The Pentagon wants to know who checks code when AI writes it (about.gitlab.com)

The Pentagon has issued a new directive, DoW Instruction 8430.01, effective September 8, 2026, that mandates stringent oversight of AI-generated code within the Department of War (DoW). This instruction classifies AI-generated code as unverified until proven otherwise, requiring organizations to review, test, and document all AI contributions at the same level as human-generated code. This significant policy change underscores the DoW's commitment to ensuring software security and accountability, especially as AI becomes increasingly integrated into software development processes. The directive emphasizes three conditions for using AI effectively and safely: tracing AI's involvement, verifying that its code is reliable, and safeguarding government data within secure boundaries. This requirement mandates that teams utilizing platforms like GitLab implement rigorous auditing and approval processes for AI-assisted code changes, ensuring that accountability rests with developers. By leveraging tools that maintain a clear audit trail and automate testing while adhering to established security protocols, the DoW aims to harness AI's capabilities without compromising on integrity and safety. As AI continues to evolve in its ability to write code, demonstrating trustworthiness and compliance with these new standards will be crucial for the defense sector.