The West is betting that earlier detection can change wildfire outcomes
As severe wildfire conditions build across the American West, utilities and state agencies are scaling up the use of AI-enabled camera networks designed to spot smoke quickly and alert responders before a blaze grows out of control. The technology is being positioned not as a replacement for firefighters or human judgment, but as a force multiplier in landscapes where vast distances and limited visibility can cost critical minutes.
The case for these systems rests on a simple operational truth: the earlier a fire is identified, the greater the chance it can be contained while still small. In Arizona, one example already serves as a proof point. On a March afternoon, artificial intelligence flagged something resembling smoke on a camera feed from Coconino National Forest. Human analysts then verified that the signal was not a cloud or dust and alerted the state forest service and Arizona Public Service. The fire that followed, later named the Diamond Fire, was contained before it grew beyond 7 acres.
That sequence captures the model now spreading across multiple states: machines scan continuously, humans verify, and authorities respond. It is an incremental change in workflow, but one with potentially large consequences in a region facing record heat and meager snowpack.
From isolated cameras to regional networks
Arizona Public Service has nearly 40 active AI smoke-detection cameras and plans to expand that total to 71 by the end of summer. The state’s fire agency has deployed seven of its own. In Colorado, Xcel Energy has installed 126 cameras and aims to have systems operating in seven of the eight states it serves by the end of the year.
California has already moved at a much larger scale through ALERTCalifornia, a network of roughly 1,240 AI-enabled cameras distributed across the state. The system works in a similar way, using AI to scan for possible smoke while keeping humans in the loop to cut down on false positives and improve the model over time. That human review layer is not incidental. It is one of the reasons these deployments are being treated as operational tools rather than experimental curiosities.
False alarms are a major risk in environmental monitoring, especially in rugged terrain where weather, dust, haze, and light conditions can easily confuse automated systems. By requiring human confirmation before escalating alerts, agencies aim to preserve trust in the technology while still capturing its speed advantage. According to ALERTCalifornia founder Neal Driscoll, that feedback loop also trains the system to become more accurate.
