The U.S. Air Force turns to GenAI to refactor millions of lines of outdated code and modernize defense systems faster.
The Department of the Air Force is using GenAI to modernize its aging software systems written in languages like COBOL and Fortran. The Air Force Life Cycle Management Center’s Bot Operations Team (DAFBOT) has successfully refactored legacy code into modern programming languages like Java using large language models (LLMs). These AI tools enable faster code translation and help overcome a decades-long backlog of technical debt.
One of the main challenges military software teams face is the complexity and scale of old codebases. Some exceeding 15 million lines—often without documentation. Traditional updates require extensive manual work, but LLM-powered code assistants can now understand and restructure obsolete code with contextual accuracy. The AI rewrites code while maintaining the original logic. Much like translating ancient texts into modern language without losing meaning.
Security remains a key concern. Many generative models require large-scale cloud infrastructure and external data access. In which they are incompatible with the classified environments where military code resides. To address this, teams like The Forge are exploring on-premise models that can run securely within defense networks. AI use is also being guided by a modular approach—gradually replacing system parts rather than complete overhauls, which minimizes risk and maximizes control.
By leveraging GenAI, the Air Force is accelerating modernization efforts and reducing dependency on legacy systems that slow innovation. With successful pilots now underway, the Department of Defense is aiming for broader adoption to help transform mission-critical platforms in the years ahead.