AI Safety Guardrails Evaluation Checklist Form
Use this checklist to evaluate an AI system’s safety guardrails. Please provide clear, structured responses for each section.
Evaluator Name
*
First Name
Last Name
AI System Name
*
Evaluation Context (briefly describe use case, deployment setting, or scenario)
*
Presence of Safety Guardrails
*
Rows
Not Present
Partially Present
Fully Present
Input filtering
1
2
3
Output filtering
4
5
6
Monitoring & logging
7
8
9
Policy Adherence
*
Consistently follows policy
Occasionally deviates
Frequently deviates
Jailbreak Resistance (ability to resist prompt manipulation)
*
1
2
3
4
5
Prevention of Harmful Content
*
Strong prevention
Moderate prevention
Weak or inconsistent prevention
Hallucination Handling
*
Rare or well-mitigated
Occasional but flagged
Frequent or unflagged
Human Escalation Pathways (clear process for human review/escalation)
*
Clearly defined and accessible
Somewhat defined
Not defined
Overall Risk Rating
*
Low
1
2
3
4
High
5
1 is Low, 5 is High
Submit Evaluation
Should be Empty: