Compliance Guide

NIST Cyber AI Profile: What Enterprises Need to Know

The NIST Cyber AI Profile refers to an NCCoE-led effort to map AI-specific risks onto the NIST Cybersecurity Framework (CSF) 2.0, building on the AI Risk Management Framework and Generative AI Profile. It is a voluntary crosswalk, not a new regulatory standard, and enterprises should use it to connect existing cybersecurity controls, including identity and access management, to AI and AI agent risk management rather than build a separate compliance track.

NIST AI Governance Landscape

The Cyber AI Profile does not stand alone. It sits within a small set of related NIST publications that build on one another, summarized below.

AI RMF 1.0

Core risk management functions: Govern, Map, Measure, Manage.

Generative AI Profile

NIST AI 600-1, addressing risks specific to generative AI systems.

CSF 2.0

Adds a Govern function and supports AI-focused Community Profiles.

Cyber AI Profile (NCCoE)

Crosswalk mapping AI risk to existing CSF categories and subcategories.

NIST Cyber AI Profile: Full Guide

The term "NIST Cyber AI Profile" refers to an ongoing effort led by NIST's National Cybersecurity Center of Excellence (NCCoE) to produce AI-related Community Profiles under the Cybersecurity Framework (CSF) 2.0. Rather than introducing a new set of technical controls, this work maps AI-specific risk considerations onto the six CSF 2.0 functions: Govern, Identify, Protect, Detect, Respond, and Recover. It is designed to extend an organization's existing cybersecurity governance structure to AI systems, rather than require a parallel compliance program.

This profile builds directly on two earlier NIST publications. The AI Risk Management Framework (AI RMF 1.0), released in January 2023, established four core functions for managing AI risk: Govern, Map, Measure, and Manage. The Generative AI Profile (NIST AI 600-1), published as a companion resource, identified risks specific to or amplified by generative AI, including confabulation, data privacy leakage, and value chain integration risks. The Cyber AI Profile effort connects these AI-specific risk concepts to the cybersecurity control language enterprises already use under CSF 2.0.

As of current publicly available NIST documentation, there is no single finalized publication with formal subcategory numbering equivalent to standard CSF Informative References. Governance leaders should treat this as an evolving, community-driven mapping project rather than a fixed, versioned standard, and confirm current status directly against NIST's published project pages before finalizing internal compliance language.

How It Relates to CSF 2.0 and the AI RMF

CSF 2.0, published in February 2024, introduced a dedicated Govern function and formally supports Community and Organizational Profiles, which allow the framework to be tailored to specific technologies, sectors, or risk contexts. AI-related Community Profiles, including the Cyber AI Profile effort, are one application of this mechanism. They are explicitly voluntary and non-regulatory, functioning as implementation guidance rather than mandatory compliance requirements.

The AI RMF and the Cyber AI Profile serve complementary but distinct purposes. The AI RMF provides a risk management lifecycle, guiding organizations through governance, system mapping, measurement, and ongoing management of AI risk. The Cyber AI Profile, by contrast, is oriented toward cybersecurity control mapping, showing how AI risk considerations align with existing CSF categories such as asset management, identity management, and access control. Enterprises should not treat these as interchangeable: AI RMF outputs, such as system inventories and risk documentation from the Map function, become inputs for aligning with CSF-based profiles.

How the two frameworks divide responsibility
FrameworkPrimary orientationWhat it produces
AI RMFRisk management lifecycleSystem inventories, risk documentation, governance ownership
Cyber AI Profile (CSF 2.0)Cybersecurity control mappingAlignment of AI risk to existing CSF categories, including identity and access control

Control Areas Relevant to AI Agent Runtime Risk

Current NIST publications do not yet define AI agent-specific subcategories with formal control numbering for identity attestation or tool-call authorization. The closest existing anchor points are found in CSF 2.0's Identify and Protect functions, particularly the Protect function's identity management and access control subcategories (PR.AA), which many enterprises already extend to non-human or machine identities.

For AI agents specifically, this means governance leaders must currently interpret general identity and access management guidance in the context of autonomous or semi-autonomous software rather than rely on AI agent-specific NIST control language. The AI RMF's Map function is relevant here as well, since it requires organizations to document AI system context, including third-party components, models, and integrated tools. This documentation requirement is a reasonable anchor point for tracking AI agent tool-call and plugin risk, even though NIST has not published a dedicated technical specification for agent tool-call authorization schemas.

Interim mapping, not a finished standard

Enterprises should document how their existing identity and access management (IAM) and privileged access management (PAM) programs apply to agent identities today, and treat this as an interim mapping subject to revision as NIST guidance matures.

Assessing Governance Maturity Against the Profile

Because the Cyber AI Profile is a mapping exercise rather than a prescriptive control set, maturity assessment starts with the AI RMF's four functions. Organizations should first complete a Map exercise: inventorying AI systems and agents in use, documenting their purpose, data flows, and third-party or tool dependencies. This inventory becomes the foundation for identifying which CSF categories apply and where gaps exist.

Following inventory, the Govern function requires organizations to establish policy-level accountability for AI risk decisions, typically requiring cross-functional ownership across security, AI governance, and legal teams. The Measure and Manage functions then guide ongoing evaluation and response to identified AI risks, which should be integrated into existing cybersecurity risk registers rather than maintained as a separate compliance track.

Governance leaders in regulated industries should also confirm whether sector-specific regulators reference the AI RMF or CSF AI profiles directly, since these NIST frameworks remain voluntary at the federal level absent such sector-specific adoption.

Where Runtime Governance Tooling Fits

Because current NIST guidance places the closest AI agent-relevant control expectations under CSF Protect function subcategories for identity and access management, enterprises implementing AI agents need practical mechanisms to enforce and evidence these controls at runtime. This includes maintaining distinct agent identities, applying least privilege to agent permissions, governing tool approval workflows, and producing audit logs of agent activity and tool calls.

Trussed AI provides runtime governance and security for enterprise AI agents, including agent identity management, permission enforcement, tool approval workflows, and audit logging. These capability areas map directly to the Identify and Protect functions that current NIST AI-related CSF guidance references, and to the system documentation requirements under the AI RMF's Map function. Organizations evaluating how to operationalize the Cyber AI Profile's guidance should treat runtime governance tooling as one component supporting evidence generation and control enforcement, not as a substitute for the underlying risk management program.

Getting Started: A Practical Checklist

The following steps translate the frameworks above into a sequence enterprises can act on now, without waiting for a finalized Cyber AI Profile publication.

  • Inventory all AI systems and agents in production, per the AI RMF Map function, including third-party models and tools.
  • Review existing IAM and PAM programs for applicability to non-human and agent identities.
  • Assign cross-functional ownership for AI risk governance across security, AI governance, and legal teams.
  • Document current interim controls for agent identity, permissions, and tool-call activity where formal NIST subcategories do not yet exist.
  • Integrate AI risk documentation into existing cybersecurity risk registers rather than a separate compliance system.
  • Track NCCoE project updates, since AI-specific CSF profiles remain under active development.

Operationalize AI Agent Governance

Align your AI agent identity, permissions, and audit logging practices with NIST guidance through runtime governance controls.

Request a Demo