MIT Technology Review maps 1,050+ deaths near US border's AI surveillance towers

MIT Technology Review, working with the Times of San Diego, has published what it describes as the first comprehensive map of migrant deaths near the US government's network of surveillance towers along the US-Mexico border. The 15-month investigation, part of a series called Dying on Camera, counts more than 1,050 people who died near the towers, and argues the real number is higher. Alongside the findings, the outlet published four recommendations for US Customs and Border Protection (CBP), from an internal audit to routine checks of whether deaths happened within camera range.

What the investigation found

According to MIT Technology Review, people walked undetected through terrain covered by advanced, AI-enabled towers and later died nearby, with their remains going unnoticed for weeks or months. The reporters attribute this to systemic flaws: towers that were out of service, detection algorithms that failed to pick people out, and agents who did not act on alerts.

The network itself is opaque. CBP publishes no map of the "virtual wall", so the team relied on mapping by the Electronic Frontier Foundation and satellite imagery to verify tower locations and estimate when they were installed. That yielded nearly 600 towers — well short of the roughly 800 the government says are operating — one reason the death toll is treated as incomplete. The analysis covers cases back to 2015, spanning multiple administrations and tower technologies.

The stakes are growing: the newest towers are built by defense contractor Anduril, which continues to win major contracts, and the US is set to spend $1 billion to triple the size of the virtual wall by 2034.

How the map was assembled

The project began with a question about why so many people died near towers designed to track them. Journalists reviewed thousands of pages of government documents, visited the border several times, and interviewed more than 45 people, including current and former White House advisors, Border Patrol agents, medical examiners, sheriffs, humanitarian volunteers and employees of tech companies.

The dataset merges existing records from organizations including No More Deaths, Humane Borders and the Electronic Frontier Foundation with material obtained from more than a dozen agencies. Texas was the largest gap, so reporters filed requests with sheriffs in 17 counties beginning in July 2025. Fourteen supplied usable records — some charged between $300 and $1,300 — adding up to more than 4,000 pages and over 1,500 Texas cases.

One methodological detail stands out for a technical audience: for three counties that sent especially large volumes of records, the team used Anthropic's Claude, accessed through an API, to extract the coordinates where remains were found, then manually checked batches to verify the model's accuracy.

The underlying records are patchy by design. No national protocol governs how migrant deaths are documented, and most agencies do not note whether a deceased person was believed to be crossing the border — even as some counties track minutiae such as whether a burglar entered through a chimney. Border Patrol estimates that more than 10,000 people have died during crossings since 2000, a figure widely seen as an undercount.

Four recommendations for CBP

First, conduct a comprehensive audit of deaths near the virtual wall. Sources told the reporters that CBP has had the ability to correlate tower activity with deaths since at least 2015 but has never done so. New Mexico state representative Sarah Silva, after being shown that more than 18 people died near towers in one small stretch of desert near her district in 2024 alone, said the federal government should audit a system that appears not to be working.

Second, fix how migrant deaths and remains are tracked. Border Patrol's internal tracking system dates to 2007, but national rules for entering cases arrived only in 2021, and a 2022 Government Accountability Office (GAO) review found cases still missing. In April 2025 the GAO found that the Missing Migrant Program collects better data but has no plan to measure whether it actually increases rescues or reduces deaths; two of its recommendations remain unimplemented.

Third, log when surveillance technology contributes to an apprehension. The GAO proposed this in 2014 and CBP agreed, but a 2017 review found the resulting data unreliable — agents in Texas credited a type of tower that did not exist in the state. A former CBP official, speaking anonymously, told the reporters the logging mandate is still not a priority, and an effectiveness analysis completed in 2022 has not been made public.

Fourth, treat deaths as possible surveillance failures. There is currently no requirement or process for checking whether a person died within range of a tower. The recommendation is to automatically cross-check every migrant death against CBP's internal tower database and investigate what the system detected and how agents responded.

Why it matters

The investigation documents a surveillance build-out expanding faster than anyone is measuring it, increasingly built on automated detection and slated for a major funding increase. For AI in particular, it is a case study in algorithmic systems deployed in life-or-death conditions without a feedback loop: when the technology fails to detect someone, nothing in the process surfaces that failure. And because no agency reliably records where and when migrants die, the human cost of those gaps stays invisible — to regulators, to lawmakers weighing the next billion dollars of spending, and to families searching for missing people.