What is a NIST AI RMF profile and how do I build one for an AI agent deployment?

A NIST AI RMF profile is a tailored application of the NIST AI Risk Management Framework to a specific AI system, such as an AI agent deployment, which helps an organization understand, assess, and manage AI risks. Building one involves addressing the four core functions of the AI RMF: Govern, Map, Measure, and Manage.

To build a NIST AI RMF profile for an AI agent deployment, consider the following controls:

• Govern: Establish an acceptable-use policy for AI/agent systems covering permitted and prohibited uses, and human-oversight expectations (NIST-GOVERN-1.1). Assign a named risk owner or accountable executive for each deployed AI/agent system (NIST-GOVERN-2.1). Define policies for human oversight of AI, including override authority and the boundary of agent autonomy (NIST-GOVERN-3.2). Address risks from third-party models, datasets, and tools, including provenance, licensing, and model-update risk (NIST-GOVERN-6.1), which also maps to OWASP LLM03/LLM05.
• Map: Document the intended purpose, deployment setting, and operating context of each AI/agent system (NIST-MAP-1.1). Maintain and keep current an inventory of AI/agent systems, including models, agents, tools, and data flows (NIST-MAP-1.5). Identify potential positive and negative impacts to individuals, groups, and society, including data-sensitivity and regulated-data exposure (NIST-MAP-5.1).
• Measure: Evaluate and document AI system security and resilience, including adversarial robustness, prompt-injection resistance, and abuse resistance (NIST-MEASURE-2.7), which cross-maps to OWASP LLM01 and LLM04. Implement mechanisms to log decisions and trace AI behavior (NIST-MEASURE-2.8). Establish approaches for tracking identified and emergent risks through monitoring, logging, and drift detection (NIST-MEASURE-3.1), which cross-maps to ISO/IEC 42001 monitoring.
• Manage: Plan, track, and resource responses to the highest-priority AI risks (NIST-MANAGE-1.3). Establish procedures to deactivate, roll back, or retire AI systems that exceed risk tolerances (NIST-MANAGE-2.3). Implement post-deployment monitoring and an AI/agent incident-response plan covering detection, escalation, containment, communication, and learning (NIST-MANAGE-4.1).

The Generative-AI Profile (NIST AI 600-1) also highlights agent-relevant risks such as confabulation/hallucination, information security (e.g., prompt injection, data exfiltration, insecure tool use, mapping to OWASP LLM01/LLM02/LLM06), data privacy (e.g., leakage of sensitive/PII data, mapping to OWASP LLM02), dangerous/harmful content, and value-chain/component integration (e.g., third-party model/tool provenance, mapping to NIST-GOVERN-6.1, OWASP LLM03/LLM05).