Financial Services · FSA / IFSA 2013 · BNM Guidelines
Every credit decision
now needs an
explanation.
Malaysia's proposed AI Governance Bill (public consultation since July 2026) and the National AI Office's AIGE principles point to explainable, accountable AI — and a clear record of every loan rejection, credit score and risk assessment is simply good governance. DecisionCore delivers FSA-cited PDF reports in minutes, from a single API call.
AI governance in Malaysia
Proposed
The AI Governance Bill is proposed and under public consultation (Ministry of Digital, July 2026). Explanations you keep now are ready for whatever it finally requires.
Per-decision output
PDF + JSON
Each decision generates a formal A4 PDF (for the applicant) and a structured JSON audit record (for your compliance log and BNM examiner access).
Integration time
< 1 hour
One REST API call. Works with any loan origination system, credit engine, or risk scoring platform. No retraining. No model replacement.
The regulatory picture
Two frameworks.
One obligation.
Financial services firms already work under BNM's policy documents and the PDPA, and Malaysia's proposed AI Governance Bill takes a risk-based approach. An AI-assisted credit decision is exactly the kind of decision worth explaining.
FSA / IFSA 2013 — BNM Policy
Credit Decision Audit Trail
Financial institutions are expected to govern the models behind their decisions and keep records that can be reviewed. A reproducible explanation for each AI-assisted credit decision makes that review — internal, auditor or supervisor — straightforward.
Proposed AI Governance Bill
Explainable, accountable AI
The Ministry of Digital describes a risk-based framework with responsibilities shared by those who build and deploy AI; the final requirements are still being consulted on. A plain-language explanation for each credit, risk or underwriting decision is the practical way to be ready.
Use cases
Where DecisionCore
applies.
Any AI system making a credit, risk or coverage decision about an individual is worth explaining. These are the most common financial services applications.
Consumer lending
Loan & Credit Approval
Every rejected applicant should be able to get a formal explanation. DecisionCore generates a PDF citing the specific DSR threshold, BNM policy provision, and credit bureau factor that drove the rejection. Delivered at the point of decision — not days later.
Risk management
Credit & Risk Scoring
AI-generated credit scores, customer risk tiers, and fraud flags all constitute Significant Decisions. DecisionCore traces each score back to the model inputs and the BNM guidelines that define acceptable criteria — in a format BNM examiners can read.
Insurance
Underwriting Decisions
Premium loading, coverage exclusions and policy rejection decisions made by AI are among the first a customer asks about. DecisionCore integrates with your underwriting engine via REST API — one call per decision generates an explanation record.
Wealth management
Investment Suitability
Robo-advisory and algorithm-driven suitability assessments that affect a client's investment profile fall within the Significant Decision definition. DecisionCore provides the explanation chain and consent record to support your SC compliance documentation.
How it works
One API call.
BNM-ready output.
DecisionCore slots into your existing credit pipeline without touching your model. Send the decision context — receive an explanation report.
01
Send the decision
POST the loan decision and applicant context to the API. Processed on TechnoNet's own servers in Kuala Lumpur — or on a sovereign node in your own network with Enterprise.
POST /explain
{
"decision": "Loan rejected",
"subject_description":
"DSR 62%, institution limit 60%",
"sector": "financial",
"tenant_id": "my-bank"
}
02
DeepSeek R1 reasons
DeepSeek R1 matches the decision against your sovereign regulatory corpus — FSA, IFSA, BNM policy documents — and constructs a traceable explanation path. Every reasoning step stored verbatim.
Matching Credit Policy §3.1...
DSR limit 60%
62% — exceeded
Generating explanation...
Audit ID: ADMT-EXP-8a2f1c
03
Deliver the report
A formal A4 PDF is returned — branded with your institution, citing specific FSA/BNM provisions, with the full XAI reasoning chain and a plain-language rights notice for the applicant.
ADMT-EXP-8a2f1c.pdf
· Rejection summary
· FSA / BNM citations
· DeepSeek R1 chain
· Applicant rights notice
What you get
Built for
financial compliance.
📋
BNM-Ready Audit Trail
Every explanation record stored in an append-only PostgreSQL log with Audit ID, timestamp, decision context, regulatory citations, reasoning chain and record fingerprint — in full for 12 months, then as a verification record for 7 years. Keep your own copy of each report for BNM examination.
📄
FSA-Cited PDF Reports
Each rejection generates a formal PDF citing the specific FSA section, BNM policy provision, or IFSA clause that drove the decision. Hand it to the applicant or your compliance officer — the same document serves both.
🔒
Hosted in Malaysia
The DeepSeek R1 reasoning model, BGE-M3 embedding model and regulatory corpus run on TechnoNet's own servers in Kuala Lumpur — no third-party AI service sees credit data or applicant details. Need everything inside your own network? Enterprise runs on a sovereign node there.
✅
Granular Consent Records
The PDPA requires consent to process personal data for specified purposes. DecisionCore logs every consent grant and withdrawal with timestamp, purpose ID, and consent version — immutably, by design.
🚨
72-hr Breach Detection
Automated monitoring for data breach indicators across your application logs. Alerts your DPO to breach indicators — with a full record written to the Sovereign Audit Log to support any notification to the PDP Commissioner.
🔗
Works with Any System
REST API (POST /explain, POST /explain/report) and MCP server. Integrates with any loan origination system, credit engine, or risk platform in under an hour. No vendor lock-in.
Get started
in under an hour.
3 free explanation reports. No credit card. Hosted in Malaysia, no third-party AI.