The Pentagon has issued a new instruction, approved by CIO Kirsten Davies and effective Sept. 8, that establishes policy and procedures for using AI in software development as part of its move toward “software-defined warfare.” The 37-page guidance, titled “Accelerated Mission Software,” emphasizes software modernization, enterprisewide reuse, and faster delivery of capabilities across the Department of Defense.
Officials describe AI-assisted development as a major force multiplier for speed and quality, and direct DOD components to use AI to automate development, improve system integration, and strengthen resilience and security. At the same time, the instruction places clear limits and accountability requirements on AI use. Developers and teams remain fully responsible for the security, functionality, and integrity of any AI-generated or AI-modified code, and human review is required for security- or safety-critical changes.
All AI-suggested code must undergo the same review and security testing as manually written code, including checks for vulnerabilities, safety issues, logical errors, bugs, intellectual property concerns, license obligations, and proper security controls. Components must also define risk-mitigation requirements, verify benchmarks before deployment, and monitor performance continuously.
To improve transparency, teams must keep records of models, versions, and significant datasets used in software generation or testing. The guidance also calls for tools to detect unintended bias and restricts the use of sensitive non-public DOD data in unapproved generative AI systems. Approved services must contractually prevent government data and prompts from being used to train external models. Training for development teams is also required.
Article timeline
Build this article’s timeline
Extract dated events, key actors, and open questions using the same timeline analysis available in Signal.
Comments
Login to view and post comments.