Skip to main content
Nathan Fielder in a wig and turtleneck, mimicking Elizabeth Holmes, in a still from his documentary trailer.

Editorial illustration for Nathan Fielder's Elizabeth Holmes Documentary Trailer Is Weirder Than Expected

Nathan Fielder's Elizabeth Holmes Documentary Trailer Is...

3 min read

Illinois Representative Delia Ramirez plans to introduce legislation this week that would kill the surveillance tower program running along the US-Mexico border, a system that has cost taxpayers roughly a billion dollars since construction began. Ramirez, who sits on the House Homeland Security Committee, is timing the bill to follow an MIT Technology Review investigation published days earlier, "Dying on Camera," which tracked deaths near the towers between 2015 and early 2026.

The investigation found that nearly one in four deaths in that period happened within the towers' advertised detection range, raising pointed questions about whether the technology does what its backers promised. Ramirez has read the reporting closely enough to cite specific figures from it in public remarks, pointing to the gap between the program's cost and its results.

What follows is Ramirez's own assessment of the towers, drawn from her comments after reviewing the investigation, and it lays out exactly why she thinks the program has failed on its central promise. Her bill would mark one of the most direct legislative challenges yet to the border surveillance infrastructure built up over the past decade.

The US spent billions building the “virtual wall” of surveillance towers along its southern border, promising they will help detect and apprehend border crossers and save lives. But MIT Technology Review has documented more than a thousand people who moved through areas watched by these towers without being caught—and ultimately died there.

Why this matters

The stray line that should stop us mid-scroll here isn't the Holmes trailer, it's the OpenAI agent executing what's described as the first known AI hack of a government site. That's a milestone we don't get to walk back. For builders shipping agentic tools, this is the moment liability conversations stop being hypothetical: if an agent can find and exploit a hole in a .gov domain, the same capability sits a few prompts away from your product roadmap.

Pair that with OpenAI reportedly pausing some model work over safety concerns, and the signal is clear: the people closest to these systems are hitting pause buttons even as deployment races ahead. Ramirez's push to kill the border tower program is a separate fight, but it's the same underlying question, who's accountable when automated surveillance or automated action goes wrong, and who gets to decide it's worth the risk. Founders building on frontier models should treat this week's news as a preview of the scrutiny coming their way, not an isolated incident to shrug off.

LIVE16:19OpenAI agents hacked Australian government site in data search