Responsible Artificial Intelligence (RAI) and Artificial Intelligence (AI) Governance Tools
Abstract
It is incumbent on the Department to ensure all its AI-enabled systems will be safe and adhere to ethical standards and that they are used in a manner that contributes to the efficiency, effectiveness, and legitimacy of the Department’s AI capabilities as outlined in the National Defense Strategy and Section 1513 of the National Defense Authorization Act (NDAA) for 2023. This requirement funds activities to develop, procure and maintain the necessary commercial and DoD-customized tools to put the DoD AI Ethical Principles into practice across the entire AI product lifecycle. RAI leaders across the Department are going to need access to utilize tools that support the capability development process, including in areas of explainability of models, bias detection, and other areas necessary for RAI. This requirement provides for the integration of commercially available tools to include an Explainable AI tool, Synthetic Data & Anonymization tool, Data Management and Traceability tool, Continuous Integration/Continual Delivery tool, Auto-Machine Learning tool, Bias Mitigation tool, and the Interference Time tool into CDAO’s AI/ML platforms, which will embed the operationalization of the DoD AI Ethics Principles into the developer’s workflow. By doing so, these tools will not only provide technical assessments through the project lifecycle, but also allow for traceability and reliability to ensure safe and ethical systems from design to development and deployment to use. This requirement also funds the development of DoD- customized assessments and tools that RAI leads will use across the AI product and acquisition lifecycles.
Document Details
- Document Type
- Accomplishment
- Publication Date
- Oct 01, 2024
- Source ID
- 6cbd14f80ebdab705748c6e2698d1fe6