Why AI border cameras often miss people in danger

Journalists cross-referenced thousands of records of migrant deaths with the locations of nearly 600 surveillance towers. They found that in many cases, people who died while attempting to cross the border were within the operating range of these systems. In one documented case, a man died just 360 feet from an AI-equipped tower, only to be found later by landfill workers. Despite billions of dollars in spending and plans to triple the network’s size, there is no public record of the government auditing these systems to see if they are effectively preventing deaths or why they so often fail to trigger a rescue.

These surveillance towers act like automated digital sentries. They are equipped with specialized lenses—including infrared cameras that see heat signatures, allowing them to spot people in total darkness or through heavy brush. The AI component acts as an automated filter; instead of a human staring at a screen, the computer runs an algorithm that analyzes the live video feed to distinguish a human form from a wandering animal or a swaying tree. When the system detects a person, it is programmed to send a digital alert to a smartphone carried by nearby border agents, showing them the exact location. If the system is working perfectly, the human agent receives a notification and heads to that location. The failures described in this report suggest that the chain breaks in two places: either the software fails to categorize a human as a person in need, or the human agents receiving the alerts are unable or unwilling to respond to every ping the system sends.

This situation reveals the limitations of relying on technology to manage complex human crises. We tend to view AI as an all-seeing eye, but the effectiveness of these tools depends entirely on what happens after the software sounds the alarm. If the agency does not have a formal process to investigate these deaths as system failures, the virtual wall may continue to expand, creating a false sense of security while systemic gaps remain unaddressed. It forces a difficult question about the role of technology in border policy: is the goal to track and stop movement at any cost, or can these tools actually be used to protect the people who find themselves in life-threatening danger?