Senior Airman Zach Wilt, 49th Communications Squadron cyber operator, installs Microsoft Windows 10 to a laptop at Holloman Air Force Base, N.M., Nov. 1, 2017. Wilt is one of a few Airmen tasked with installing Windows 10, and various other programs such as Adobe and Java, to Holloman’s computers following the Air Force’s transition to a Microsoft Windows 10 secure host baseline. (U.S. Air Force by Airman 1st Class Alexis P. Docherty)
Cyber operator installs software to a laptop. Photo: Airman 1st Class Alexis P. Docherty/US Air Force

The US Air Force Research Laboratory has partnered with Washington-based startup EdgeRunner AI to test a generative artificial intelligence chatbot for the military’s nonclassified internal network.

The project builds on a December 2024 cooperative research and development agreement tasking the company with fine-tuning an AI-enabled prototype conversation system for air force use.

Once complete, the chatbot will be plugged into the military’s private network, the Non-Secure Internet Protocol Router Network or “NIPERNET,” where it will help manage and relay sensitive but unclassified information.

Built-In ‘Expert’

EdgeRunner’s chief technology officer and co-founder Colton Malkerson said the company is focused on delivering an algorithm trainable for specific mission sets.

This on‑device virtual assistant acts as a real-world expert tailored to specific roles such as pilot, logistics officer, or field medic, providing instant access to critical technical data and calculated decisions during operations.

“We’re trying to give every warfighter expert advice in a bottle on the fly to help them make better decisions faster and make better decisions more safely,” Malkerson told National Defense.

Additional Features

One standout feature? It works offline. EdgeRunner’s chatbot system is independent from the internet and does not share user locations, which is a major plus for security-conscious environments.

When activated, the system can absorb new data, pulling in fresh info not included in the original download. And for transparency, every recommendation comes with source citations so users can double-check the details.

To keep things sharp, the chatbot gets quarterly updates packed with “faster models, better features, more performance, more optimizations, more capabilities.”

“Even though you’re running it locally on-device and you’re not connected to the cloud at all times, you still effectively have frontier model performance,” Malkerson said.

AI Chatbot at Sea, DoD Devices

Currently, EdgeRunner is engaged with a separate project to assemble and field generative AI technologies across ships.

It also launched an effort that allows US Department of Defense users to install and use the company’s proprietary chatbot on their respective devices for free.

You May Also Like

South Korea Demos ‘World’s First’ Photonic Radar for Small Objects

South Korea showcases its AI-based photonic radar system, marking the first real-life test of this drone-detecting tech powered by lasers and AI.

British Army Debuts AI Targeting Spear for Modern Warfare

Asgard taps advanced communication networks to monitor threats in a matter of minutes, supporting rapid strike decisions that once “took hours.”

AI Beats Human Planners in DASH 3 Battle Planning Challenge

The event brought together US operators and coalition partners from Canada and the UK to see how AI tools could speed up and expand multi-domain decision-making.