AI-powered cameras are now scanning construction sites in real time, spotting missing hard hats and unsafe behaviours before anyone gets hurt. Image: Surveillance Secure.
Construction is one of the most dangerous industries in the UK. That has been true for decades. But something is quietly shifting on building sites across the country, and it has nothing to do with new regulations or another training course.
It is a camera. Or more precisely, it is what the AI camera system can now see.
What Changed in Computer Vision Systems
Computer vision systems are now being deployed on construction sites that can watch a live video feed and identify hazards in real time. Missing hard hat. Harness not clipped in. Worker standing too close to a crane swing radius. Two operatives in a confined space without a spotter.
These systems do not rely on someone watching a bank of monitors all day. The AI does the watching. It flags the problem, sends an alert to the site manager's phone, and logs the incident. All of this happens in seconds, not hours.
According to Occupational Health & Safety, roughly 28% of EHS functions already use artificial intelligence, and nearly half plan to invest in AI-enabled safety capabilities within the next year. The shift from reactive to predictive safety management is well underway.
40-50% Incident reduction reported by firms using AI safety
28% of EHS teams already using AI tools
30 UK construction worker deaths in 2023/24 (HSE)
Beyond Hard Hat Detection
The early versions of these systems were basic. They could spot a missing hard hat with reasonable accuracy. Now they are doing much more.
Geofencing technology creates virtual perimeters around high-risk zones: excavation edges, crane swing areas, fall hazard locations. If an unauthorised worker enters one of these digital boundaries, both the worker and the supervisor are alerted instantly. No clipboard sign-in required.
Predictive analytics go further still. By combining weather data, project schedules, workforce patterns, and historical incident records, some platforms now flag high-risk activities before they even begin. A forecast of rain on a day with planned work at height? The system flags it the night before.
The Implementation Problem
Here is the part that does not get enough attention. A safety AI system that sends 200 alerts a day quickly becomes background noise. Site managers start ignoring it. Workers resent it. The whole thing backfires.
Poor implementation can create alert fatigue, mistrust, and a false sense of security.
That warning comes from OHS Online, and it is worth taking seriously. The technology works. The question is whether it is deployed thoughtfully.
There are real governance questions here. Who decides what counts as a flaggable risk? What happens to the footage? Is it used for coaching or for punishment? If the AI starts scoring individual workers, does that create a surveillance culture that actually makes people less safe?
What This Means for the Built Environment
For quantity surveyors and project managers, AI safety monitoring is increasingly showing up in tender requirements and principal contractor obligations. Some major clients are now specifying computer vision monitoring as a condition of contract. That changes the cost model for site supervision and the insurance picture for the project.
And the insurance industry is paying attention. Gallagher's 2026 AI Adoption and Risk Survey found that one in five insurance professionals said a client had experienced a loss or claim related to AI-related risks in the past year. The classes of business most likely to be affected include product liability and employers' liability. AI safety tools that reduce incidents could meaningfully shift premiums, but only if they are properly governed.
Construction firms operating in or supplying the EU market also need to be aware that the EU AI Act becomes generally applicable in August 2026, with specific obligations for high-risk AI systems including those used for worker monitoring and safety-critical applications.
The technology is genuinely useful. But deploying a camera that watches people is not the same as deploying a tool that makes people safer. The gap between those two things is governance.
Sources: Occupational Health & Safety, AI Transforming Construction Safety (February 2026) | ABC, AI in Construction Site Safety (December 2025) | Kwant AI, Construction Safety & Workforce Tracking (2026) | Gallagher, 2026 AI Adoption and Risk Survey | Browne Jacobson, AI in Construction Legal Risks (January 2026) | Autodesk, 2026 AI Construction Trends.
