The DHS Study That Said Its Own Towers Don't Work

Introduction

A DHS-funded study found in 2020 that border surveillance towers were "associated with decreased apprehension levels," because migrants just walked around the cameras. CBP's answer was a $1.8 billion contract for AI towers that now run with "minimal human oversight," awarded to the same contractor whose existing towers were two-thirds broken when the study dropped.

A 25-Year Loop, Five Different Names

The contract was awarded under the Consolidated Tower and Surveillance Equipment program on September 5, 2023. CTSE is the fifth federal program in 25 years to put cameras on the southern border. The first four were the Integrated Surveillance Information System in 1997, America's Shield Initiative in 2003, SBInet in 2006, and the Integrated Fixed Tower program in 2014. Each was canceled or quietly replaced after the GAO, the DHS Inspector General, or both reported it didn't work.

ISIS cost $43 million through a no-bid contract; a DHS IG investigation found its maintenance center had "little or no work performed." America's Shield Initiative absorbed $163.6 million before cancellation. SBInet, the Boeing-led $1 billion "virtual fence," was canceled by DHS Secretary Janet Napolitano in 2011 after delivering 53 miles. Elbit Systems of America won the next contract for $145 million in 2014. By 2017, the GAO found Border Patrol stations in the Rio Grande Valley had recorded roughly 500 "asset assists" from IFT towers in a sector that didn't have any.

Each cancellation became the sales pitch for the next contract, and the story that never makes it into a single piece is the system that turns every failure into the procurement justification for whatever replaces it. CTSE is built directly on top of GDIT's Remote Video Surveillance System Upgrade infrastructure, the same system Congress subpoenaed DHS over three months after CTSE was awarded.

The Receipt Congress Already Pulled

On October 31, 2024, House Homeland Security Chairman Mark Green subpoenaed DHS Secretary Alejandro Mayorkas over the same towers GDIT was being paid to upgrade. The committee's letter said directly: "Multiple sources informed the Committee that more than 66 percent of cameras in the RVSS-U program are inoperable." The same letter put the timeline on the record: "since June 2023, roughly two-thirds of those cameras have gone completely off-line and have not been repaired."

June 2023 matters because the CTSE solicitation was open in Q1 2023, and the contract was awarded September 5, 2023. The cameras started going dark inside the same window the agency was picking who would inherit them, and the agency picked the same company. The committee also reported that whistleblowers said maintenance was "sometimes performed by non-U.S. citizens, possibly in violation of CBP security policy."

NBC News had broken the lower-bound version a few weeks earlier. An internal Border Patrol memo from early October 2024 said 30 percent of the cameras (about 150 of 500) were inoperable nationwide and that "the nationwide issue is having significant impacts on [Border Patrol] operations." One CBP official told NBC the problem was "something that has not been properly managed for the last 20 years," which lines up with the start of America's Shield Initiative in 2003.

The 30 and 66 percent figures don't reconcile cleanly. The NBC memo cited RVSS broadly; the committee cited RVSS-U, the GDIT-modernized variant. GDIT has pointed out that under the FAA interagency agreement, the FAA serviced the cameras, not GDIT directly. That's true, and it doesn't help: CBP wrote the contract structure that handed maintenance to the FAA, and CBP picked the contractor that would inherit the next $1.8 billion ceiling without resolving the maintenance question first.

The Lobbying Window

The Senate Lobbying Disclosure Act database is where this story stops being a coincidence. General Dynamics' Q1 2023 LDA filing reports $2,810,000 in lobbying expenses for the first three months of 2023. It lists FY24 Homeland Security Appropriations under specific legislation lobbied, DHS, CBP, and ICE as agencies contacted, and "border security technologies" verbatim as a lobbying issue.

That filing covers the exact quarter the CTSE solicitation was live, and the contract was awarded that September. For scale, General Dynamics spent $11.58 million on federal lobbying in 2022 and $12.21 million in 2024, per OpenSecrets, so the Q1 2023 figure tracks the company's normal cadence. What jumped out was the specificity: "border security technologies" as the issue and CBP as the named agency contacted — not a generic line item.

The sequence isn't a smoking gun on its own. Defense contractors lobby their customer agencies; that's the system. The receipt only becomes a receipt when paired with the operational record: the agency was about to award a new tower contract to a company whose existing towers were failing, while that company was actively lobbying about "border security technologies." Those two pieces sat in primary documents and were never published in the same article.

"Extremely Expensive Theater"

The phrase belongs to Dave Maass at the Electronic Frontier Foundation, not to any internal CBP document. Maass told AZFamily this month: "What's going on at the border is just extremely expensive theater." He framed the AI rebranding directly: "Now they're coming back with the same thing, but saying it's the same thing plus AI — as if AI has actually deserved to win anybody's trust."

The hardware is real enough. GDIT announced on April 22, 2026 that its autonomous surveillance towers had received CBP certification to meet requirements of the One Big Beautiful Bill Act. The towers use edge AI, patented video analytics, and 5G plus Starlink for backhaul, and the company has already deployed more than 200 of them along 566 miles of border. The same press release describes the towers operating with "minimal human oversight" and "limited human intervention."

"Minimal human oversight" was already the problem when 30 to 66 percent of the cameras were broken and nobody noticed for a year, and the fix on the table is to do even less of it. The EFF count is at least 585 autonomous towers already along the U.S.-Mexico border, with Border Patrol planning 1,500 more.

Per-tower installation cost runs up to $1.5 million, and the towers run on a Phoenix-based AI vision system from PureTech Systems. The point Maass is making, and the point the procurement record supports, is that functional hardware deployed in a quarter-century failure cycle doesn't fix the cycle — it just adds the word "AI" to the marketing language for the next round.

Who Benefits

GDIT is the named beneficiary, paid first to build the cameras that failed and now paid again to replace them. The 14-year ceiling pushes the revenue horizon out to 2037, with an initial $20.5 million task order plus parallel awards to Elbit Systems and Advanced Technology Systems Company. The cover is structural: CTSE resets the accountability clock on RVSS-U by absorbing the old infrastructure under a new contract number and new technology language. Any future audit will measure CTSE against its own baseline, not against the failures it inherited.

Anduril is the secondary beneficiary, courtesy of the One Big Beautiful Bill Act signed July 4, 2025. The bill directed over $6 billion to border technology, with $2.77 billion for Autonomous Surveillance Towers. At enactment, a CBP spokesperson confirmed to The Intercept, "Anduril is now the country's only approved border tower vendor." Anduril's political connections are on the public record. Trump's Under Secretary of the Army nominee Michael Obadal worked there through June 2025 and refused to divest his stock before the Senate confirmation hearing. Co-founder Trae Stephens served on Trump's 2016 transition team; Palmer Luckey, the company's founder, has hosted multiple Trump fundraisers. GDIT obtained the same CBP certification on April 22, 2026.

CBP's procurement apparatus benefits structurally. Every cycle of failure gives the agency new technology language for Congress and a new contractor to manage. The 2014 GAO report made six recommendations on the IFT program; CBP rejected two. The 2024 GAO report found CBP "failed to address all six of the main privacy protections" required under the Fair Information Practice Principles. After the administration cut staff at the DHS Office for Civil Rights and Civil Liberties in June 2025, DHS asked GAO to close the related recommendations. GAO refused.

The Procurement Cycle Is the Product

The 2020 RAND study is the strongest piece of evidence the towers don't achieve what CBP says they're for, and nobody talks about it. The 88-page report was sponsored by DHS's Science and Technology Directorate, meaning DHS paid for a study about a program DHS was running. The most robust finding: "deploying integrated fixed towers (IFTs) is associated with decreased apprehension levels in the zones of deployment." RAND called it "strong evidence" of a negative effect, with the mechanism that migrants began to "walk around the viewshed" of the cameras. A Border Patrol official in Tucson confirmed this on the ground.

CBP would invoke a caveat: decreased apprehensions could also mean deterrence, and RAND acknowledged this directly. The harder fact is that the testing CBP did was insufficient to determine effectiveness in the first place. GAO-14-368 had already made that point in 2014: CBP's IFT test plan was designed "to determine its mission contributions, but not its effectiveness and suitability for the various environmental conditions, such as weather, in which it will be deployed." DHS did not concur. Ten years later, CBP's response to RAND was to lock the same architecture into a 14-year contract.

A procurement system with no mechanism to punish failure produces more failure, and inside that system the failure is a feature for everyone in the room except the taxpayer. CBP gets new contracts to wave at Congress when somebody asks why the last cameras went dark, and the contractor that built the dark cameras is the same contractor paid to build whatever comes after them. None of those incentives point at fixing the maintenance problem the internal Border Patrol memo called "not properly managed for the last 20 years."

The Next Failure Already Has a Contract

The CTSE contract is the predictable next turn of a cycle that has run five times since 1997, with each successive contract larger and each successive contractor drawn from the same ecosystem. $1.8 billion is the headline figure; what doesn't appear in any line item is any institutional path for the cycle to stop. The "minimal human oversight" language in GDIT's April 22 press release describes the accountability environment for the next 14 years, the same environment that produced a year of broken cameras nobody noticed.

The question worth holding onto is whether anything about the 2026-2037 contract window is different from the 1997-2024 history that produced it. If the answer is no, the next failure already has a successor contract scheduled to absorb it. The thread covers the September 2023 timing and the lobbying receipt. The full RAND apprehensions finding, the GAO data-fabrication record from 2017, the Anduril/OBBBA monopoly story, and the GAO-25-107302 privacy-protection failures only live here, because the trail took 25 years to assemble.