Skip to content

Product1 publisher3 min readPublished

Graciela Gomez Hernandez died a mile from a tower rated to track at seven miles

Times of San Diego found at least 138 sets of migrant remains inside the range of California's border cameras over four years, and an MIT Technology Review terrain analysis estimates more than half lay in open view.

The Product Desk · Product desk

Photograph accompanying Graciela Gomez Hernandez died a mile from a tower rated to track at seven miles
Photo: technologyreview.com

What happened

  • On Sept. 14, 2025, Graciela Gomez Hernandez, 30, crossed from the eastern edge of Tijuana into Southern California, sending voice messages to her mother and sister as she walked into the Otay Mountain Wilderness.
  • A General Dynamics tower in place on a nearby hilltop since 2019 carries two sets of electro-optical and infrared cameras with long-range lenses rated to track targets 5 to 7 miles away.
  • A Times of San Diego analysis counted hundreds of cases along the California border, at least 138 of them in the past four years, where migrants' remains were found within range of those cameras.
  • The tower network has grown over recent years at an overall cost of more than a billion dollars, with contractors adding newer AI camera systems along the way.

Compiled by The Product DeskSomething wrong?How this is made

Why it matters

  • constraint A rated range of 5 to 7 miles bounds what a camera can resolve and commits nobody to climb the hill, so the response half of the system sits outside anything the procurement measured.
  • decision The audit method here is cheap and repeatable: asset coordinates, outcome records, terrain. Every operator of a deployed detector now chooses whether to run it or wait for a reporter to run it for them.
  • exposure More than a billion dollars of camera spending has an outside measure of its human outcomes for the first time, and two newsrooms built it, not the agency holding the contracts.
  • contradiction Proximity and visibility are counted; causation is not tested. The strong reading, that tower coverage itself drives people into terrain that kills them, is not what these two analyses measure.

Before she died, Gomez Hernandez told her family in a WhatsApp message that she could see helicopters overhead and white Border Patrol trucks making periodic laps on the road below [5]. Between 2 and 3 p.m., her sister said, she sent her mother a last message. "I can't do it anymore," she whispered into her phone [3]. She had told the family about the plan only the day before. She hoped to work and send money to her three children and her parents in southern Mexico; her oldest son, 11, had asked for a bicycle [18]. A forensic crew collected her two weeks later, when much of her body had decomposed [4].

The camera systems added to these towers do two jobs, as the investigation describes them: they tell border-crossers from other people and animals, and they track movement [8]. A person who has stopped walking falls outside both. Dozens of the deaths in the California count were slow deaths of exposure that emergency aid might have averted [12].

At least 138 cases in four years works out to about 35 a year, one set of remains inside camera range roughly every 11 days [1]. MIT Technology Review's line-of-sight estimate was made across the wider set of hundreds of cases. Applied to the 138 alone, more than half means more than 69 people found where terrain did not block a camera's view of them [2]. The General Dynamics tower on the hilltop is rated for targets five to seven miles out, and Gomez Hernandez was at one mile, between a fifth and a seventh of the rated distance [3].

Range and sightline are two different tests. Marcia Dutra, a 36-year-old Brazilian woman, was found next to a border fence in July 2023, likely dead of heat exposure. The nearest tower with AI cameras sat 2,000 feet northeast above the edge of a plateau, out of sight of the slope where she lay, and according to a medical examiner's report the Border Patrol learned of her body from a group of migrants who had been apprehended [14]. Elvi Vasquez Bortolon, a 28-year-old Mexican man, had been dead more than five days from suspected hypothermia when he was found on a ridge near the wall in February 2024 [15].

Dying on Camera, the collaboration that produced both analyses, describes a year of reporting by MIT Technology Review and Times of San Diego [17]. On why anyone is in that terrain, the reporting is plain. The eastern edge of San Diego is a mile from the wall, and migrants walk 10 or 20 miles north through the mountains to evade the Border Patrol, coming out in smaller towns and on highways with fewer checkpoints [16]. The camera positions in the two analyses are matched to where bodies were found, not to where routes moved.

If you are standing up a detector of any kind, the grid worth drawing has two axes. Down the side, whether the sensor can produce the signal. Across the top, whether a named role owns the action, with a clock on it. Vendor decks and budget arguments live almost entirely in the first column, where range and classification accuracy are measured and the response is assumed. The test for each class a model can emit is which person gets paged and how many minutes they have. Where that name is missing, the deployment is producing a record for somebody to count later. The people found inside camera range on the California border were as old as 70 and as young as 5 [13].

What to watch

  • Whether CBP releases tower coordinates and coverage maps that would let the line-of-sight overlay be reproduced outside California.
  • Whether any contract for the AI camera systems adds an alert class for a person who has stopped moving, and names who responds to it.
  • Whether medical examiner records in Arizona and Texas get the same terrain-and-camera treatment the California border records just received.
Loading claim ledger
Loading source directory links
Loading share composer
Loading topic controls
Loading related stories