# AI functional risk assessment

Version 2.0 · Updated 2026-09-21 · Planning

Connect a specific failure to its consequences, controls, tests and residual-risk decision.

Adapt this template to your intended use, applicable requirements, and approved procedures.

## When to use it

Use when you need to connect a specific failure to its consequences, controls, tests and residual-risk decision.

## Information needed before starting

- Defined AI use, authority limits and assessed configuration
- Controlled sources and independent reference results relevant to this worksheet
- Applicable risk assessment, acceptance basis and responsible reviewers

## Field instructions

- **Requirement or function:** Identify the specific intended behavior and stable requirement ID.
- **Failure and cause:** Describe a missed, wrong, unauthorized or unavailable behavior and credible contributing factors.
- **Consequence and severity:** Trace the process effect and apply the organization’s severity rubric.
- **Existing control:** Identify the control that interrupts the failure, including its owner.
- **Control evidence:** Link tests or other evidence showing the control works.
- **Remaining gap:** Record uncertainty, failed criteria and unresolved controls.
- **Residual-risk decision:** State accept, restrict, defer or other locally defined disposition with rationale.
- **Owner and trigger:** Identify who acts and which change or result triggers reassessment.

## Completed fictional example

Fictional risk rows from PRA-01. No universal numerical risk score is implied.

| Risk / requirement | Failure and consequence | Control and evidence | Residual decision |
| --- | --- | --- | --- |
| R-01 / REQ-01 | AI omits critical alarm challenge; protocol coverage could be inadequate | Full required review and T-01/T-03 evaluation; two AI critical misses and one final assisted critical miss | Expanded reliance withheld; DEV-01/02 open |
| R-02 / REQ-02 | Unsupported requirement misdirects review | Approved-source verification; T-02 identifies six false AI findings, T-03 leaves two accepted | DEV-03 open; improve and retest |
| R-04 / REQ-04 | Unauthorized action or source disclosure | Permission boundaries; T-04 blocks six of six planned attempts | Supported only for tested scope; retain technical restrictions |

## Related guidance

- [Read chapter 6: AI Risk Assessment](https://csvtocsa.com/ai-assurance-academy/ai-in-csv-csa-chapter-6-risk-analysis-for-ai-failure-and-uncertainty)

- [Read chapter 7: AI Validation Evidence Plan](https://csvtocsa.com/ai-assurance-academy/ai-in-csv-csa-chapter-7-data-governance-provenance-and-alcoa-for-ai)

- [Read chapter 20: AI Release Decision: Complete Case Study](https://csvtocsa.com/ai-assurance-academy/ai-in-csv-csa-chapter-20-end-to-end-case-study-and-inspection-readiness)

Canonical page: https://csvtocsa.com/worksheets/ai-functional-risk-assessment