The National Institute of Standards and Technology (NIST) Artificial Intelligence Risk Management Framework (AI RMF 1.0) is a voluntary, non-sector-specific, and use-case-agnostic resource designed to help organizations manage the risks of artificial intelligence and promote trustworthy, responsible development.
1. Framing Risk & Characteristics of Trustworthy AI
Risk Definition: Risk is a composite measure of the probability of an event occurring and the magnitude of its consequences (positive or negative). AI risks can impact individuals, groups, organizations, society, the environment, and the planet.
Trustworthy Characteristics: To minimize negative risks, AI systems should embody seven core characteristics:
Valid and Reliable: Performing accurately and reliably as required.
Safe: Not endangering human life, health, property, or the environment.
Secure and Resilient: Withstanding unexpected adverse events and maintaining protection against attacks.
Accountable and Transparent: Providing appropriate information and clear oversight structures.
Explainable and Interpretable: Ensuring underlying operations and outputs can be understood.
Privacy-Enhanced: Safeguarding human autonomy, identity, and dignity.
Fair with Harmful Bias Managed: Addressing equality, equity, and systematic, computational, or human-cognitive biases
2. The Core Functions
The framework is structured around four high-level, continuous functions that guide organizational activities:
*Govern: Cultivates and implements an organizational culture of risk management, defining accountability structures, workforce diversity, and third-party policies.
*Map: Establishes the context to frame risks, identifying intended purposes, legal norms, system categorization, and potential impacts.
*Measure: Employs quantitative, qualitative, or mixed methods (such as Test, Evaluation, Verification, and Validation - TEVV) to analyze, assess, and track AI risks and trustworthiness.
*Manage: Allocates resources to prioritize, respond to, and treat mapped and measured risks, maintaining post-deployment monitoring and incident recovery.
Note: NIST is currently updating the framework
(Generated by Gemini, which can make mistakes)
To download the full framework:https://lnkd.in/dc8frw25
To download the full framework:https://lnkd.in/dc8frw25
No comments:
Post a Comment