Appendix A: Tool Profile Template

Use this when adding an AI tool to the project's tool register. See Step 2: Check the Tool for what each fact means and why it matters.

This records what the tool is and what its supplier promises. It does not say what the tool may be used for: that is settled use by use, in each risk assessment.


Tool	[Name and provider]
Tier or plan assessed	[Enterprise and consumer versions are different tools for this purpose]
Facts gathered by	[Name]
Date	[Date]
Next review	[Date]

Is it excluded?

Any "yes" rules the tool out, whatever the intended use.

	Yes / No	Notes
Trains on your inputs, or will not commit not to
No DPA, or one that does not meet UK GDPR requirements
A free or consumer tier under consumer terms
Fails a client or department requirement

Excluded: [Yes / No]. If yes, stop here unless a documented exemption is approved by the SRO.

What the tool is

Design facts. These shape the inherent risk in Step 3.

Fact	Details
Where data is processed and stored	[Countries and jurisdictions]
Permissions the tool holds	[What it can reach: calendar, mailbox, files, repositories, ticketing systems. What it can access, not what you intend to give it]
What it can do	[Produces output only, or can take actions: run commands, edit files, call APIs, change records. Record the most it is capable of]
Control over what it can do	[For tools that can act: who defines the available actions, and whether the limit can be enforced outside the tool's own configuration]
Authentication and access management	[SSO, MFA, team/project access control, attribution of individual use]
Audit logging	[What is logged, and who can read it]

What the supplier promises

Contractual and configurable guarantees. These are mitigations in Step 3.

Fact	Details
Training on inputs	[Contractual commitment, or a changeable setting?]
Retention	[How long inputs are kept; deletion after processing]
Data processing agreement	[Available? Meets UK GDPR?]
Security certifications	[ISO 27001, SOC 2, Code of Practice for the Cyber Security of AI]
IP terms	[Rights claimed over outputs; indemnification]
Change and notice	[Can the model or behaviour change without notice? How much notice, and is it contractual?]
Handling of AI-specific risk	[How the provider addresses hallucination, bias, prompt injection]

Decision

A judgement, not a fact. Needs authority.

Highest classification cleared: [e.g. OFFICIAL]. Do not infer this from the facts above; UK residency and a DPA do not by themselves clear a tool for OFFICIAL-SENSITIVE.

Role	Name	Decision	Date
Assessed by
Approved by