamic, mobile intelligence.

Data and Decision Intelligence

AI’s effectiveness in mass transit depends increasingly on data flow and computing architecture. With the rise of onboard connectivity and edge processing, vehicles themselves are becoming rolling data centers. They analyze video, telemetry, and environmental inputs in real time without sending every byte to the cloud.

Edge computing reduces latency, safeguards privacy, and enables instant response during incidents. These “smart trains” and buses feed continuous safety and performance insights to operators, creating an invisible nervous system that underpins modern transit.

A 2024 study on AI and IoT in public transit found that AI and IoT together can predict component failures, detect security anomalies, and even forecast demand surges. It’s a convergence of operational intelligence and situational awareness.

Operational Analytics and Predictive Performance

While safety often drives adoption, efficiency is what funds it. AI’s ability to optimize scheduling, routing, and maintenance can yield measurable savings while indirectly strengthening security.

For example, Via TransitTech's 2025 Via Intelligence platform uses digital-twin modeling to simulate and test routes before launch. In Fort Worth, Texas, the technology improved paratransit efficiency by 13% and cut excessive trip times by 86%. In Barcelona, predictive ventilation controls now balance air quality, energy use, and comfort in real time, enhancing both passenger safety and health.

Cubic Transportation Systems has deployed similar tools to detect fare evasion, optimize bus frequency, and monitor for fraud in open-loop payment systems. Cogent Infotech advises agencies on integrating such analytics across budgeting, compliance, and customer experience functions. These are all part of an emerging decision-intelligence ecosystem where security, operations, and service quality are deeply intertwined.

AI for Emergency Management

AI is proving equally powerful in crisis response. Real-time analytics can detect a collapsed passenger, an unauthorized track crossing, or an escalating altercation, prompting immediate alerts to police or medical personnel.

In several Ambient AI-powered systems, response times have been cut in half. Elsewhere, drones and mobile cameras now stream live imagery to command centers during floods, fires, or derailments. The fusion of AI detection, human verification, and rapid dispatch is replacing decades-old passive monitoring models.

Such advances point toward “predictive prevention,” a shift from reacting to incidents to anticipating and interrupting them before harm occurs.

Balancing Innovation and Public Trust

As AI reshapes security, it also challenges traditional governance. The benefits of visibility and efficiency come with concerns about privacy, bias, and workforce impact. Over-monitoring, opaque algorithms, and data misuse can erode public confidence just as easily as they enhance safety.

Experts stress that AI must augment, not replace, human decision-making. “Technology can help us see more, but leadership determines what we do with what we see,” says Lambrinakos. Responsible adoption hinges on transparency, data protection, and accountability frameworks like the EU AI Act, ISO/IEC 42001, and the U.S. Blueprint for an AI Bill of Rights.

Labor impacts are another dimension. Rather than displacing staff, AI is likely to elevate roles, shifting personnel from passive observation to active incident management and analysis. Still, the human element remains irreplaceable in empathy, discretion, and judgment.

The Road Ahead

For now, AI in transit security remains unevenly distributed. It’s accelerating in some cities and idling in others. As WMATA’s Leon notes, progress depends as much on procurement and budgets as on technology itself. But the direction is unmistakable.

From predictive analytics and sensor fusion to drone reconnaissance and weapons detection, AI is redefining how transit systems see, think, and respond. The subway-surfing deaths that opened this story are a reminder of what’s at stake. Cameras recorded what happened, but by the time the footage was reviewed, two young lives were gone.

If future systems can interpret those images fast enough to trigger a warning, then AI will have fulfilled its most important mission: tuning visibility into safety. And maybe one day, when a child boards a train or bus, her or his safety will depend not on who happens to be watching but on what the system has learned to see.