Photo
INDUSTRIES · BFSI · BANKING, FINANCIAL SERVICES AND INSURANCE
Put AI to work on regulated data, with proof a regulator can check.
Models that run inside your perimeter, personal data masked before any model sees it, and a recorded trail behind every answer.
IN SHORT
For banking, financial services and insurance (BFSI), AgentAnywhere is designed to keep customer data inside the institution. Aavaran, our data-masking layer, masks personal and financial data before any model reads it, Kuber answers regulatory questions from public RBI, SEBI and IRDAI documents and shows the passage it relied on, Taksha modernises legacy code on your own network, and every answer is recorded, with a signed Trust Receipt for each masking call. Underwriting, claims and KYC assistance are work we build with you, with a person deciding.
01 · Customer data has rules about where it sits
IRDAI keeps policy and claims records in data centres in India, and the RBI's outsourcing directions set need-to-know and isolation provisions. A foreign API is a hard sell to your own risk committee.
02 · The regulator will ask
You need a record of what a model did with which data, in a form that can be checked without trusting the vendor.
03 · Decades of legacy code
Core systems need modernising without a silent guess changing what the business logic does.
WHY IT FITS
Sovereign, our models, and the Trust Platform.
Three things, in the terms of BFSI: where it runs and who controls it, which models do the work, and the controls and the record around every call.
01 · Sovereignty
It runs where your data already sits, and you control where that is.
- Inside your perimeter: your own cloud account in an Indian region, or your own hardware. Region-pinned, and designed to deploy air-gapped.
- Swaraj, the sovereign deployment, is designed to pin a deployment to the region you choose, which helps keep policy and claims records in the data centres your regulator expects.
- It runs where you run your other systems, under your own governance and your own change process.
02 · Our models
The line we select, measure and serve, built for regulated work.
- Kuber answers regulatory questions from public RBI, SEBI, IRDAI and statutory documents and shows the passage it relied on. Your own policies and circulars are added with you.
- Taksha, our coding line, modernises legacy code on your own network, with an independent referee that checks every claimed success.
- Manthan reasons over long documents and drafts for your people, inside the same perimeter. Kuber, Taksha and Manthan are built on open foundation models, measured and served by us.
- Underwriting, claims and KYC assistance are work we build with you, with a person deciding. You can also bring your own open model.
03 · The Trust Platform
The controls and the record, around every call.
- Aavaran, our data-masking layer, masks personal and financial data before a model reads it, with a signed Trust Receipt for each masking call.
- Kavach is designed to screen each request for prompt injection and unsafe output. The Agent Gateway applies your policy at the point of the call.
- Every step is recorded, in a trail you can query and export as CSV. Your compliance function owns the filing.
INSIDE YOUR PERIMETER
What happens to one customer request.
Follow one request from the moment it arrives to the receipt an auditor can check. Raw values are designed not to cross the boundary.
One request · start to receipt
Illustration
Your perimeter · your cloud, your region, or designed to deploy air-gapped
Step 1 of 5
A request arrives
A question comes in. It carries a name, an account, a card number: the things that must not travel.
THE RULES
The rules that bind you.
What each instrument requires in plain words, where it stands today, and what it means for AI in BFSI. Dates that have not arrived are written as dates.
RBI (Commercial Banks – Managing Risks in Outsourcing) Directions, 2025
Issued 28 Nov 2025
Source: RBI, 28 November 2025
What it requires
Issued 28 November 2025 in place of the 2023 IT outsourcing directions. They include need-to-know (paragraph 24) and isolation (paragraph 27) provisions for outsourced work.
What it means for AI
A model service is an outside party that sees your data. A model that runs inside your perimeter, with personal data masked before it reads anything, leaves your risk team less to isolate and a record of what was shared.
What it requires
A framework report, not a direction. Recommendation 23 asks for an inventory of the AI an institution uses, and recommendation 24 for an AI audit framework.
What it means for AI
The agent Registry and Model Hub are designed to hold the inventory, and the recorded trail is the material an AI audit works from. The Trust Center describes our mapping of the platform to all twenty-six recommendations, available under NDA.
What it requires
Circular IRDAI/GA&HR/CIR/MISC/51/4/2026 of 6 April 2026, revising the 2023 guidelines. Incidents go to CERT-In within 6 hours of noticing, with a copy to IRDAI. Cloud only from MeitY-empanelled providers whose data centre holds valid STQC audit status. ICT logs are kept 180 days within Indian jurisdiction.
What it means for AI
If AI runs on a public cloud, the provider and its data centre have to qualify. AgentAnywhere is designed to run in your own cloud account or on your hardware, and its record stays in the deployment you control.
IRDAI (Maintenance of Information by the Regulated Entities and Sharing of Information by the Authority) Regulations 2025, regulation 9(iii)
In force
What it requires
Records of policies and claims in India are held in data centres located and maintained in India only.
What it means for AI
A policy or claims assistant should run where those records may sit. Kuber and the platform are designed to run in India, region-pinned.
What it requires
Mutual funds using AI or ML report to SEBI through AMFI each quarter. We found no other issued SEBI AI or ML guideline when we read the rules on 30 October 2026.
What it means for AI
If you are a mutual fund, the recorded trail is the material for that report. The report and the filing are yours.
What it requires
Commenced by G.S.R. 843(E) of 13 November 2025; most duties start about May 2027. Section 8(5): protect personal data, including data with a processor, by reasonable security safeguards. Rule 7(2)(b): a fuller breach report within 72 hours. Rule 6(1)(e): logs and personal data kept one year.
What it means for AI
A hosted model service can be a data processor. Masking before the model reads, and a trail you can query when something goes wrong, help you show the safeguards and rebuild an incident. Government may also require Significant Data Fiduciaries to keep specified data in India; as far as we read, no list or categories have been notified.
Also read: the RBI cybersecurity and technology risk directions of 31 July 2026, which replaced the 2023 IT governance directions and the 2016 cyber framework, and SEBI's CSCRF (circular 2024/113 of 20 August 2024, in force).
This is what the rules say, not legal advice. Your compliance function owns the reading and the filing.
USE CASES · BY FAMILY
What runs inside a bank or an insurer.
Each card opens the page for that family or machine. Tell us your setting and we will show you the closest demonstration.
WORK SCENES
What it looks like on a working day.
The problem, what runs, and what the person at the desk gets. Each scene is an illustration of how the work is designed to go, not a customer account.
Scene 01
A branch officer asks about a circular
The problem
- The answer a customer needs sits in a circular nobody has time to read.
What runs
- Kuber answers from public RBI, SEBI, IRDAI and statutory documents and shows the passage it relied on.
What they get
- The officer reads an answer with its passage beside it, and the question is recorded.
Scene 02
A disputed card transaction
The problem
- A dispute arrives carrying a customer's name and card number, and neither may be pasted into a model.
What runs
- Aavaran masks both before the model reads the message, and the Agent Gateway applies your policy to each call.
What they get
- The reviewer works from masked text, with a signed Trust Receipt for each masking call and a person approving the outcome.
Scene 03
A credit file to prepare
The problem
- A small-business loan file is days of reading for a two-page opinion.
What runs
- Agents prepare the file on masked documents, and Manthan served inside your perimeter drafts the memo.
What they get
- The credit officer reads a draft with its sources and decides the loan.
Scene 04
Legacy code on a core system
The problem
- Decades of business logic sit in code where a silent guess can change what the bank does.
What runs
- Taksha's modernisation specialist works on your own network, and an independent referee checks each claimed success.
What they get
- Your reviewer receives each change with the referee's verdict beside it and decides what ships.
Scene 05
A voice note in the customer's language
The problem
- A customer sends a voice note about a failed payment, in Hindi or Tamil, and the recording is sensitive.
What runs
- AgentAnywhere Vaani is designed to understand, mask, classify and route it on Indian infrastructure, and to keep only an anonymised tally.
What they get
- The right desk receives a masked, categorised request, and uncertain or urgent notes go to a person.
THE SOVEREIGNTY TEST
The sovereignty test, for BFSI.
If an AI system cannot answer these from its own records, you are renting it. Beside each question: what a rented AI service usually answers, and how we answer. Read the test.
01Where does the data run?
What a rented AI service usually answers
“In a region near you, per the contract.” The vendor decides when the model changes.
How we answer
Inside your perimeter: your cloud in an Indian region or your own hardware, region-pinned, and designed to deploy air-gapped. Designed so that policy and claims records sit where the IRDAI regulations expect them.
02Which model made that decision?
What a rented AI service usually answers
“The AI.” The model version lives in a vendor's release notes.
How we answer
Model Hub is designed to register each model with an owner, lineage and evaluation history. Kuber, Taksha and Manthan are built on open foundation models, measured and served by us.
03What did the model actually see?
What a rented AI service usually answers
The raw customer record: name, account and card number.
How we answer
Aavaran masks personal and financial data before the model reads it, with a signed Trust Receipt for each masking call.
04What stops the agent doing harm?
What a rented AI service usually answers
A system prompt that asks the model to behave.
How we answer
A policy outside the model. Kavach is designed to screen each request for prompt injection and unsafe output, and the Agent Gateway applies your policy at the point of the call.
05Who approved the consequential step?
What a rented AI service usually answers
“The system.”
How we answer
A step your policy holds waits for a named person, and the approval or refusal is recorded. On a credit file the agents prepare it and an officer decides.
06What will you show the regulator?
What a rented AI service usually answers
A screenshot and a slide.
How we answer
A recorded trail, queryable and exportable as CSV, with a signed Trust Receipt for each masking call. Your compliance function owns the filing.
STATED PLAINLY
What this is not.
- Kuber answers regulatory questions from public documents and shows its source. Underwriting, claims and KYC assistance are work we build with you, with a person deciding.
- We describe Kuber and AgentAnywhere Vaani without accuracy figures. AgentAnywhere Vaani's first pilot deployment is being built, and its speech recognition and synthesis are open-licensed external components used for now.
- A model's answer supports a decision. Your own officers make it.
- The full FREE-AI mapping is shared under NDA. Your compliance function owns the filing.
- Masking reduces what a model reads. It does not take data outside any law.
THE OFFER · THE SOVEREIGN PILOT
Bring your hardest workload. We come to your data.
Pick the one workload your risk committee keeps deferring because the data cannot go anywhere. We run the pilot where the data already sits, with our architects beside your team, and you decide at the end whether it goes live.
For banks and insurers: bring the workload your risk committee keeps deferring, such as a circular desk, a dispute queue or a credit file. We set up inside your cloud account or on your hardware.
What the pilot is
- One workload, one team, six to ten weeks.
- A sovereign environment by default.
- Solutions Architecture paired with your team for the duration.
- It ends with one agent in production under your governance, or a documented decision not to proceed.
What we add
- The Trust Platform is on from the first call: data masking, policy screening, the Agent Gateway and a recorded trail.
- The record stays with you and is exportable as CSV.
- You can bring your own open model.
- A reply within one business day.
A risk committee has no reason to refuse, because the data does not have to go anywhere to find out. The pilot ends one of two ways: one agent in production under your governance, or a documented decision not to proceed, with the record behind it.
FAQ
Frequently asked questions.
- Can a bank in India use AI without customer data leaving the country?
- Yes. AgentAnywhere deploys inside your perimeter, in your cloud account in an Indian region or on your own hardware, region-pinned and designed to deploy air-gapped. Aavaran masks personal and financial data before any model reads it, and is designed so that raw data does not leave your boundary.
- How does a bank prove to a regulator what an AI model did?
- Every consequential call is recorded: what was asked, which policy applied, what was masked and what was returned, and each masking call carries a signed Trust Receipt. The trail is queryable and exportable as CSV. Your compliance function owns the filing.
- What does Kuber do today?
- Kuber answers regulatory questions from public RBI, SEBI, IRDAI and statutory documents and shows the passage it relied on, inside your perimeter. Your own policies and circulars are added with you. Underwriting, claims and KYC assistance are work we build with you, with a person deciding. Kuber is built on open foundation models, measured and served by us.
- What does the RBI's FREE-AI report ask banks to do?
- FREE-AI is a framework report, not a direction. Recommendation 23 asks for an inventory of the AI an institution uses, and recommendation 24 for an AI audit framework. The agent Registry, Model Hub and the recorded trail are designed to give both something to work from. The Trust Center describes a mapping of the platform to all twenty-six recommendations, available under NDA.
- Where must an insurer keep policy and claims records?
- In India. Regulation 9(iii) of the IRDAI (Maintenance of Information by the Regulated Entities and Sharing of Information by the Authority) Regulations 2025, in force, says records of policies and claims in India are held in data centres located and maintained in India only. A model that reads those records should run there too.
- Which cloud may an insurer use for AI?
- IRDAI's Information and Cyber Security Guidelines 2026, version 2.0, issued on 6 April 2026, allow cloud only from MeitY-empanelled providers whose data centre holds valid STQC audit status. AgentAnywhere is designed to run in your own cloud account or on your hardware, so the choice of provider stays yours.
- When do the DPDP rules start to apply to banks and insurers?
- Most duties start from May 2027, after the commencement notified on 13 November 2025. Section 8(5) asks for reasonable security safeguards, including for data with a processor. Rule 7(2)(b) asks for a fuller breach report within 72 hours, and rule 6(1)(e) for logs and personal data to be kept one year. Your compliance function owns how they apply to you.
- What does the Sovereign Pilot involve for a bank or insurer?
- One workload, one team, six to ten weeks, in a sovereign environment, with Solutions Architecture paired with your team. The Trust Platform is on from the first call, the record stays with you and is exportable as CSV, and it ends with one agent in production under your governance or a documented decision not to proceed.
Bring your hardest workload.
We come to your data. One workload, one team, six to ten weeks, in a sovereign environment. Tell us the workload and a solutions architect will reach out within one business day.
