aigovdev/governance-risk-classifier
Tabular Classification • Updated
id stringlengths 7 7 | use_case stringlengths 25 65 | sector stringlengths 8 23 | impact stringclasses 4
values | decision_autonomy stringclasses 3
values | human_oversight stringclasses 3
values | monitoring stringclasses 3
values | traceability stringclasses 3
values | technical_documentation stringclasses 2
values | governance_risk stringclasses 4
values | primary_gap stringclasses 5
values | notes stringlengths 50 88 |
|---|---|---|---|---|---|---|---|---|---|---|---|
gov_001 | Automated creditworthiness assessment | financial_services | high | automated | partial | partial | partial | strong | high | human_oversight | High-impact automated decision with incomplete oversight and monitoring. |
gov_002 | Internal document summarization assistant | enterprise_productivity | low | assistive | strong | strong | strong | strong | low | none | Assistive use with documented controls and review. |
gov_003 | Automated eligibility decision for access to an essential service | public_services | critical | automated | none | partial | partial | partial | unacceptable | human_oversight | Directly affects individuals while lacking effective human oversight. |
gov_004 | AI-assisted fraud investigation prioritization | financial_services | high | recommendation | strong | strong | strong | partial | limited | technical_documentation | Strong operational controls with incomplete technical documentation. |
gov_005 | Customer support response drafting | customer_service | low | assistive | strong | partial | partial | strong | limited | monitoring | Low-impact assistive system with some monitoring and traceability gaps. |
gov_006 | Automated employee performance scoring | employment | high | automated | partial | none | partial | partial | high | monitoring | High-impact employment use with missing post-deployment monitoring. |
gov_007 | Clinical decision support recommendation | healthcare | critical | recommendation | strong | strong | strong | strong | limited | none | Critical domain but strong human oversight, monitoring, traceability, and documentation. |
gov_008 | Automated insurance claims rejection | insurance | high | automated | none | partial | none | partial | unacceptable | traceability | Automated adverse decision without effective oversight or reliable traceability. |
gov_009 | Marketing copy generation | marketing | low | assistive | strong | partial | partial | partial | limited | technical_documentation | Low-impact generative use with lightweight governance controls. |
gov_010 | Automated university admissions ranking | education | high | automated | partial | partial | strong | partial | high | human_oversight | High-impact ranking process with only partial oversight. |
gov_011 | AI coding assistant for internal developers | software_engineering | medium | assistive | strong | strong | partial | strong | limited | traceability | Human-reviewed outputs with incomplete traceability of generated suggestions. |
gov_012 | Autonomous access-control decision system | security | critical | automated | none | none | none | partial | unacceptable | human_oversight | Critical autonomous system without essential governance controls. |
AI Governance Scenarios
A small synthetic dataset of AI-system deployment scenarios annotated with governance-control and risk signals.
The dataset accompanies the AIGov AI Governance Lab Hugging Face Space.
Purpose
The dataset is designed for:
- AI governance prototyping
- governance-control evaluation
- risk-analysis experiments
- testing deterministic governance heuristics
- responsible AI engineering demonstrations
- educational and portfolio use
It is not intended as a legal or regulatory benchmark.
Schema
Each record contains:
| Field | Description |
|---|---|
| id | Unique scenario identifier |
| use_case | Description of the AI system use case |
| sector | Deployment sector |
| impact | Potential impact level |
| decision_autonomy | Degree of AI decision autonomy |
| human_oversight | Human oversight maturity |
| monitoring | Post-deployment monitoring maturity |
| traceability | Traceability maturity |
| technical_documentation | Documentation maturity |
| governance_risk | Synthetic governance-risk label |
| primary_gap | Main governance weakness |
| notes | Short annotation explanation |
Governance signals
The examples cover signals including:
- human oversight
- decision autonomy
- post-deployment monitoring
- traceability
- technical documentation
- system impact
Data creation
All examples are synthetic and manually constructed for this project.
They are not real compliance assessments and do not represent legal conclusions about any organization, product, or deployment.
Intended use
The dataset can support experiments involving:
- governance scenario classification
- risk-control mapping
- responsible AI baselines
- governance dashboards
- control-gap detection
- evaluation of governance heuristics
Relationship to AIGov AI Governance Lab
The associated AIGov AI Governance Lab converts similar observable system characteristics into a deterministic governance-engineering assessment.
This dataset provides controlled examples for testing and extending that approach.
Limitations
The dataset is intentionally small and simplified.
The governance_risk labels are engineering annotations, not regulatory classifications.
No label should be interpreted as determining compliance with the EU AI Act or any other legal regime.
Future work
Potential extensions include:
- larger scenario sets
- structured control evidence
- lifecycle events
- auditability signals
- incident and monitoring records
- human-review workflows
- governance-control taxonomies
- benchmark evaluation
AIGov
AIGov builds infrastructure for transparent, auditable, and accountable AI systems.
Website: https://govbase.dev
License
Apache-2.0
- Downloads last month
- 15
Models trained or fine-tuned on aigovdev/ai-governance-scenarios