AI-powered border surveillance towers fail to prevent deaths despite billions spent, investigation finds

A joint investigation by MIT Technology Review and Times of San Diego examined the failures of AI-powered border surveillance towers along the US-Mexico border, finding that over 1,050 people died within the advertised range of these systems between 2015 and early 2026. The analysis cross-referenced nearly 4,000 locations where human remains were found with data on nearly 600 towers identified by the Electronic Frontier Foundation, combining this with field reporting from the border.

The investigation determined that deaths occurred within the advertised range of nearly two-thirds of the analyzed towers, including modern autonomous systems from Anduril—one of the most advanced deployments by US Customs and Border Protection (CBP)—since 2021. Topographical analysis further revealed that while some towers have as little as 10% visibility of their advertised surveillance area due to terrain, most deaths did not occur in these blind spots.

One case involved José Morales Bernal, who died of environmental exposure on April 8, 2024, just 360 feet from the closest of three Anduril surveillance towers in southern New Mexico. Despite the proximity of the AI-enabled systems, landfill workers—not Border Patrol—were the first to spot his body hours after he entered the area. Agents arrived 45 minutes after being called and reported him as 'probably one of the migrants crossing through there,' unaware of his earlier movements near the towers.

Additional cases highlighted by the investigation include a man who dragged his 30-year-old brother into the shade near two AI towers after the brother began having trouble breathing; the pair were only spotted when the man waved down a helicopter for help, and the brother was already dead by the time agents arrived. Another instance involved a 24-year-old woman whose decomposed body lay unnoticed for weeks near an Anduril tower with a clear sightline, discovered only during a routine patrol.

In response to inquiries, an Anduril spokesperson stated that once a tower is delivered, it is operated by CBP, and directed questions about specific incidents to the agency. The spokesperson noted that actual surveillance ranges vary depending on terrain, physical obstructions, and virtual boundaries set by CBP for privacy and other reasons, and alleged inaccuracies in the reporting without specifying what was inaccurate or providing further clarification.

The investigation underscores a fundamental failure of the 'virtual wall'—a decades-long, multibillion-dollar effort to deploy surveillance technology along the 1,951-mile southern border. CBP has credited this system with providing 'persistent surveillance' and 'situational awareness,' as well as saving lives, but the documented deaths challenge these claims and raise questions about the system's ability to effectively identify and locate migrants entering the United States.