TLDR: Seattle-based startup EdgeRunner AI has signed an agreement with the Air Force Research Laboratory to advance NIPRGPT, a generative AI chatbot designed for military applications. This initiative focuses on providing secure, on-device AI assistance tailored for various military roles, ensuring data privacy and operational security by operating independently of the internet.
WASHINGTON, D.C. – Seattle-based startup EdgeRunner AI has entered into a cooperative research and development agreement with the Air Force Research Laboratory (AFRL) to further the development of NIPRGPT, a cutting-edge generative artificial intelligence initiative. The agreement, awarded in December, aims to leverage generative AI on a private, non-classified internal network for transmitting sensitive information within the Air Force. The news was reported on July 14, 2025.
EdgeRunner AI specializes in creating air-gapped, on-device AI assistants equipped with occupation-specific adapters. These adapters are meticulously trained on the nuances of particular military roles, such as combat medics, logistics officers, or pilots. According to Colton Malkerson, cofounder and COO of EdgeRunner, the company’s platform is designed to provide immediate answers to technical questions, thereby supplementing knowledge and significantly saving time for military personnel. Malkerson articulated the company’s mission, stating, “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.” This statement was made during an interview at the recent AI+ Expo.
In its collaboration with the AFRL, EdgeRunner is focused on developing “domain-specific generative AI deployed at the edge,” which is customized for the Air Force’s diverse occupational specialties. A critical feature of EdgeRunner’s system is its independence from the internet. Malkerson emphasized that this design choice ensures “increased data security, data privacy, performance and operation security.” He further highlighted the security advantage, noting that users’ locations are not shared, unlike “satellite-based system or network-based system” that could ping a user’s location, making them “immediately a target.”
Also Read:
- Generative AI: A Transformative Force in Australian Defence and Global Military Strategy
- CellBot Revolutionizes Medical Research with AI on CellVerse Platform
The EdgeRunner team is committed to fine-tuning the AI assistant’s responses, rigorously testing for quality and aiming for the model to “think, act and behave like a military officer.” The system also allows for the upload of documents not initially included and provides citations for all answers, ensuring traceability and further reference for users.
