1.1 — Overview and Purpose of NIST AI RMF
The NIST AI Risk Management Framework (AI RMF 1.0) was released in January 2023 by the National Institute of Standards and Technology. It provides a structured, flexible approach for organizations to manage risks associated with AI systems throughout their lifecycle.
Unlike prescriptive regulations, the AI RMF is a voluntary framework designed to be adopted by organizations of any size, sector, or jurisdiction. It complements existing risk management practices and can be mapped to other frameworks like ISO 42001 and the EU AI Act.
The NIST AI RMF is voluntary — it is not a law or regulation. However, it has become a de facto standard referenced by U.S. executive orders and is increasingly expected by regulators and customers. Know the distinction between 'voluntary framework' and 'mandatory regulation' for exam questions.
The framework defines 'trustworthy AI' across seven characteristics. These characteristics are not independent — they interact and may require trade-offs. For example, increasing explainability may reduce accuracy in certain model architectures. The framework emphasizes that organizations must balance these characteristics based on context.
- Valid and Reliable — AI systems perform as intended and consistently produce accurate outputs under expected and challenging conditions.
- Safe — AI systems do not endanger human life, health, property, or the environment under defined conditions of use.
- Secure and Resilient — AI systems withstand adversarial attacks, unexpected inputs, and environmental changes while maintaining confidentiality and integrity.
- Accountable and Transparent — Clear accountability structures exist, and information about the AI system is available to appropriate stakeholders.
- Explainable and Interpretable — Stakeholders can understand AI system outputs (explainability) and the mechanisms producing them (interpretability).
- Privacy-Enhanced — AI systems protect privacy through design, data minimization, and appropriate handling of personal and sensitive information.
- Fair — with Harmful Bias Managed — AI systems are designed to identify, assess, and mitigate harmful biases across the lifecycle.
The AI RMF is structured around two main parts: Part 1 discusses foundational information about AI risks and trustworthiness, while Part 2 introduces the Core — a set of four functions, categories, and subcategories that guide risk management activities. The Core is hierarchical: Functions contain Categories, which contain Subcategories, each with suggested actions documented in the companion Playbook.