AI Risk & Governance
When to Use
- Classifying AI use cases by risk tier and impact
- Drafting AI acceptable-use policies and governance frameworks
- Building AI risk registers with likelihood/severity/mitigation tracking
- Preparing model cards, system cards, or DPIAs for AI deployments
- Reviewing third-party LLM vendors (data terms, fine-tuning, safety commitments)
- Mapping AI products to frameworks (NIST AI RMF, ISO 42001, EU AI Act concepts)
- Aligning product and engineering teams with compliance requirements
- Designing human-in-the-loop oversight for consequential AI decisions
When NOT to Use
- Implementing RAG pipelines, agents, or production features →
ai-engineer
- Running jailbreak tests or adversarial campaigns →
ai-redteam
- SOC 2/ISO evidence automation and technical control mapping →
compliance-engineer
- General SOC 2 IT controls without AI scope →
cybersecurity
- Commercial/enterprise AI solution architecture →
applied-ai-architect-commercial-enterprise
- Skills portfolio governance and publish gates →
ai-skill-manager
Related skills
| Need |
Skill |
| Building LLM products |
ai-engineer |
| Adversarial testing |
ai-redteam |
| Research and benchmarks |
ai-researcher |
| Enterprise security program |
cybersecurity |
| Pipeline and data security |
devsecops |
| SOC 2/ISO evidence and technical controls |
compliance-engineer |
| AI solution architecture (commercial/enterprise) |
applied-ai-architect-commercial-enterprise |
| Agent skills governance |
ai-skill-manager |
| Safeguard gateways, classifiers, rollout |
ml-infrastructure-engineer-safeguards |
| Safety classifier research and benchmarks |
ml-research-engineer-safeguards |
| Privacy research for safeguards |
privacy-research-engineer-safeguards |
| Enterprise security risk registers (non-AI scope) |
security-risk-analyst |
| M&A/investment cyber diligence and board packs |
cyber-diligence-governance |
Core Workflows
1. Use-case intake and classification
Capture:
| Field |
Purpose |
| Purpose and users |
Scope and accountability |
| Data types |
PII, special categories, IP |
| Automation level |
Human-in-loop vs autonomous |
| Impact if wrong |
Safety, legal, financial, reputational |
| External exposure |
Customer-facing vs internal |
Risk tier (example):
| Tier |
Criteria |
Controls |
| Low |
Internal, low impact, no sensitive data |
Standard policy + logging |
| Medium |
Customer-facing or internal PII |
Review + eval + monitoring |
| High |
Regulated domain, high impact, autonomous actions |
Governance board + red-team + enhanced oversight |
See references/risk_classification.md for EU AI Act–oriented mapping (non-legal).
2. Risk assessment
Use structured worksheet:
- Identify hazards (bias, hallucination, leakage, misuse, dependency)
- Estimate likelihood and severity
- Define mitigations (technical, process, legal)
- Assign owner and review date
- Residual risk acceptance sign-off
See references/risk_assessment.md for worksheets and NIST AI RMF functions.
3. Documentation artifacts
| Artifact |
When |
| Model card / system card |
Every production model or vendor model |
| Data sheet |
Training/fine-tune data described |
| Eval summary |
Pre-deploy and periodic |
| Incident log |
AI-specific harms and near-misses |
See references/documentation.md for model card sections and change log.
4. Policy and oversight
- Acceptable use policy (prohibited uses, approval paths)
- Human oversight rules for consequential decisions
- Escalation for policy violations and serious incidents
- Training for builders and reviewers
See references/policy_oversight.md for governance committee cadence.
5. Vendor and third-party models
Review: data processing terms, subprocessors, retention, fine-tuning on customer data, safety commitments, breach notification, exit plan.
See references/vendor_review.md for vendor questionnaire topics.
When to load references
- Tiering and regulation mapping →
references/risk_classification.md
- Assessments and frameworks →
references/risk_assessment.md
- Model cards →
references/documentation.md
- Policies and committees →
references/policy_oversight.md
- Vendors →
references/vendor_review.md
1---2name: ai-risk-governance3description: Guides AI risk management and governance—use-case risk assessment, model/system documentation, policies and guardrails, human oversight, third-party model/vendor review, and mapping to frameworks (NIST AI RMF, ISO 42001, EU AI Act concepts, internal AI policy). Use when classifying AI use cases, drafting AI acceptable-use policies, building risk registers, preparing model cards or DPIAs for AI, reviewing vendor LLMs, or aligning product teams with compliance—not for implementing RAG/agents (ai-engineer), running jailbreak tests (ai-redteam), or SOC 2/ISO evidence automation and technical control mapping (compliance-engineer), or general SOC 2 IT controls without AI scope (cybersecurity). For AI solution architecture in commercial or enterprise deployments, use applied-ai-architect-commercial-enterprise. For agent skills catalog standards and security review before publish, use ai-skill-manager. Safeguard infra/research: ml-infrastructure-engineer-safeguards, ml-research-engineer-safeguards.4---56# AI Risk & Governance78## When to Use910- Classifying AI use cases by risk tier and impact11- Drafting AI acceptable-use policies and governance frameworks12- Building AI risk registers with likelihood/severity/mitigation tracking13- Preparing model cards, system cards, or DPIAs for AI deployments14- Reviewing third-party LLM vendors (data terms, fine-tuning, safety commitments)15- Mapping AI products to frameworks (NIST AI RMF, ISO 42001, EU AI Act concepts)16- Aligning product and engineering teams with compliance requirements17- Designing human-in-the-loop oversight for consequential AI decisions1819## When NOT to Use2021- Implementing RAG pipelines, agents, or production features → `ai-engineer`22- Running jailbreak tests or adversarial campaigns → `ai-redteam`23- SOC 2/ISO evidence automation and technical control mapping → `compliance-engineer`24- General SOC 2 IT controls without AI scope → `cybersecurity`25- Commercial/enterprise AI solution architecture → `applied-ai-architect-commercial-enterprise`26- Skills portfolio governance and publish gates → `ai-skill-manager`2728## Related skills2930| Need | Skill |31|---|---|32| Building LLM products | `ai-engineer` |33| Adversarial testing | `ai-redteam` |34| Research and benchmarks | `ai-researcher` |35| Enterprise security program | `cybersecurity` |36| Pipeline and data security | `devsecops` |37| SOC 2/ISO evidence and technical controls | `compliance-engineer` |38| AI solution architecture (commercial/enterprise) | `applied-ai-architect-commercial-enterprise` |39| Agent skills governance | `ai-skill-manager` |40| Safeguard gateways, classifiers, rollout | `ml-infrastructure-engineer-safeguards` |41| Safety classifier research and benchmarks | `ml-research-engineer-safeguards` |42| Privacy research for safeguards | `privacy-research-engineer-safeguards` |43| Enterprise security risk registers (non-AI scope) | `security-risk-analyst` |44| M&A/investment cyber diligence and board packs | `cyber-diligence-governance` |4546## Core Workflows4748### 1. Use-case intake and classification4950**Capture:**5152| Field | Purpose |53|---|---|54| Purpose and users | Scope and accountability |55| Data types | PII, special categories, IP |56| Automation level | Human-in-loop vs autonomous |57| Impact if wrong | Safety, legal, financial, reputational |58| External exposure | Customer-facing vs internal |5960**Risk tier (example):**6162| Tier | Criteria | Controls |63|---|---|---|64| Low | Internal, low impact, no sensitive data | Standard policy + logging |65| Medium | Customer-facing or internal PII | Review + eval + monitoring |66| High | Regulated domain, high impact, autonomous actions | Governance board + red-team + enhanced oversight |6768**See `references/risk_classification.md` for EU AI Act–oriented mapping (non-legal).**6970### 2. Risk assessment7172Use structured worksheet:73741. Identify hazards (bias, hallucination, leakage, misuse, dependency)752. Estimate likelihood and severity763. Define mitigations (technical, process, legal)774. Assign owner and review date785. Residual risk acceptance sign-off7980**See `references/risk_assessment.md` for worksheets and NIST AI RMF functions.**8182### 3. Documentation artifacts8384| Artifact | When |85|---|---|86| Model card / system card | Every production model or vendor model |87| Data sheet | Training/fine-tune data described |88| Eval summary | Pre-deploy and periodic |89| Incident log | AI-specific harms and near-misses |9091**See `references/documentation.md` for model card sections and change log.**9293### 4. Policy and oversight9495- Acceptable use policy (prohibited uses, approval paths)96- Human oversight rules for consequential decisions97- Escalation for policy violations and serious incidents98- Training for builders and reviewers99100**See `references/policy_oversight.md` for governance committee cadence.**101102### 5. Vendor and third-party models103104Review: data processing terms, subprocessors, retention, fine-tuning on customer data, safety commitments, breach notification, exit plan.105106**See `references/vendor_review.md` for vendor questionnaire topics.**107108## When to load references109110- **Tiering and regulation mapping** → `references/risk_classification.md`111- **Assessments and frameworks** → `references/risk_assessment.md`112- **Model cards** → `references/documentation.md`113- **Policies and committees** → `references/policy_oversight.md`114- **Vendors** → `references/vendor_review.md`
daemon-blockint-tech