MIT Technology Review Reveals Systemic Failures in U.S. Border 'Virtual Wall' Surveillance

ME AI Message: Today, MIT Technology Review published our investigative report on the U.S. government’s deployment of surveillance towers along the U.S.-Mexico border—known as the “virtual wall”—focusing on how many people have died near these towers. We found that some individuals passed undetected through surveillance zones, overlooked by advanced AI-powered towers, and later died nearby, their bodies undiscovered for weeks or even months. These cases expose systemic flaws in the “virtual wall”: damaged towers, algorithms failing to detect humans, and agents receiving alerts but taking no action. This article is part of the series “Death by Surveillance,” a collaboration between MIT Technology Review and the Times of San Diego. Over the past year, journalists from both outlets have investigated failures in border surveillance technology and uncovered the stories of those who died along the border. Each death should have prompted U.S. Customs and Border Protection (CBP) to reassess the flaws in its surveillance network—especially those occurring near newly built towers from defense contractor Anduril. Anduril continues to secure massive contracts and stands to benefit from billions in federal funding earmarked for border security. The U.S. plans to invest $1 billion by 2034 to triple the scale of the “virtual wall.” Here are four steps CBP should take immediately to address these issues. 1. Conduct a full audit of deaths near the “virtual wall” In this June 16, 2021, file photo, a family from Venezuela, along with other migrants, surrenders to law enforcement after crossing the border in Del Rio, Texas, before boarding a Border Patrol transport vehicle. ERIC GAY/AP PHOTO “If I were the federal government, I would conduct an audit, because it clearly isn’t working,” said New Mexico Representative Sarah Silva after MIT Technology Review presented her with our findings: in just 2024, over 18 people died near surveillance towers in a small stretch of desert within her district. An audit is necessary because, despite reviewing hundreds of police reports and other documents, we still have limited information about how many towers exist or where exactly they are located. CBP has not released a complete map of the “virtual wall.” We relied on mapping efforts led by the nonprofit Electronic Frontier Foundation, verified with satellite imagery, to pinpoint tower locations and approximate installation dates. We analyzed nearly 600 towers—but this is far fewer than the 800 the government claims are currently operational. Our estimates of installation dates are also imprecise. This means our count of over 1,050 deaths near these towers is an undercount and fails to fully reveal which areas along the border face the most severe problems. An internal audit would avoid these limitations. In theory, CBP should maintain records of when each tower was activated and its operational status. Since at least 2015, when this type of tower began deployment, the agency has had the capacity to assess its effectiveness and tally deaths near its towers. But sources told us CBP has never conducted such an audit. Our findings indicate that conducting one is now urgent. 2. Fix how the agency tracks migrant deaths and remains If CBP is to analyze deaths occurring near its surveillance technology, it must know when and where these deaths occurred. But the current systems for tracking this data are deeply flawed. As noted in our methodology, no national agency collects and publishes all available records of migrant deaths. In Texas alone, we combed through thousands of pages of records from dozens of agencies—but due to inconsistent documentation standards across public records, our data remains incomplete. The Border Patrol has had an internal system since 2007 called the Border Safety Initiative Tracking System (BSITS). But it wasn’t until 2021 that national guidelines were established for how different agencies should enter cases into this system. The following year, the Government Accountability Office (GAO) found that many cases remained unrecorded—particularly those not discovered by Border Patrol agents. In 2023, GAO reported that Border Patrol had taken steps to improve, yet even today, external researchers continue to find undercounting in its estimates. Beyond tracking migrant deaths, CBP launched the Missing Migrant Program in 2017 to prevent such deaths. In April 2025, GAO found that the program improved data collection and tracking of 911 calls and rescue efforts. But GAO noted that the program has no mechanism to use this data to measure whether it has actually increased rescues or reduced deaths—the very goals it was designed to achieve. GAO made two recommendations for improvement, neither of which have been implemented. 3. Track cases where surveillance technology led to apprehensions In 2014, when the “virtual wall” was much smaller, GAO recommended that Border Patrol agents document whether a specific technology helped them apprehend individuals attempting to enter the U.S. illegally. While this would require extra time from agents, recording how often these technologies led to actual apprehensions would help Congress understand their real-world impact. GAO also stated that once collected, this data should be used to evaluate whether surveillance technologies are effective. CBP agreed and began requiring agents to log this information. On July 17, 2024, U.S. Border Patrol agents apprehend an individual near San Land Park, New Mexico. GLENN FAWCETT / CBP PHOTO However, in 2017, GAO reported that the collected data was inaccurate and unreliable. For example, GAO found that agents in Texas had credited hundreds of apprehensions to a specific type of surveillance tower—even though no such towers existed in Texas. Border Patrol again pledged improvements and agreed to train agents on the importance of collecting such data. Yet years later, agents told us this task is still not prioritized. A former CBP official described the requirement as “honestly like spitting into the wind,” arguing that the chaotic reality of apprehensions cannot be neatly converted into standardized metrics. Moreover, more than seven years after CBP agreed to implement this process—in August 2021—the agency still had not completed any analysis evaluating tower effectiveness. This means the expansion of the “virtual wall” has outpaced assessments of its impact. 4. Investigate deaths as potential surveillance failures CBP should also investigate new deaths promptly to determine whether the deceased was within range of a surveillance tower or other monitoring technology. For cases confirmed within coverage areas, the agency must determine why these individuals were not detected sooner. Multiple agents and officials told us that if Border Patrol agents discover a body believed to have died during an attempted border crossing, there is no policy requiring them to check whether the death occurred near surveillance equipment or any established procedure to follow. At minimum, this means potentially critical evidence related to deaths is never investigated—and migrant deaths are rarely treated as possible indicators of “virtual wall” failure.CBP should automatically cross-reference every migrant death with its internal surveillance tower database and investigate cases that occurred within the surveillance range to determine what the system detected and whether agents responded appropriately. Agents already record GPS coordinates when apprehending groups, discovering footprints, and handling other incidents. Without this verification, law enforcement, medical investigators, and families may never know that someone died near a surveillance tower—or that footage potentially capturing the moments before their death may exist. Unless an internal review is initiated, CBP typically deletes surveillance tower footage after 30 days, and internal reviews are usually triggered only for deaths occurring in custody. At a minimum, flagging such cases would preserve footage for investigators and allow families the opportunity to request access. Such reviews could also reveal flaws in the “virtual wall” itself. For example, at a site we identified in New Mexico, multiple individuals died within clear view of an advanced AI surveillance tower, one after another over several months, sometimes only hundreds of feet apart. Investigating these cases might uncover issues such as cameras repeatedly failing to detect or identify humans, or agents receiving alerts but never responding. Conducting these post-incident investigations is the only way to answer the most common questions posed by families, researchers, and volunteers as we investigate hundreds of migrant deaths near surveillance towers: Did the cameras see them, but no one came? Or did no one see them at all? (Source: MLion)