How to use the checklist
PDF · 35 KB
Free resource · Version 1.0 · Reviewed 6 Nov 2026
Can your AI agents act in your systems? Review their boundaries before expanding use.
A practical checklist for data, permissions, approvals, logs and failure response. Assign owners and prioritize verifiable changes without treating it as a certification.
Get the checklistFor teams using connected agents
For small-business leadership, operations, product and IT teams with agents that can access data, use tools or take actions. If you only use assistants without integrations, start with the broader AI kit.
What you get
Agent security checklist
Editable XLSX · 9 KB
Review copy
Printable PDF · 39 KB
How to use it
- 1Choose one agent and record the data, tools and actions it can use.
- 2Mark each control Yes, Partial, No or Not applicable; record actual evidence.
- 3Prioritize three gaps, assign owners and verify changes before expanding access.
Download the checklist
We record your request and provide access here. We do not email the files.
Frequently asked questions
Is this an audit or certification?
No. It is an internal review tool. It does not technically verify an agent or certify ISO/IEC 42001.
What counts as evidence?
A verifiable setting, log or test with a date and owner. Do not put credentials or sensitive data in a shared copy.
What if we do not use agents?
Start with the Safe AI Adoption Kit. This checklist focuses on systems that can act or use tools.
Must I send the completed checklist?
No. Keep it within your company with limited access; we do not need sensitive integration or permission details.
This does not guarantee security, compliance or absence of incidents. Technical decisions require context and risk-based review.