Border surveillance towers were built to save lives. Over 1,050 people died near them anyway
A 15-month investigation found migrants dying within sight of AI-enabled US border towers, and found the government has never audited whether the towers work.
What happened: A joint investigation by MIT Technology Review and Times of San Diego built the first comprehensive map linking migrant deaths to US border surveillance towers. Reviewing nearly 4,000 death cases and records on roughly 600 towers since 2015, reporters found that more than 1,050 people died within the towers' likely viewing range, and the true number is almost certainly higher, since the government's own tower count runs to about 800 and doesn't include everyone who was tracked down.
Why it matters: The towers exist specifically to spot people crossing so agents can respond, whether to arrest them or rescue them. Instead, reporters documented cases of people who died slowly of dehydration or drowned quickly in rivers, sometimes within a few hundred feet of a functioning camera, with no agent response recorded until it was too late. One tower operator reportedly watched a father and toddler fall into the Rio Grande and did not get agents there in time to save the child. CBP has never audited how often that happens, despite having the technical ability to do so since at least 2015.
How it works, plainly: The border camera network mixes older Remote Video Surveillance System cameras, run mainly by General Dynamics, with newer AI-powered towers from companies like Anduril that automatically detect and classify people, vehicles, and animals and alert agents. In practice, both systems have struggled: congressional staff found over 66% of first-generation towers unusable in 2024, and 30% were still broken in January 2026. Even working cameras overwhelm the agents watching them, who also juggle radios, sensor alerts, and records checks, and camera duty is widely treated as an unwanted assignment rather than a monitored, staffed priority.
The rollout: CBP is expanding, not pausing: it plans to spend $1 billion on roughly 1,500 more towers by 2034, tripling the network, even as government watchdogs have repeatedly found the agency's data on tower effectiveness unreliable since 2014. Reporters and outside experts recommend CBP audit existing towers, fix fragmented death-tracking since there is still no national database, properly log when tech actually leads to apprehensions, and automatically check new deaths against tower locations before footage is deleted after 30 days. None of these fixes are currently required.