RBI FREE-AI report
Recommends a board-approved AI policy.
BiltIQ AI for finance is Agent OS and AI Factory installed on a bank's, NBFC's or insurer's own servers. Agents read loan files, policies, circulars and contracts, cite the page behind each finding and pause high-consequence actions for a named approver. The models run on the institution's GPU server, so its policy covers systems it controls.
We'll walk you through it on sample data.
Recommends a board-approved AI policy.
The regulated entity answers for what its models produce, even through third-party tools.
Sets duties and penalties for how organisations handle personal data.
What it means for your AI: a board that answers for its models' outputs needs a record of what the model read, what it found and who approved the action. Owning the AI that keeps that record makes the policy easier to enforce.
Your AI does the reading and the checking; your staff make the decisions.
| Work | What your AI does | Who decides |
|---|---|---|
| Loan files | Checks documents against your credit policy and cites the page | Your credit officer |
| Policies and circulars | Answers staff questions from your own policy files, with the page cited | — |
| Contracts | Tests clauses, amounts and dates against your checklist | Your legal or ops reviewer |
| Approvals | Irreversible and high-consequence actions pause for a named approver | Your approver |
| Own it | Subscribe | |
|---|---|---|
| Agents and models | Your GPU server, in your data centre | BiltIQ's own servers in India |
| Outbound calls | Blocked by default. In on-prem-required mode, no outbound host can be allowed at all. | — |
| Approvals | Irreversible and high-consequence actions pause for a person | Same |
| Status | On sale now, starting with a paid pilot | Subscription: coming in 2027 |
Write and approve the board policy that RBI's report recommends.
Validate models under your own risk framework.
Collect customer consent and answer data-rights requests under the DPDP Act.
Report breaches to the regulator.
Your AI gives you the record; your governance decides what to do with it.
RBI's FREE-AI report recommends that regulated entities adopt a board-approved AI policy. Read the report itself for its full recommendations. BiltIQ supports such a policy with your own AI: models on your servers, page-cited findings and an approval record for high-consequence actions. The policy and its governance remain the bank's responsibility.
Yes. On an Own-it install, BiltIQ Agent OS and its models run on your GPU server, and outbound calls are blocked by default. Your AI agents check loan documents against your credit policy and cite the page behind each finding. The credit decision stays with your officer. BiltIQ proves this in a 4–6 week paid pilot on past files.
Not on an Own-it install. The security gate in BiltIQ Agent OS blocks outbound calls by default, and an unconfigured install blocks all outbound calls. In on-prem-required mode, no outbound host can be allowed at all. Customer files are read by your AI on your own GPU server. Patent application 202631094208 covers the gate.
Owning your AI with BiltIQ costs a site licence of ₹50–90 L one-time plus a managed service of ₹16–18 L a year, with hardware passed through at cost. Buyers start with a paid pilot of ₹5–8 L over 4–6 weeks, credited against the licence. All prices are ex-GST. The subscription tier is coming soon.