RBI FREE-AI

RBI FREE-AI implementation.

RBI’s FREE-AI report is a set of recommendations, not a binding rule. We turn it into an AI register, working controls and evidence for banks, NBFCs and the fintechs that serve them.

Recommendations, not rulesControls built
RBI FREE-AI

What we put in place

Where AI is usedCredit, collections, support, fraud and KYC.
Who owns each systemA named person for every AI use.
Vendors and modelsThird-party models and where data goes.
Testing and monitoringChecks before launch and while live.
Customer impactExplanations, complaints and human review.
EvidenceRecords a reviewer can open and read.
Six areasMapped to your AI uses
Evidence collected

Where FREE-AI stands, said plainly.

Read the report itself before you rely on any summary, ours included. For RBI’s separate draft text on models, including AI, see RBI’s draft model risk guidance for GenAI.

What FREE-AI is

The report of an RBI committee on a Framework for Responsible and Ethical Enablement of Artificial Intelligence in the financial sector. RBI released it on 13 August 2025. It sets out 7 Sutras (principles) and 26 recommendations under six pillars.

What it is not

It is not a direction, circular or master direction. The recommendations are not binding on their own. We will not tell you that RBI requires something because the report suggests it.

What else applies

As of November 2026, RBI’s AI-specific regulatory text is the draft model risk management guidance of 24 June 2026. It is a draft, not final, so it does not bind anyone yet. RBI directions that already apply to your entity do not stop applying because a system uses AI. Check with your counsel which ones cover your AI uses.

What we build, and the evidence it leaves.

We map the recommendations that apply to your AI uses to these controls. Where personal data is involved, the same evidence supports your DPDP work for BFSI.

AI register

What a reviewer checksEvery model and AI feature in use: credit scoring, collections, chatbots, fraud, KYC and internal tools, including those built by vendors.

Evidence we collectA register with an owner, purpose, data used, vendor and risk level for each entry.

Policy and ownership

What a reviewer checksWho approves a new AI use, who can change it, and who answers for it to your board and to RBI.

Evidence we collectAn AI policy that matches your real systems, named owners, and a record of approvals.

Vendor and model-provider review

What a reviewer checksWhich third-party models and AI vendors touch customer data, in which region, on what terms, and what you do if one fails.

Evidence we collectVendor files with contract terms, region and retention settings, and an exit or fallback plan.

Testing and monitoring

What a reviewer checksHow a model is checked for accuracy, bias and drift before and after it goes live, and what happens when it degrades.

Evidence we collectTest results, monitoring dashboards, and logs of changes and reviews.

Customer-facing AI

What a reviewer checksWhether customers can tell they are dealing with AI, get an explanation of a decision, and reach a person.

Evidence we collectProduct screens and scripts, escalation paths, and a log of complaints that involve AI.

Data protection and security

What a reviewer checksHow personal data in prompts, training sets and outputs is protected, kept and erased.

Evidence we collectAccess controls, retention settings and logs, aligned with your DPDP work.

AI incidents

What a reviewer checksWhat counts as an AI incident, who is told, and how you fix and record it.

Evidence we collectAn incident process that names AI cases, with a worked example.

For fintechs that serve banks.

If you sell credit models, collections tools or AI support to a bank or NBFC, your client’s AI review becomes yours. The AI section of their questionnaire is where SIG and AI-CAIQ questions arrive. Our questionnaire help answers them with evidence.

Credit and risk models

Which inputs drive a score, whether you can explain a decline, and how you test for unfair outcomes.

Bank due diligence

Banks and NBFCs may send vendors AI questions. Accurate answers need evidence from your systems, not memory.

Contract flow-down

A regulated client may ask you to meet AI controls by contract, even where you are not regulated by RBI yourself.

Third-party models

If you call an external model API, the client will ask where its data goes and what happens if the provider changes terms.

How it runs.

We prepare and build. Your counsel, auditors and RBI decide what you must meet and whether you meet it.

  1. Step 1

    Scope

    We agree which AI uses, products and entities are in scope. Your counsel confirms which RBI rules already apply to them.

  2. Step 2

    Map

    We read the FREE-AI report with you and map each recommendation that applies to your use cases.

  3. Step 3

    Build

    We set up the register, policy, vendor files, testing and logging. Our engineers fix gaps in the systems.

  4. Step 4

    Evidence

    We collect the records into a pack you can show your board, auditors, supervisors or bank clients.

RBI FREE-AI FAQs.

Is RBI FREE-AI mandatory?

No. It is a committee report with recommendations, not a binding direction. RBI could turn parts of it into rules later, and its draft model risk management guidance of June 2026 is separate. Check the report and any RBI notice that applies to you, and take legal advice on what binds you.

Does FREE-AI apply to a fintech that RBI does not regulate?

Not directly. The report is aimed at the financial sector. But banks and NBFCs are your clients, and they may ask you to meet AI controls through contracts and due diligence questionnaires. Working to the report helps you answer them.

How is FREE-AI different from the draft model risk guidance?

FREE-AI is a broad set of recommendations on AI in finance. The draft model risk management guidance is a separate RBI text on how regulated entities should manage models, including AI and machine learning models. As of November 2026 it is still a draft. We cover it in our note on RBI model risk management for GenAI.

Can you give us a FREE-AI certificate?

No. We are not a certification body, auditor or law firm, and we know of no RBI certificate for FREE-AI. We prepare your systems and evidence, so that your own auditors, your supervisor or your bank clients can review them.

Where do we start if we have no AI inventory?

With a list of every place AI is used, including features inside vendor products and tools staff use on their own. That register shows which uses touch customers or money, and those come first. A shadow AI audit is a quick way to build it.

Is this legal advice?

No. We work on the engineering and the evidence: systems, settings, logs and documents. Contract terms and legal questions go to your counsel, or to our partner law firm if you want us to involve them.

Turn the FREE-AI report into controls you can show.