AI Model Card

AI Model Card

Applicability and Documentation Profile

Model Details

Model Name:
[________________________________]

Version & Identifier:
[________________________________]

Model Type:
☐ Classification
☐ Regression
☐ Language Model (LLM)
☐ Computer Vision
☐ Recommendation System
☐ Autonomous Agent
☐ Other: [________________________________]

Organization & Owner:
[________________________________]

Date Released:
[__/__/____]

Model Architecture:
[________________________________]

Framework & Libraries:
[________________________________]

Hardware Requirements:
[________________________________]

Software/Firmware Dependencies:
[________________________________]

Intended Use

Primary Use Case:
[________________________________]

Intended Users/Stakeholders:
[________________________________]

Operational Context (e.g., consumer-facing, internal automation, embedded system):
[________________________________]

Geographic/Jurisdictional Scope:
[________________________________]

Decision Impact Level:
☐ Non-material (advisory, informational)
☐ Moderate (affects resource allocation, eligibility screening)
☐ High impact under the organization's risk taxonomy — legal classification separately recorded above

Foreseeable High-Impact Uses (e.g., hiring, loan approval, criminal justice, child safety):
[________________________________]

Out-of-Scope Uses (explicitly NOT intended for):
[________________________________]

Factors & Input Features

Input Feature List:

Demographic Variables in Model:
☐ Age
☐ Gender/Sex
☐ Race/Ethnicity
☐ Geographic Location
☐ Disability Status
☐ Other Protected Characteristics: [________________________________]

Missing Data Handling:
[________________________________]

Preprocessing & Feature Engineering:
[________________________________]

Metrics & Performance

Overall Performance Metrics:

Disaggregated Performance (Subgroup Analysis):

Intersectional Performance:
[Describe performance across intersecting demographic groups (e.g., older women, younger men, geographic + income combinations)]

Fairness Metrics Assessed:
☐ Demographic Parity
☐ Equalized Odds
☐ Predictive Parity
☐ Calibration
☐ Other: [________________________________]

Evaluation Data

Evaluation Dataset Name & Source:
[________________________________]

Dataset Size:
[____] samples

Time Period Covered:
[__/__/____] to [__/__/____]

Geographic Coverage:
[________________________________]

Data Split Methodology:
☐ Random stratified split ([____]% train / [____]% validation / [____]% test)
☐ Temporal split (cutoff date: [__/__/____])
☐ Other: [________________________________]

Inclusion/Exclusion Criteria:
[________________________________]

Known Limitations or Biases in Evaluation Data:
[________________________________]

Training Data

Training Dataset Name:
[________________________________]

Data Source(s):
[________________________________]

Collection Period:
[__/__/____] to [__/__/____]

Total Records:
[____] samples

Data Types Included:
☐ Structured (tabular)
☐ Unstructured (text, images, audio)
☐ Mixed

Demographic Composition:

Data Preprocessing & Cleaning:
[________________________________]

Data Anonymization/De-identification:
☐ Yes
☐ No
☐ Partial

Proxy Variables (indirect indicators of protected characteristics):
[________________________________]

Reviewed Data Authority and Restrictions:

Data Retention Period:
[________________________________]

Quantitative Analyses

Confusion Matrix (Classification Models):

Sensitivity Analysis (feature importance, input perturbations):
[________________________________]

Error Analysis (common failure modes, systematic errors):
[________________________________]

Stress Tests / Adversarial Robustness:
[________________________________]

Drift & Degradation Monitoring:
☐ Data drift monitoring in place
☐ Model performance monitoring in place
☐ Retraining frequency: [________________________________]

Ethical Considerations

Fairness Risks Identified:
[________________________________]

Mitigation Measures (technical, process, human oversight):
[________________________________]

Transparency & Explainability:
☐ Model predictions are explainable to end users
☐ Feature importance documented
☐ Uncertainty quantified
☐ Explanation method: [________________________________]

Accountability Mechanisms:
☐ Human review process established
☐ Audit trail maintained
☐ Appeal/recourse process available
☐ Details: [________________________________]

Potential for Discriminatory Outcomes:
[Assess likelihood of disparate impact on protected classes or vulnerable populations]

Secondary/Unintended Uses & Misuse Risk:
[________________________________]

Consent & Transparency with Subjects:
☐ Individuals informed of automated decision-making
☐ Right to explanation provided
☐ Opt-out mechanisms available
☐ Other: [________________________________]

Environmental & Resource Consumption:
[Describe training computational cost, inference latency, model size, energy use]

Caveats & Recommendations

Known Limitations:
[________________________________]

Performance Degradation Scenarios:
[E.g., out-of-distribution inputs, concept drift, demographic shift]

Recommended Monitoring & Governance:
☐ Performance review at this risk- and requirement-based cadence: [________________________________]
☐ Fairness or impact review at this risk- and requirement-based cadence: [________________________________]
☐ Real-time data monitoring
☐ Incident response protocol
☐ Retraining trigger thresholds: [________________________________]

Stakeholder Communication Plan:
[Describe how users, subjects, and regulators will be informed of material changes, incidents, or discontinuation]

Version Control & Change Log:

Framework and Regulatory Mapping

NIST AI RMF 1.0 Documentation Map

AI RMF Functions Addressed:

☐ GOVERN: Governance and accountability evidence: [________________________________]

☐ MAP: Context and risk-mapping evidence: [________________________________]

☐ MEASURE: Assessment, testing, and monitoring evidence: [________________________________]

☐ MANAGE: Prioritization, treatment, response, and improvement evidence: [________________________________]

Regulation (EU) 2024/1689 Article 11 / Annex IV Map

Applicability gate: Complete this section only when the reviewed profile determines that Article 11 applies to the high-risk AI system. A model card alone is not the full Annex IV technical documentation.

☐ General system description and instructions for use — evidence location: [________________________________]

☐ Development methods, design specifications, architecture, resources, data, oversight, validation/testing, and cybersecurity — evidence location: [________________________________]

☐ Monitoring, capabilities, limitations, accuracy, foreseeable unintended outcomes, discrimination risks, and input specifications — evidence location: [________________________________]

☐ Performance-metric appropriateness and risk-management system — evidence location: [________________________________]

☐ Lifecycle changes, harmonised standards/common specifications, declarations, and post-market monitoring plan — evidence location: [________________________________]

☐ Documentation drawn up before placement on the market or putting into service and kept up to date — owner / process: [________________________________]

Risk Management Plan:
[Reference internal risk register or describe high-level approach]

Post-Market Monitoring Plan:
[Describe how performance will be tracked after deployment; monitoring frequency, responsible parties, escalation procedures]

Retention & Update Schedule:
Documentation will be retained and reviewed under the approved schedule recorded in the Applicability and Documentation Profile: [________________________________].

Attestation & Approvals

Prepared By:
Name: [________________________________]
Title: [________________________________]
Date: [__/__/____]
Signature: [________________________________]

Technical Review (Model Developer/Data Scientist):
Name: [________________________________]
Date: [__/__/____]
Approval: ☐ Approved ☐ Approved with conditions ☐ Not approved

Compliance/Legal Review:
Name: [________________________________]
Date: [__/__/____]
Approval: ☐ Approved ☐ Approved with conditions ☐ Not approved

Executive Sponsor/Model Owner:
Name: [________________________________]
Date: [__/__/____]
Approval: ☐ Approved ☐ Approved with conditions ☐ Not approved

Comments & Conditions:
[________________________________]

Sources & References

• NIST, Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1 — https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.100-1.pdf
• European Commission AI Act Service Desk, Article 11 text — https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-11
• European Commission AI Act Service Desk, Annex IV text — https://ai-act-service-desk.ec.europa.eu/en/ai-act/annex-4

Document Classification: [Internal / Regulated / Public]
Next Review Date: [__/__/____]
Questions? Contact: [________________________________]