Investigation Finds More Than 1,050 Deaths Within Advertised Range of U.S. Border Towers
The investigation cannot determine every breakdown, but it finds AI detection, human response, and evidence retention are hard to assess as the network grows.
Listen to this story
The audio brief
Story brief
3 key pointsAn MIT Technology Review investigation linked more than 1,050 deaths between 2015 and early 2026 to locations within the advertised coverage of U.S. border-surveillance towers, including 110-plus near Anduril’s autonomous systems since 2021. The proximity does not prove a detection failure: terrain, virtual boundaries, alert handling, and agent response can all limit coverage or outcomes. Yet CBP has no...
- 01
Researchers cross-referenced nearly 4,000 remains locations with nearly 600 towers; CBP operates about 800, making the tally an undercount.
- 02
Most identified deaths were not in analyzed blind spots, though terrain substantially constrained some tower views.
- 03
José Morales Bernal was found 360 feet from an Anduril tower after landfill workers—not Border Patrol—called for help.
An MIT Technology Review investigation has found more than 1,050 people died within the advertised range of U.S. border-surveillance towers from 2015 through early 2026. The reporting identifies cases where people passed through areas watched by AI-enabled towers and later died nearby, raising the possibility of failures in detection, response, or both.
A system built to spot movement and send alerts
Customs and Border Protection says its autonomous surveillance towers use AI to detect and classify people, vehicles, and animals, then alert Border Patrol agents to activity in monitored areas. The towers are one part of a wider system that includes older camera towers, drones, blimps, seismic sensors, and tunnel-sensing robots.
That design creates a chain: automated detection, an alert, and a human response. A death near a tower does not establish which link broke. Anduril says CBP operates its towers after delivery, and that terrain, physical obstructions, and agency-set virtual boundaries can affect actual coverage.
The count is large, but not a verdict on every tower
The journalists cross-referenced nearly 4,000 locations where human remains were found with information on nearly 600 towers. The tally is an undercount, the reporting says, because CBP does not publish a comprehensive tower map and operates about 800 towers.
The findings are not confined to older equipment. The investigation estimated that more than 110 people have died within range of modern autonomous Anduril towers since 2021. Its topographical analysis found that most identified deaths did not occur in analyzed blind spots, though some tower views were substantially constrained by terrain.
A missing review process leaves the key question open
José Morales Bernal died 360 feet from an Anduril tower in southern New Mexico after landfill workers, not Border Patrol, noticed him and called for help. In the same desert area that year, the investigation found 18 other deaths within range of three Anduril towers; five were visible from the closest tower.
Former agents and officials told the publication that when a body is found, CBP does not formally investigate whether surveillance should have detected that person or why agents did not reach them. CBP typically deletes tower footage after 30 days unless an internal review is opened, and such reviews usually involve deaths in custody.
Expansion continues as oversight questions mount
A June Government Accountability Office report cited by the investigation said Border Patrol had 803 fielded towers, planned another 175, and expected the network to reach 2,300 by 2034. CBP says it evaluates autonomous surveillance technology by its effect on detection, response coordination, agent safety, and mission outcomes.
MIT Technology Review recommends that CBP audit deaths near towers, improve death records, track technology-assisted apprehensions, and investigate deaths within surveillance range. It reported that CBP has not conducted a comprehensive audit quantifying such deaths and evaluating tower effectiveness, despite having the ability to do so since at least 2015.
Editorial analysis
Our Read
The investigation’s hardest finding is not that every death near a tower proves a camera failure. Terrain, software settings, and human response all shape what happens after a detection. It is that CBP is scaling a system without a public accounting that can distinguish those failures. The next meaningful test is whether the agency conducts tower-by-tower reviews of deaths within surveillance range and preserves relevant footage long enough for those reviews. Without that record, neither claimed life-saving benefits nor the investigation’s most serious questions can be conclusively measured.
Citation desk / original work
Cite this
Citation desk / original work
Cite this
The investigation’s hardest finding is not that every death near a tower proves a camera failure.
/posts/investigation-finds-more-than-1-050-deaths-within-advertised-range-of-u-s-border-towers#finding-1
Sources
- technologyreview.com4 ways to address the failures we found along the US border’s “virtual wall”
- technologyreview.comShe died at the San Diego border. A surveillance camera was in plain sight
- technologyreview.comThe US spent billions on border surveillance. Why can’t it catch people before they die?
Reader comments
Newest comments first. Replies stay oldest first.