On June 9, 2026, U.S. Customs and Border Protection quietly signed a $71 million task order with GDIT, a General Dynamics business unit, for a new generation of autonomous surveillance towers along the southern border. It's a small line item inside a much larger story: a $1.8 billion indefinite-delivery contract that has been running for three years, funded in part by the more than $6 billion the 2025 One Big Beautiful Bill Act set aside for border surveillance technology -- roughly $1 billion of which is going specifically toward these AI-powered towers. This is what "AI at the border" actually looks like in practice right now, and it's worth separating the real deployment from the hype around it.
What's Actually Being Deployed
The towers come in two sizes, 80 feet or 120 feet, and in testing can detect, track, and identify objects up to eight miles away -- though GDIT says real-world range depends heavily on terrain. They're built to run off-grid, with solar panels and batteries plus propane generators as backup, and use edge computing so they can operate in remote locations without a fast round-trip to a data center. Multiple cameras and sensors track hundreds of targets simultaneously; thermal infrared imaging helps in darkness, and laser rangefinding improves distance accuracy.
The AI's actual job, for now, is classification: telling the difference between a car, a person, an animal, or another object. Michael Wagner, GDIT's VP of biometrics, border and transportation security, told FedScoop the longer-term goal is for the system to determine whether a person is carrying a weapon or a large backpack, with a human operator still in the loop. A second layer of AI links multiple towers together into a single operating picture for command-and-control centers -- arguably the more novel piece: not just smarter cameras, but a networked sensor grid feeding one operational view.
Despite the "autonomous" branding, the actual autonomy on offer is narrower than the word suggests. ICE's own CIO, Dustin Goetz, told an AFCEA Bethesda panel in May 2026 that the agency's real-world work with AI agents remains "minimal." DHS's own updated AI use-case inventory, published months after its compliance deadline, reportedly raised more questions than it answered -- including walking back the deployment status of some systems it had previously listed as live. The towers detect and classify; for now, humans still decide.
Not an Isolated Program
The GDIT towers sit alongside other AI-driven CBP initiatives, including the Rapid Tactical Operations Reconnaissance program (RAPTOR), which fuses radar, infrared, and video data to flag suspicious activity along the border closer to real time. CBP has also run an internal AI Center of Innovation since 2022, applying machine learning to tasks like automatically redacting personally identifiable information from surveillance footage and reducing false positives in cargo and vehicle screening.
Zoom out further and this fits a pattern the broader homeland-security and defense trade press has started calling "border intelligence": fusing signals, electronic, imagery, and open-source intelligence into a single operational picture, then layering predictive analytics on top so agencies can flag risk before it reaches a physical crossing point rather than after. It's the framing vendors like Cognyte have been pushing in industry outlets -- worth reading with the understanding that it's partly a sales pitch -- but the underlying shift, away from isolated sensors and toward fused, AI-assisted analysis, is real, and it shows up in how CBP itself is now structuring its technology roadmap.
The Part That Connects to the Chip Story
One detail in Wagner's interview is easy to miss but tells its own story: GDIT has faced supply chain and logistical constraints building these towers over the past several months, driven in part, in Wagner's words, "by the AI boom and the shortage of chips." CBP Commissioner Rodney Scott told lawmakers in April 2026 that signing contracts was never the hard part; sourcing the hardware has been. That's the same global compute crunch BorderTrend covered in the context of U.S.-China chip export controls -- the domestic border-security buildout is competing for some of the same silicon.
What to Watch
GDIT says it's exploring counter-UAS capabilities for future tower iterations in response to rising drone activity at the border, though for now the systems are ground-focused. Whether agentic AI moves meaningfully beyond the "minimal" stage ICE's own CIO described, and whether DHS's AI oversight documentation catches up with what's actually being deployed, are the two threads most worth tracking as this buildout continues through the rest of 2026.