Invest1 publisher3 min readPublished
MIT Technology Review counts 1,050 deaths within range of the border's camera towers
Deaths were found inside the advertised range of nearly two-thirds of the roughly 600 towers mapped, and more than 110 of them sit within range of Anduril's autonomous systems installed since 2021.
The Investor · Invest desk
What happened
- Jose Morales Bernal died in the southern New Mexico desert on April 8, 2024, 360 feet from the closest of three surveillance towers, and an autopsy concluded the cause was environmental exposure.
- MIT Technology Review and Times of San Diego cross-referenced nearly 4,000 locations of human remains against nearly 600 towers identified by the Electronic Frontier Foundation.
- The investigation counted more than 1,050 people who died within range of border surveillance towers between 2015 and early 2026, including more than 110 near Anduril systems since 2021.
- Its topographical analysis found some towers can see as little as 10% of their advertised surveillance area, while most of the deaths did not fall inside tower blind spots.
Compiled by The InvestorSomething wrong?How this is made
Why it matters
- constraint With most deaths outside blind spots, sight-line coverage stops being the obvious explanation, so another tranche of sensors has no clear path to changing the outcome the spending is meant to produce.
- exposure Anduril's newest towers now have a measured number attached to them, more than 110 deaths in range since 2021. That is the kind of number a vendor takes into the next contract review.
- contradiction CBP credits the same equipment with saving lives, and the two readings rest on identical hardware, so whoever funds the next order has to pick which measurement the towers are accountable to.
- decision The saving the towers were bought for is control-room labour, and a buyer who wants apprehensions instead has to fund the dispatch end: agents, vehicles and access to the ground they watch.
The chain those towers are supposed to set off has four links, and MIT Technology Review describes each one. Cameras and AI detect and track a person automatically. Live video goes to a nearby control room, and an alert reaches the government-issued smartphone of the closest available agent, who is meant to respond [3]. Three of those links are equipment.
Spread across 2015 to early 2026, the count of more than 1,050 works out at roughly 95 deaths a year inside tower range [1]. The Anduril subset runs at about 22 a year since 2021 [2], close to a tenth of the whole count [3]. Nearly two-thirds of the roughly 600 towers in the analysis had a death inside their advertised range, so on the order of 400 towers [4].
The topographical work cuts two ways. Some towers can see as little as 10% of the area their specification claims, a defect in what was bought and where it was put; most of the deaths, though, were not in blind spots [12]. The second finding is the harder one for the buyer, because a person moving in view of a working camera who dies anyway is a failure somewhere after detection.
MIT Technology Review puts the virtual wall at 25 years and billions of dollars, a fleet that also takes in blimps, drones, seismic sensors and tunnel-sensing robots [14]. CBP has credited the technology with saving lives [15]. The story does not break out what CBP paid for the Anduril towers. The method also sets a limit on how far the count can be pushed. It matched locations of remains and estimated dates of death against tower locations and installation dates [9]. That establishes who was inside a coverage circle, not whose alert fired.
The towers were sold as a labour saving, freeing agents from hours in front of video monitors along 1,951 miles of border [4]. That saving is control-room hours. What the agency needs out the other end is an agent at a location, and the Morales case is the one dispatch the investigation times. Landfill workers saw him lying in the sand at 1 p.m. and called Border Patrol at 4 p.m., when he had not moved. Agents arrived 45 minutes after that, about three hours and 45 minutes after he first went down [5][5]. When an agent phoned 911 he said the body was "probably one of the migrants crossing through there" [6].
In my view the binding constraint sits after detection, and I would expect another order of towers to leave the 95-a-year figure roughly where it is. Two findings would change that. If the towers with 10% sight lines turn out to be a large share of the fleet and not a handful, then siting and hardware carry more of the failure than I am allowing. And if agency logs showed alerts firing and agents rolling promptly across most of those 1,050 cases, the gap would sit in drive time. That is a different purchase: vehicles, staffing and road access.
What to watch
- Whether CBP releases tower alert logs and agent dispatch times for any of the 1,050 sites the investigation mapped.