Barksdale was a warning shot
In March 2026, days after Operation Epic Fury began, Barksdale Air Force Base battled multiple waves of 12 to 15 drones operating over sensitive areas of the installation, including the flight line. Barksdale hosts B-52s and nuclear weapons storage. NORTHCOM commander Gen. Gregory Guillot confirmed the trend to Congress: 'We've seen an increase from last year in the number of detections over military installations.' The baseline was already 350 detections across more than 100 installations in 2024.
Then the number that should end any complacency: 'about a quarter of the ones that we detect we're able to defeat.' One in four. Over strategic bases, inside the continental United States, in 2026. Guillot's threat assessment was equally direct: 'the primary threat I see for them in the way they've been operating is detection and perhaps surveillance of sensitive capabilities on our installations.' Ukraine's Operation Spiderweb put more than 100 drones over Russian strategic airfields from container trucks. The homeland version of that raid is a planning assumption now, not a hypothetical.
Authorities caught up. Capability did not.
The policy layer moved fast. In December 2025, the Secretary of War signed updated guidance empowering base commanders to treat unauthorized drone surveillance as a threat, control an enlarged defensive airspace, and share UAS sensor data across agencies. JIATF-401 now coordinates the joint counter-drone fight. Lt. Col. Adam Scher: 'Installation commanders are fully empowered to develop base defense plans in accordance with their unique capabilities and threats.'
The capability layer did not keep pace. A House Oversight hearing on unauthorized drone activity found base commanders lack monitoring capability and face a near-total lack of counter-drone equipment, with unclear procedures layered on top. Guillot told the Senate what a real defeat capability requires: 'You can't have just one or two different defeat mechanisms. They have to be across multiple mediums, everything from denying satellite guidance, all the way to first-person-view and everything in between.' Multiple sensors, multiple effectors, hundreds of installations. That is a fleet coordination problem, not a gadget procurement.
The autonomy line just moved
On June 9, JIATF-401 approved an autonomous counter-drone system for joint-force-wide use after testing at MCAS Yuma. The system runs automated sense-and-shoot algorithms around the clock, with jamming out to 40 miles and net effectors to four. Scher again: 'now validated for use as one component of a layered C-UAS defense across the entire Joint Force.' The Pentagon also stood up a directed-energy pilot at five bases, pairing high-energy lasers with high-power microwaves.
Read the June 9 approval carefully, because it marks a doctrinal line: the joint force just accepted machine-speed detection and defeat for base defense, inside CONUS, with humans supervising rather than pulling each trigger. Drone raids arrive faster than a watch floor can adjudicate. The assignment of effectors to tracks has to run locally and immediately, under policy a human set in advance. That is the same architecture the interceptor fight demands, applied to the perimeter fence.
Base defense is an edge AI problem
Look at what an installation actually needs: persistent multi-sensor watch (EO cameras, RF detection, acoustic), local fusion into a single threat picture, automated alerting with evidence attached, effector cueing under policy, and all of it running when the network to higher headquarters is slow, saturated, or down. CRS notes that guardrails and responsibilities remain unclear and inconsistent across installations. Meanwhile hundreds of bases, armories, and critical infrastructure sites will never get a program-of-record air defense battery. The defense they get is the one they can stand up themselves.
That is precisely the bottom-up capability EdgeLance exists to provide. A command node and a handful of cameras and RF sensors give a base defense operations cell local AI watch over the fence line today: detection, fusion, CoT publish to the TAK displays the security forces already run, and an evidence trail for every alert. It does not replace the layered kinetic defense the big installations are getting. It closes the gap between detecting a quarter of incursions and seeing all of them, at the hundreds of sites the programs of record will not reach this decade.