AI Governance Assessment  ·  NIST AI RMF 1.0

⏱ Estimated time: 15–20 minutes total
Section 1 of 10

Organizational Profile

Provide foundational information about your organization to contextualize the AI risk assessment.

GOVERN · MAP · MEASURE · MANAGE

Select the option that best describes your organization's relationship with AI systems.

1.1 Organization Information GOVERN 1.1
1.2 Primary Contact Internal
Section 2 of 10

AI System Inventory & Classification

Identify and classify the AI system being assessed. Complete one assessment per AI system, or provide consolidated answers for an organization-wide assessment.

MAP 1: Context & Use Classification
2.1 AI System Identification MAP 1.1

3–5 sentences describing what the system does, its inputs, outputs, and primary function.

2.2 Purpose and Use Context MAP 1.1, 1.3, 1.4
2.3 AI Lifecycle Responsibility GOVERN 1.4
Section 3 of 10

Data Governance

Document the data inputs, training data characteristics, and quality controls for this AI system.

MAP 2: Data Context & Risk
3.1 Training & Input Data MAP 1.1, 2.3, GOVERN 6.1
Your AI vendor controls model training. Respond to the questions below based on the data your organization provides as input. Where vendor-managed items are not applicable to your operations, select Not applicable / managed by vendor.
3.2 Data Quality Controls MEASURE 2.5
Section 4 of 10

Stakeholder & Impact Mapping

Identify affected populations and characterize the potential harms and benefits of this AI system.

MAP 1 · MAP 5: Impact & Harm Assessment
4.1 Affected Populations MAP 5.1, 5.2, 1.2
4.2 Impact Characterization MAP 5.1, 5.2 Rate the likelihood and severity of each harm category relative to your AI system's use case. Select N/A for categories that are not applicable to your deployment context.

Rate the likelihood and potential severity of each harm category. Select N/A where a category is not applicable to your deployment context. View harm type definitions ▾

Rights & Legal, Outcomes that may impair an individual's legal rights, entitlements, or access to services (lending, hiring, benefits eligibility).
Fairness, Differential outcomes across demographic groups that may constitute or approximate disparate impact under applicable law.
Financial, Direct or indirect monetary losses to individuals, customers, or the organization attributable to AI system outputs.
Safety, Physical harm arising from AI-driven decisions or system failures; most applicable in healthcare, manufacturing, or autonomous systems.
Wellbeing, Emotional distress, reputational injury to individuals, or loss of autonomy resulting from AI outputs or automated decisions.
Reputational, Damage to organizational trust or brand equity from AI behavior, inaccurate outputs, or incidents made public.
Operational, Disruptions to business continuity or service availability caused by AI system failures or degraded performance.
Societal, Broader negative externalities including environmental impact, contribution to misinformation, or systemic community effects.
Security, Unauthorized access to or exfiltration of sensitive data facilitated by AI system vulnerabilities, including inadvertent exposure through commercial tool inputs.
Harm Type
Likelihood
Severity
Unfair treatment or denial of rights Rights & Legal
Biased or unequal outcomes for different groups of people Fairness
Financial harm to individuals or the organization Financial
Physical safety risk to people Safety
Psychological or emotional harm to individuals Wellbeing
Harm to the organization's reputation or trust Reputational
Disruption to business operations Operational
Environmental or broader societal harm Societal
Data breach or privacy violation through the AI system Security
4.3 Benefits Characterization MAP 5.2
Section 5 of 10

Trustworthiness Characteristics Assessment

Rate the current implementation maturity of each trustworthiness characteristic for this AI system.

MEASURE 2: Trustworthiness Evaluation
ⓘRate each item on a four-point scale. These ratings directly inform the risk posture sections of your generated policy.
5.1 Valid and Reliable Section 3.1
Are accuracy metrics defined and documented for this system?
Is the system tested against representative datasets that reflect real deployment conditions?
Are false positive/false negative rates tracked and reported?
Is there a defined acceptable performance threshold below which the system triggers a review?
Is model robustness tested across varied conditions and edge cases?
Is ongoing post-deployment performance monitoring in place?
5.2 Safe Section 3.2
Has a formal safety risk assessment been conducted for this system?
Are there defined conditions under which the system will be shut down or overridden?
Is there a human intervention mechanism for safety-critical outputs?
Are safety incidents documented and tracked?
Does the system have a graceful degradation mode if it exceeds its knowledge limits?
5.3 Secure and Resilient Section 3.3
Has the AI system been assessed for adversarial attack vulnerabilities (data poisoning, model evasion, model extraction)?
Are access controls in place to prevent unauthorized use or modification of the model?
Is the system included in the organization's incident response plan?
Are the model, training data, and outputs protected against exfiltration?
Has a threat model specific to this AI system been developed?
Is the system subject to penetration testing or red-teaming?
5.4 Accountable and Transparent Section 3.4
Is documentation available describing how the system works, what data it uses, and how outputs are generated?
Are end users or affected individuals informed that an AI system is involved in decisions that affect them?
Is there a mechanism for individuals to understand why a specific decision was made?
Are audit logs maintained for system inputs, outputs, and decision events?
Is there a designated accountable owner for this AI system?
Are there documented escalation paths when the system produces disputed or harmful outputs?
5.5 Explainable and Interpretable Section 3.5
Can the system produce explanations of its outputs in terms understandable to non-technical users?
Are explanation methods (e.g., SHAP, LIME, rule extraction) in place?
Are explanations provided to decision-makers using the system's outputs?
Is there documentation mapping system outputs to the inputs or features that drove them?
Does the organization distinguish between explainability (how the system works) and interpretability (what the output means in context)?
5.6 Privacy-Enhanced Section 3.6
Has a Privacy Impact Assessment (PIA) been conducted for this system?
Are data minimization practices in place (only collecting data necessary for the system's purpose)?
Are de-identification or anonymization techniques applied where applicable?
Are privacy-enhancing technologies (PETs) employed (e.g., differential privacy, federated learning)?
Does the system comply with applicable privacy regulations (GDPR, CCPA, HIPAA, etc.)?
Are individuals provided with rights to access, correct, or opt out of AI-driven decisions about them?
5.7 Fair, with Harmful Bias Managed Section 3.7
For commercial AI tools, bias responsibility is shared. Answer these questions based on steps your organization has taken to evaluate and mitigate bias in how the tool is used, not solely on what the vendor has implemented.
Has bias in this system been evaluated or tested? For commercial AI tools: includes reviewing the vendor's published bias and fairness documentation, or assessing outputs in your own use context for unequal treatment. For proprietary systems: includes formal disparity analysis or fairness testing.
Are fairness metrics defined and tracked (e.g., demographic parity, equalized odds)?
Are outputs disaggregated by demographic group to assess disparate impact?
Is there a process for surfacing and addressing bias complaints from users or affected individuals?
Has a disparity analysis been conducted comparing system performance across protected classes?
Section 6 of 10

Governance Structure

Assess the organizational policies, roles, risk tolerance, and workforce practices governing AI risk management.

GOVERN 1 · GOVERN 2 · GOVERN 6
6.1 Policies and Procedures GOVERN 1–6
6.2 Roles and Accountability GOVERN 2
6.3 Risk Tolerance MAP 1.5
6.4 Third-Party and Supply Chain Risk GOVERN 6
6.5 Workforce and Culture GOVERN 3, 4
Section 7 of 10

Testing, Evaluation, Verification & Validation

Document pre-deployment testing practices and ongoing production monitoring for this AI system.

MEASURE 2 · MEASURE 4: TEVV & Monitoring
7.1 Pre-Deployment Testing MEASURE 1
7.2 Ongoing Monitoring MEASURE 2.4, MANAGE 4.1
7.3 Metrics in Use MEASURE 2
Section 8 of 10

Incident History and Risk Response

Document any known adverse outcomes and the organization's risk treatment approach for this AI system.

MANAGE 2 · MANAGE 3: Incident & Risk Response
8.1 Known Incidents MANAGE 1–4
8.2 Risk Treatment MANAGE 1.2, 1.3
Section 9 of 10

Legal and Regulatory Compliance

Identify applicable legal requirements and document your organization's compliance posture for this AI system.

GOVERN 1 · GOVERN 4: Legal & Regulatory

Select federal, international, and sector-specific frameworks that apply to your organization and this AI system.

Federal: Privacy & Consumer Protection

Federal: Government & Security

International

State AI & Privacy Laws

Other / Specify

Section 10 of 10

Open-Ended Supplemental Questions

These questions capture nuances not covered by the structured assessment. Your responses directly shape the custom policy provisions generated for your organization.

All RMF Functions: Supplemental Context

Include emerging risks, novel use cases, or gaps you have identified.

This is your opportunity to add context, priorities, or specific policy language requirements.

Section 1 of 10
Changes saved automatically