The cameras were always there. On street corners, in stadiums, at transit hubs, and across the venues where millions of people gather. For decades they recorded. They stored. They waited.
Today, what these cameras are doing now is fundamentally different. And the business being built around that shift is growing faster than most people realize.
Security cameras move from passive recording to real-time AI intelligence
Public safety infrastructure is undergoing a significant transformation as artificial intelligence becomes embedded into video surveillance systems. What were once passive recording tools are increasingly evolving into systems capable of interpreting activity, identifying anomalies, and supporting real-time decision-making as events unfold.
The market reflects that shift. The global market for AI in video surveillance is expected to more than double to $14.6 billion in the next four years, with use of the technology for video analytics projected to reach $69 billion by 2028, according to Lumana.
The driver behind those numbers is straightforward. Security teams monitoring hundreds of camera feeds simultaneously face what the industry calls monitoring fatigue, the cognitive limits of human attention at scale.
AI addresses that by filtering the feed, surfacing only what requires attention, and compressing the gap between an event and a decision. “Video data is no longer just stored for review after an incident. It is being processed continuously, turned into structured insights that can influence responses in real time,” said Lumana CEO Sagi Ben Moshe, an AI video intelligence platform provider that works with organizations across enterprise, commercial, and public environments.
How the FIFA World Cup 2026 is testing AI video surveillance at scale
Large-scale global events offer the clearest real-world test of how these systems perform under pressure.
The FIFA World Cup 2026, spanning venues across the United States, Canada, and Mexico, requires real-time coordination across security, crowd management, and operational teams at a scale that traditional monitoring cannot support, according to the Department of Homeland Security.
More AI:
- Micron sits at the center of a red-hot chip rally
- IBM CEO sends blunt message on AI and quantum computing
- Anthropic CEO makes shocking admission about AI
In these environments, the shift from passive to active surveillance is operational, not theoretical. Ben Moshe explained the core change to TheStreet.
“AI shifts that model from passive monitoring to proactive, real-time situational awareness by continuously analyzing video streams and surfacing only the activity that actually requires attention,” Ben Moshe said. “That dramatically reduces monitoring fatigue while enabling security teams to respond to incidents as they unfold instead of after the fact.”
The challenge of managing public spaces at this scale is not new. What is new is the infrastructure being deployed to meet it.
Vendors, including Motorola Solutions, Axon, and a growing tier of AI-native startups, are competing to define what the next generation of public safety technology looks like, according to the Security Industry Association.
The gap between AI marketing and actual AI performance in surveillance
Despite the acceleration in adoption, a significant gap persists between what organizations believe they are purchasing and what AI video surveillance systems actually deliver. The term AI has become pervasive in vendor marketing to the point where it no longer distinguishes between meaningfully different architectures.
That matters because performance under pressure is where differences become visible and consequential. A system adequate in a controlled environment can fail in a high-density, dynamic venue where the stakes are highest.
The underlying model architecture, GPU processing capacity, and whether the system was purpose-built for video intelligence rather than adapted from general computer vision all determine real-world accuracy.
For public institutions and event organizers, the procurement decision is becoming more complex. Evaluation criteria need to move beyond feature lists to questions of how systems perform at enterprise scale, how they handle edge cases, and what happens when they encounter scenarios outside their training data.
MTStock/Getty Images
Accountability is key for AI real-time decision-making in public safety
The acceleration of AI-powered surveillance prompts a separate and equally urgent conversation. Who is accountable when these systems flag an individual, trigger a response, or inform a decision in real time?
That question is arriving faster than the regulatory frameworks designed to answer it. The European Union’s AI Act, which classifies real-time biometric surveillance in public spaces as high-risk, represents one approach.
In the United States, the regulatory picture remains fragmented, with cities and states setting different rules around facial recognition and automated decision-making, according to the Electronic Privacy Information Center.
“With any AI system, especially one influencing real-time decisions, governance, transparency, cybersecurity, and human oversight have to be built in from day one,” Ben Moshe told TheStreet. “At the end of the day, long-term trust in these systems will depend on balancing operational efficiency with accountability, privacy, and responsible deployment. The companies that get that balance right will define the future of AI in physical infrastructure.”
The practical requirements include role-based access controls, audit trails, clear data retention and governance policies, and maintaining humans in the loop for consequential decisions.
Those are not features most organizations think about during procurement. They become visible later, when something goes wrong or when a regulator asks for documentation of how a decision was made.
Key figures on the AI video surveillance market:
- Global AI video surveillance market forecast: More than doubling to $14.6 billion in the next four years; video analytics use projected to reach $69 billion by 2028, according to Lumana
- North America: Largest regional market for AI video surveillance in 2025; Asia-Pacific expected to be the fastest-growing region through 2030, according to Research and Markets
- Key growth drivers: Expansion of smart city projects, integration of AI with IoT cameras, growth of public safety investment, and adoption of predictive security solutions, according to Research and Markets
- EU AI Act classification: Real-time biometric surveillance in public spaces listed as high-risk, requiring stricter oversight and accountability frameworks, according to the Electronic Privacy Information Center
What the transformation of public safety infrastructure means going forward
The shift from passive recording to real-time intelligence is happening faster than public awareness of it. Most people passing through a monitored space have no visibility into what the systems watching them are actually doing. That invisibility is partly operational by design. The goal is better outcomes, not visible friction.
What is now becoming clear is that the infrastructure of public safety is being rebuilt around a fundamentally different set of assumptions. The camera is no longer the endpoint. It is the starting point of a system that interprets, alerts, coordinates, and increasingly acts. The companies building that layer, the institutions deploying it, and the regulators overseeing it are all still working out what the rules should be.
The market is not waiting for that conversation to conclude. Investment is flowing, deployments are scaling, and the technology is evolving faster than the governance around it.
How those gaps close, and who closes them, will shape how AI-driven public safety is remembered by the people it is designed to protect.
Related: Ondas CEO says AI-powered autonomous warfare is becoming a reality