How we audit AI systems: our approach

Blog-author
Sandeep Gokhale

Insights Time 5 mins

Insights DateNovember 17th, 2026

More and more organisations rely on AI for customer service, decisions and daily work. That raises a fair question: how do you know it is safe, fair and under control? An AI security audit answers it. This page explains how we run one, step by step, so you know what to expect before we begin.

What an AI security audit is

An AI security audit is an independent check of the AI systems your organisation uses. We look at whether they are protected from misuse, whether they treat people fairly, whether someone is accountable for them, and whether people stay in control. We use the CERT-In AI Security Auditing Guidelines as our framework, so our work follows a recognised national approach.

The principles we follow
  • Independent: We never audit anything we designed, built or run, and our fee does not depend on the result.
  • Based on risk: The more a system affects people or the business, the deeper we look.
  • Evidence, not promises: We check how things work in practice, not only what a policy says.
  • Fair and clear: You see our findings before the report is final, you can respond, and we explain every point in plain language.
The journey at a glance

Eleven stages take you from the first conversation to a wrapped-up audit. The highlighted stage is the signed engagement letter, which comes before anything sensitive is shared.

The journey, stage by stage

Getting started
  1. Enquiry: A first conversation of about 45 minutes. We ask what AI you use and what you hope to learn. You do not need to send any documents.
  2. Proposal: A short proposal that sets out what we would do, who would do it, how long it would take and what it would cost.
  3. Engagement: We agree and sign an engagement letter. It sets the scope, what we may test and how we handle your information. Nothing sensitive is shared before it is signed.
  4. Discovery: You tell us about your AI systems through a structured set of questions and share the documents we ask for. One person on your side coordinates.
  5. Scoping: We list your AI systems and rate each one by risk. This decides how deep the audit goes for each system.
  6. Planning: We plan the evidence we need, the interviews, and any testing. Testing happens only with your written permission.
  7. Fieldwork: We review the evidence, speak with the people who run and oversee your AI, and carry out the agreed checks. If we find something serious, we tell you straight away, not at the end.
  8. Analysis: We work out what the findings mean and why they happened. Fixing the cause matters more than fixing the symptom.
  9. Reporting: We share a draft so you can correct facts and respond. Then we issue the final report with our overall opinion.
  10. Follow-up: You plan the fixes and tell us how they are going. We check that they work.
  11. Close-out: We return or securely delete what we hold on the agreed date, and close the engagement.

What we look at

Depending on your systems, we look at nine areas:

  • Governance: Who is accountable, and how decisions are made.
  • Data: Where information comes from, who can see it and how long it is kept.
  • Models: How your AI is built, changed and retired.
  • Fairness: Whether outcomes are even-handed.
  • Explainability: Whether decisions can be explained to the people they affect.
  • Security: Whether the system resists attack and misuse.
  • Monitoring: Whether it keeps working as intended over time.
  • Human oversight: Whether people can review, challenge and override the AI.
  • Suppliers: The companies and AI services you depend on.

If you use chat assistants, document search or AI agents, we add checks designed for those.

How we rate what we find

Every finding gets one of four ratings. The rating tells you how quickly to act.

Rating What it means What we ask
Critical An immediate or severe risk Act straight away
High A significant weakness Tracked until it is fixed
Moderate A control exists but is incomplete or inconsistent A plan within 30 business days and a fix within 60 days
Low A minor issue or an improvement Within three months

What you receive

  • A short summary your leadership can act on.
  • Findings, each rated and with a practical recommendation.
  • Our overall opinion, which is one of four levels: Satisfactory, Satisfactory with Reservations, Requires Significant Improvement, or Unsatisfactory. It is a reasoned judgement, not a score.
  • A check that your fixes work.

If the opinion is Unsatisfactory, or there are two or more Critical findings, we recommend a follow-up audit within three months.

How we protect your information

  • We sign confidentiality terms and the engagement letter first.
  • Wherever possible we ask for redacted or sample data, not personal records.
  • We test only what you have approved in writing, and only in the way we agreed.
  • We do not use AI tools on your information without your written permission.
  • We store evidence securely and delete it on the date we agree.

What an audit is not

An audit is not a certification and it is not a guarantee. It is an independent view of where you stand on the day we finish, with a clear plan for what to do next. It does not certify that you comply with any law.

How to get ready

A few things make an audit smoother, and none of them is needed for the first conversation:

  • A list of the AI systems you use, including AI built into software you buy and tools your staff use on their own.
  • One person who can coordinate on your side.
  • The name of the person who owns each system.
  • Any policies you have on how AI may be used.

Do not worry if some of this does not exist yet. Finding the gaps is part of the point.

Talk to us

Tell us what AI you use and we will explain what an audit would involve. Contact Us or email us at hello@techvito.in

Let's Connect!

Have a question or looking to work with us? Fill out the form below, and we'll get back to you soon.


We are capturing your information please wait...

Your Response was submitted. our team will contact you shortly.
success iconThank you for being fan our content.You will start receiving amazing contents soon.