What should I ask an AI vendor before using it with client data?

Where the data is stored and processed, whether it trains models, how long it is kept, who can see it, which other companies handle it, what security audit or certificate the vendor holds, and whether your plan includes a data processing agreement. Client data and AI sets out what can go into an approved tool.

What is a SOC 2 Type 2 report?

An independent audit of a company’s security controls over a period of time, often six to twelve months. Ask for the report and the period it covers.

Do small AI tools need to answer all of this?

Any tool that will see client data should. A free or small tool that cannot answer basic questions about storage, training and retention is a reason to keep client data out of it. For meeting tools, see what to ask any AI notetaker vendor.

Free template

AI vendor security questionnaire

Training is only one question. Before your firm approves an AI tool, find out who can access the data, where it is stored and processed, how long it is kept, which other companies handle it, and how you stay accountable for it. Send this to the vendor, file the answers, and record what you decided.

Updated 11 November 2026 · Editable Word document (.docx) · Free with a firm profile

Free with a firm profile

Get the editable AI vendor security questionnaire

Or build it with AI

Open Claude or ChatGPT with the prompt filled in. It writes the finished document for your firm straight away, then asks a few questions to tailor it. No client data needed.

Open in Claude Open in ChatGPT

Works without client data. Don't paste client names, tax file numbers or ledgers into an AI tool your firm hasn't approved for them. AI data safety playbook.

How to use it

How to use this template

Check the trust centre first

Many answers are on the vendor’s security and privacy pages. Fill in what you can, then send the rest. Not sure what the terms mean? AI vendor jargon, explained covers SOC 2, DPAs and subprocessors.
Send it before client data goes in

Ask the vendor to answer in writing and attach their documents: the SOC 2 Type 2 report or ISO 27001 certificate, the data processing agreement and the subprocessor list.
Ask about the plan you are buying

Protections often apply only to business or enterprise plans. Make sure every answer is for the plan your firm will use. AI privacy settings for ChatGPT, Claude, Copilot and Gemini shows what changes between personal and business plans.
Record the decision and a review date

Approved, approved with conditions, or not approved, who decided and when you will look again. Add it to your AI tools register.

The template

[Firm name] AI vendor security questionnaire

Replace everything in [square brackets].

1. About the tool

Question	Vendor answer
Product name and the plan we would use	[ ]
What we will use it for	[ ]
Which AI models it uses, and from which provider	[ ]
Vendor contact for security questions	[ ]

2. Your data

Question	Vendor answer
Where is our data stored?	[ ]
Where is it processed? (This can differ from where it is stored.)	[ ]
Is our data, or are our prompts and files, used to train any model on this plan?	[ ]
Can training be switched off for every user in the firm?	[ ]
How long are prompts, files, outputs and logs kept?	[ ]
Can we delete our data, and how quickly is it removed?	[ ]
Can your staff see our data? When, and who approves it?	[ ]
Which subprocessors handle our data, and where are they?	[ ]

3. Security

Question	Vendor answer
Do you hold a SOC 2 Type 2 report or ISO 27001 certificate? Please attach it and give the period or scope.	[ ]
Is data encrypted in transit and at rest?	[ ]
Can we require single sign-on and multi-factor authentication for everyone?	[ ]
Do we get audit logs of who did what, and when?	[ ]
How quickly will you tell us about a security incident affecting our data?	[ ]

4. AI-specific questions

Question	Vendor answer
What happens when the tool reads a document or email with hidden instructions (prompt injection)?	[ ]
If it connects to other systems, what can it read, and can it change or send anything?	[ ]
Can we limit connections to read only, or to certain clients, folders or users?	[ ]
Does a thumbs up or down, or other feedback, share our conversation with you?	[ ]

5. Contract

A data processing agreement is included on our plan, and we have a copy.
The vendor will tell us before adding a new subprocessor or changing where data is stored.
We can export and delete our data if we leave.

6. Our decision

Item	Record
Decision	[Approved / Approved with conditions / Not approved]
Conditions (for example: no client identifiers, read only connectors)	[ ]
What client data may go in, under our AI policy	[ ]
Decided by	[Name, role]
Date	[ ]
Review date	[ ]

This template is a starting point, not legal advice. Your obligations depend on where you practise, your professional body and your clients; take advice on anything you are unsure of.

Common questions

Frequently asked questions

What should I ask an AI vendor before using it with client data?: Where the data is stored and processed, whether it trains models, how long it is kept, who can see it, which other companies handle it, what security audit or certificate the vendor holds, and whether your plan includes a data processing agreement. Client data and AI sets out what can go into an approved tool.
What is a SOC 2 Type 2 report?: An independent audit of a company’s security controls over a period of time, often six to twelve months. Ask for the report and the period it covers.
Do small AI tools need to answer all of this?: Any tool that will see client data should. A free or small tool that cannot answer basic questions about storage, training and retention is a reason to keep client data out of it. For meeting tools, see what to ask any AI notetaker vendor.

More templates

The thinking behind these: AI for accountants and bookkeepers: what actually works.