RBI Model Risk Management
RBI model-risk readiness for every lending decision system
If you came for the framework rather than the product, start with the explainer and self-benchmark: RBI model risk management framework for NBFCs.
An evidence-ready control plane for the models, BRE rules, scorecards, AI use cases, third-party dependencies, and co-lending workflows behind your lending decisions, so when the board or the regulator asks, the answer and the evidence are already in one place.
Built by the team behind Apache Fineract · Append-only evidence ledger · Four-eyes governance · Human-authoritative by design
The expectations just widened. The evidence work didn't get easier.
RBI's June 2026 draft Guidance on Model Risk Management, with the FREE-AI framework, widens what lenders must govern: a board-approved framework, a complete model inventory, risk tiering, independent validation, AI/ML controls, and accountability for third-party and co-lending models. The hard part was never the policy; it is proving, on demand, what exists, who owns it, what's validated, and what's ready for the board.
Most lenders can answer only by stitching spreadsheets together the week before a Risk Committee. Lokta replaces the stitching with a system of record.
For a clause-by-clause read of what the June 2026 draft actually asks for, see our breakdown of the RBI model risk management guidance. The same reading applied to a second draft, on the proposed revolving-credit restriction for NBFCs, covers what a term-loans-only rule would change about how a book is designed, booked and serviced.
One control plane, from scattered evidence to a board-ready pack
- 01
Inventory every decisioning asset
Models, rules, scorecards, AI use cases, vendor APIs, and co-lending, one catalogue.
- 02
Score your readiness
Against the draft, dimension by dimension, with the gaps ranked.
- 03
Track and close evidence gaps
A live queue of what's missing, who owns it, and what “ready” means.
- 04
Govern the record
Approvals, exceptions, and changes with four-eyes sign-off and an immutable trail.
- 05
Generate the board pack
A Risk-Committee-ready evidence pack in minutes.
See where your model-risk evidence actually stands
From one inventory to a board-ready pack: built so the evidence is examination-ready, and a human stays the authority on every assertion.
One inventory of every decisioning asset
Models, BRE rules, scorecards, pricing calculators, AI use cases, bureau and vendor APIs, and co-lending dependencies: each with an owner, a risk tier, an evidence score, and a status. The whole decisioning surface in one catalogue, not a dozen spreadsheets.
Readiness scored against the draft, worst gaps first
A live board-readiness score across the twelve RBI model-risk themes, each one ranked, with the agent summarising which gaps to close first to lift the number before the Risk Committee meets.
The agent proposes. A human decides
Ask about your estate and get an answer with citations to the exact record behind every claim. The agent drafts and surfaces; it never approves evidence or activates a control. Nothing moves without a person.
A board-ready pack, assembled from current evidence
Eleven sections (executive summary, readiness heatmap, high-risk assets, validation status, exceptions, third-party exposure), all drafted from what is on record, with citations throughout. The agent never finalises; you review and snapshot it.
Every claim traces back to a record
Open Sources on any pack and each line links to the evidence behind it: the validation, the approval, the gap finding, the snapshot. Read-only references, so the pack holds up to the question no one prepared for.
Four eyes on every action, and two separate steps
The agent's toolbox is read and create-draft only. Every action is logged as Proposed, then Accepted or Rejected by a named reviewer under two-person sign-off. Accepting a draft and activating a control stay gated separately, each requiring its own sign-off.
Built for how RBI defines "model": the whole decisioning surface
The draft's definition is deliberately broad: AI/ML, algorithms, decision rules, calculators, even the spreadsheets that materially shape a lending decision, plus the vendor and co-lending models you are now accountable for.
Most governance tools were built to watch AI models. Lokta was built for lending decisioning: deterministic BRE rules, scorecards, pricing logic, and third-party dependencies sit in the same catalogue as your AI use cases, governed the same way, because that is the surface a supervisor will examine.
Line-mapped to the June 2026 MRM draft
Every readiness dimension and gap in the self-benchmark traces back to a specific expectation in RBI's June 2026 draft. No generic global checklist retrofitted to India: purpose-built for the RBI examination your board is preparing for.
Governance you can trust, because it's governed itself
Append-only evidence ledger
Every action is preserved with who, what, when, and source. Records cannot be quietly altered or deleted.
Four-eyes on every assertion
Validation, approval, exception, tier confirmation, and decommission all route through two-person sign-off.
Human-authoritative by design
Assistive agents propose and surface. They never approve, alter evidence, or activate a control without a person. Every agent action is logged.
You stay the authority
Lokta is your evidence system of record. Your validators, auditors, and board remain the decision-makers: we organise and ready the evidence, and they render judgment on the models.
The people accountable when the regulator asks
- CROs
- Heads of Compliance
- Model-risk & credit-policy teams
- Data science leaders
- Internal audit
- CTOs
- Founders of fintech lenders
Designed for Middle Layer NBFCs, co-lending partners, and fintech lenders who carry the same accountability as a large bank, without a large bank's model-risk department.
Up and running on uploads, not a six-month integration
- 01
Upload
Your assets and existing records, with guided templates.
- 02
See your readiness
On a live dashboard, scored against the draft.
- 03
Close the gaps
From a prioritised queue, with owners and "ready" criteria.
- 04
Generate your board pack
Evidence-ready, in minutes.
Find out where your model-risk evidence stands today.
Take the RBI MRM draft self-benchmark. 10 questions, anchored to the June 2026 draft. Get your readiness score instantly. No email needed to see where you stand.
Take the self-benchmark →Evaluate your model-risk evidence workflow with Lokta
RBI Model Risk Management is available now. Share your institution type and role, and we will arrange a product evaluation and readiness walkthrough.
Prefer a general conversation? Talk to us
Questions a risk officer asks
The sharp ones: compliance, independence, third-party scope, integration, and the draft itself.
Does Lokta make us "RBI compliant"?
No tool can. Lokta helps you build and maintain the evidence and readiness a lender needs to demonstrate its model-risk governance: the inventory, the validations on record, the approvals, the gaps, and the board pack. Compliance is your board's determination; we make it provable.
You also build lending technology. Can you be our governance system?
Lokta RBI MRM is your own system of record: it organises and evidences the controls you run. Your validators, auditors, and board remain the independent authority over your models. We do not validate or certify models; we make the evidence examination-ready. Where required, your data stays walled from any other Lokta product.
We use third-party and co-lending models. Are those covered?
Yes. That is a core reason Lokta exists. Vendor, bureau, and co-lending dependencies sit in the same catalogue as your own assets, with ownership and oversight evidence tracked against the draft's expectation that accountability is not outsourced.
Do we need to integrate our systems first?
No. Start with guided uploads and get a readiness dashboard and board pack from day one. Deeper source-system evidence APIs are available when you are ready.
Is the June 2026 guidance final?
It is a draft, out for public consultation. Lokta is built around evidence and board-readiness that serve your Risk Committee today, and we track the guidance as it finalises so your mapping stays current.