Public transport agencies are under pressure to keep riders safe, manage growing networks, and deliver a smoother journey, often with constrained budgets and staff. Artificial intelligence (AI) and computer vision are helping to close that gap.
What began as a way to reduce fare evasion at gates is now reshaping how agencies think about safety, security, and customer experience. Platforms such as Cubic’s FLARE (Fare Loss Avoidance Reporting Engine) demonstrate that the same technology that spots a tailgater can also help prevent serious incidents, support vulnerable riders, and keep crowds moving, while protecting passenger privacy.
In simple terms, computer vision is software that can “understand” what is happening in a video feed. Combined with AI, it can detect patterns, highlight anomalies, and send real-time alerts without a human watching every camera.
From fare evasion to safer networks
FEnX fare gates, for example, use computer vision to recognise behaviours such as tailgating or pushing through. The system can tell the difference between someone deliberately slipping through and a parent helping a child, then generate real time alerts and reports for staff.
The immediate benefit is clear, agencies can reduce revenue loss and act on hotspots rather than guessing. More importantly, clamping down on unauthorised entries reduces the risk of associated incidents, such as people entering restricted areas or moving into unsafe parts of the platform, for example beyond safety lines or close to the tracks.
Many agencies are now using the same tools to train them to recognise a broader range of behaviours of concern. Instead of focusing only on fare evasion, AI models can flag fights, harassment, vandalism, trespassing, and people entering tunnels and track areas. In Seoul, for example, the metro is developing AI driven surveillance that spots unusual or dangerous activity and automatically notifies staff so they can intervene quickly. The intent is not just to enforce rules, but also to reassure riders that the system is being actively monitored at all hours.
Supporting vulnerable riders and inclusive journeys
Singapore’s Bukit Panjang LRT line is a good example. Its “iSafe” system monitors platforms for passengers on the tracks or in danger zones, sending instant alerts and triggering announcements that remind everyone to stand back. On other lines, trials are underway that detect wheelchairs, prams, or unattended objects and notify staff to assist or investigate. Instead of a generic alarm, staff get specific information about who might need help and where.
These kinds of capabilities directly support riders with disabilities, older passengers, and anyone who needs extra time or space to travel. When people know the system is actively looking out for them, public transport feels safer and more welcoming.
Smoother everyday journeys through better data
- Operational response in real time: for example, adjusting gate direction, redeploying staff, or managing platform access when a station is becoming overcrowded.
- Planning and design: for example, understanding where bottlenecks occur, how long transfers really take, and which platforms or entrances are under pressure during big events.
Some agencies are already experimenting with live crowding forecasts, using AI to predict where and when a station will hit certain thresholds, then nudging interventions before the situation becomes unsafe. That might be as simple as targeted announcements or as complex as automatically adjusting train dispatch patterns.
For buses and light rail, which often do not have fare gates, the same principles apply. Cameras on board or at stops can count passengers, detect falls or altercations, and flag when an emergency button is pressed but the driver cannot see what is happening. In Ottawa, OC Transpo’s network of thousands of cameras is being paired with AI tools that can detect anomalies and help special constables respond faster, rather than relying on delayed reports.
Over time, these data sets can also support more equitable service planning. For example, understanding where and when wheelchair users, stroller users or large groups typically board can help agencies match the right vehicle types and stop designs to real demand. The same insights can be fed into early station planning so designers can test and refine new layouts before they are built, rather than discovering problems after opening day.
Global proof points
Dubai, the Roads and Transport Authority has equipped inspection vehicles with AI powered cameras to patrol Metro and Tram corridors. These vehicles continuously scan the right of way for trespassers, infrastructure damage, or other violations. By spotting issues early, the agency can intervene before passenger services are affected, and staff can spend more time solving problems rather than hunting for them.
In Chicago, the transit authority is piloting AI based track intrusion detection at key stations. When a person or object is detected on the tracks, alerts go straight to the control centre, allowing operators to slow or stop trains and send help. With thousands of track intrusions reported each year, even a modest reduction can save lives and reduce knock on delays.
On another major metropolitan rail network, AI analysis across a large CCTV estate has helped staff detect tunnel trespassers in real time and intervene before anyone is harmed. This is particularly important on networks that experience frequent trespassing and the tragic incidents that can result.
In all these cases, AI is not replacing humans. It is extending their reach, acting as a tireless watcher that highlights the few seconds that really matter in thousands of hours of footage.
Privacy, equity, and building public trust
Any discussion about AI and video in public spaces must deal directly with privacy and equity. Riders need to feel that the system exists to protect them, not to profile them.
Modern transit AI platforms are therefore being designed with privacy as a core requirement, not an afterthought. Systems such as FLARE focus on patterns of behaviour rather than personal identity. They analyse movements, posture, and context, not faces. They avoid storing biometric data and aim to use anonymised information in aggregate, for example to identify fare evasion hotspots or crowding patterns, rather than to track individuals across a network.
Most deployments also keep a human firmly in the loop. An AI alert is treated as a prompt, not a verdict. Trained staff review the event, decide whether to act, and apply judgement that an algorithm cannot. That balance reduces the risk of overreaction and helps prevent biased outcomes.
Agencies are also exploring how AI insights can be used constructively. Fare evasion analytics can inform better station design, clearer signage, or outreach campaigns, not only more fines. Behaviour detection can be paired with improvements in lighting, staffing, or service, not only enforcement.
Transparency matters as well. Public communication about what the systems do, what data they use, and how they are governed helps to counter misconceptions. Sharing examples where AI enabled staff to help a vulnerable rider or prevent harm is often more powerful than any technical specification.
Looking ahead, AI as a standard safety layer
As AI matures, it is likely to become as standard in transit environments as CCTV is today, with a much higher return on investment. The focus will shift from standalone pilots to integrated safety and analytics platforms that span gates, open station entrances, platforms, tunnels, and vehicles.
Systems like FLARE point to that future. By combining advanced image analysis, edge computing, and configurable rules, they can adapt to both gated and ungated environments. The same core platform can monitor a dense metro station one day and a light rail stop the next, feeding relevant alerts to local staff and high-level insights to central teams.
A key strength of AI is its ability to learn. When new fare evasion tactics emerge, passenger behaviour changes, or agencies introduce new policies, models can be retrained and redeployed without ripping out hardware. Over time, the system becomes better at distinguishing genuine risk from everyday behaviour, thereby reducing false alarms and building trust.
For public transport leaders, the opportunity is to treat AI and computer vision not as buzzwords but as practical tools within a broader strategy for safer, more inclusive, and more efficient networks. Used well, these technologies help ensure that:
- People pay their fares fairly.
- Staff have the information they need to respond early and confidently.
- Vulnerable riders feel seen, supported, and safe.
- Everyday journeys are smoother, with less crowding and fewer surprises.
Ultimately, the promise of AI in public transport is simple. If something goes wrong, help arrives faster. If a pattern of risk emerges, it is fixed before it becomes a headline. And for most people, most of the time, the journey feels easier. That is how technology earns its place in the fabric of a transit system, by quietly making every trip a little safer and more human.