Skip to content

How to answer a customer's AI security questionnaire

By Shiva Dhali · Updated

If you sell software to larger companies, the security questionnaire probably has a new section: how do you use AI? For many teams the honest answer starts with a scramble through Slack and spreadsheets to work out which tools are in use and who owns them. Preparing once, and keeping it current, turns the next questionnaire into a lookup.

What customers usually ask

  • Which AI systems or providers do you use, and for what?
  • Does any of our data go into those systems? Is it used to train models?
  • Where is that data processed and how long is it kept?
  • Who in your company is accountable for each AI system?
  • Is there human review of AI outputs that affect us?
  • Do you follow a framework such as the NIST AI RMF or ISO/IEC 42001?

The exact wording varies, but nearly every question comes back to the same facts: what you use, what data touches it, and who is responsible.

Step 1: Build an AI inventory

List every AI system in use, both in your product and internally: model APIs, AI features inside other SaaS tools, coding assistants and internal automations. For each one, record the purpose, the owner, the vendor, what data it receives, and whether it is in testing, live or retired.

Step 2: Decide what matters most

Not every tool needs the same scrutiny. A model that processes customer data and makes decisions on its own deserves more attention than an internal writing helper. Score each system on a few plain questions about data, autonomy and impact, and review the high-priority ones first.

Step 3: Write reusable answers

Draft standard answers from the inventory: a short AI usage statement, a list of AI subprocessors, and your data-handling position for each. Keep them with your other security documents so sales and security teams use the same wording.

Step 4: Have evidence ready

Reviewers trust a dated export more than a paragraph. A PDF or CSV showing each system, its owner, its review status and any open actions shows the register is real and maintained.

Common questions

Do small SaaS companies really get AI questions in security reviews?
Increasingly, yes, especially when selling to US and EU companies with formal vendor-risk processes.
Is an AI register the same as legal compliance with AI regulation?
No. A register is the foundation for answering questions and managing risk, but it isn't a legal assessment. Get legal advice for regulatory obligations.
What should an AI inventory include?
At minimum: the system, its purpose, owner, vendor, the data it receives, and its lifecycle status.

Try Aegistra

A lightweight register of every AI system your company uses, with owners, priority scores and the evidence customers ask for.