MIT Technology Review, in collaboration with Times of San Diego as part of the "Dying on Camera" project, examined how many people have died near the surveillance towers the US government has installed along the US–Mexico border. The investigation identified cases in which people walked through areas monitored by advanced, AI-enabled towers and later died nearby, with bodies remaining undiscovered for weeks or months.
The reporting points to systemic problems in the virtual wall: towers that were broken, algorithms that failed to detect people, and alerts that agents did not respond to.
The issue has policy and budget implications: some deaths occurred near the newest towers built by defense company Anduril, even as the US plans to spend $1 billion to triple the size of the virtual wall by 2034.
Key findings and data limitations
- Reporters analyzed nearly 600 towers using mapping from nonprofit groups and satellite imagery. That number is lower than the roughly 800 towers the government says are in operation today, so the investigation’s count of more than 1,050 people who have died near towers is likely an undercount.
- New Mexico state representative Sarah Silva told MIT Technology Review that, in 2024 alone, more than 18 people died near surveillance towers in a single small stretch of desert in or near her district, a finding that prompted her to call for an audit.
- Customs and Border Protection (CBP) does not publish an authoritative map of the virtual wall. External researchers therefore face gaps and imprecision when trying to determine how many towers exist and when each was installed.
Four recommended actions for CBP
1) Conduct a comprehensive audit of deaths near the virtual wall
The investigation argues that only an internal, comprehensive audit by CBP could reliably determine how many towers exist, when they were installed and active, how often they were operational, and how many deaths occurred in their vicinity. CBP has kept records that should make such an audit possible—it began installing the types of towers included in the analysis at least as far back as 2015—but sources told reporters the agency has never carried out this kind of audit.
2) Fix how the agency tracks migrant deaths and remains
Current systems are fragmented. There is no single national agency that collects and publishes all available records on migrant deaths. The journalists reviewed thousands of pages of records from dozens of agencies in Texas alone and still found an incomplete picture because there is no shared protocol for documenting those deaths in public records.
- Border Patrol has used its internal Border Safety Initiative Tracking System (BSITS) since 2007, but national rules for entry into that system were only established in 2021. The Government Accountability Office (GAO) found in subsequent reviews that many cases were still not being tracked in BSITS—especially deaths not discovered by Border Patrol.
- CBP launched a Missing Migrant Program in 2017 to help prevent deaths. An April 2025 GAO report found the program improved how it collected and tracked data such as 911 calls and rescues, but it lacked a plan to use the data to measure whether the program actually increased rescues or reduced deaths. The GAO made two recommendations; neither has been implemented.
3) Track instances when surveillance technology leads to an apprehension
The GAO suggested in 2014 that Border Patrol agents log when technology aided an apprehension, to let Congress and oversight bodies measure how often various technologies led to arrests. CBP agreed and required agents to record such information.
But a 2017 GAO report found the collected data to be inaccurate and unreliable; for example, agents in Texas were recording specific tower types hundreds of times even though such towers did not exist in that state. CBP said it would improve training and data quality, but agents told reporters this task still isn’t always prioritized in the field. Although CBP completed an analysis of tower effectiveness in 2022, the results are not public, and as of August 2021 it had not produced such an analysis.
The result is that the virtual wall is expanding faster than its impact is being independently assessed.
4) Investigate deaths as potential surveillance failures
CBP should examine new deaths to determine whether they occurred within sight of towers or other surveillance technology and, when they did, determine why those people were not detected sooner. Several agents and officials told reporters that if Border Patrol agents find a body they suspect died while crossing, there is no requirement to check whether the death occurred near surveillance infrastructure, nor is there a routine process to do so.
Without that check, potentially crucial evidence about whether the system saw someone and whether footage exists can go unexamined. CBP typically deletes surveillance footage from towers after 30 days unless an internal review is opened, which usually occurs only for deaths in custody. Automatically flagging cases that occurred within surveillance range could preserve footage for investigators and give families a clearer avenue to request recordings.
Postmortems could also expose how the virtual wall itself is failing. The investigation cites one location in New Mexico where several people died within months of one another and within a few hundred feet of an AI-enabled tower. Looking into those cases could reveal, for example, that a camera repeatedly failed to detect people in a particular area or that agents received alerts and did not respond.
Only by conducting these postmortem reviews, the reporters argue, can authorities answer the central question raised by families, researchers, and volunteers who study hundreds of migrants who died near towers: did the cameras see them and no one came, or did the system not see them at all?
Conclusion
The MIT Technology Review and Times of San Diego investigation documents structural gaps in the oversight, tracking, and evaluation of border surveillance technology. It recommends four concrete steps for CBP—an internal audit, better death-and-incident tracking, systematic logging when technology assists with apprehensions, and automatic checks of deaths against tower records—to help determine whether surveillance failures contributed to preventable deaths along the border.
