RI

AGT-1021 · Financial Crime

AML Investigation Agent

Assembles alert narratives and proposes STR disposition.

critical riskcompliant

Trust score

85

grounding, fairness, oversight

Requests / 30d

31,438

model Azure OpenAI · gpt-4.1

Human approval

Required

data class Restricted

Last audit

2026-05-06

version v4.3

Agent architecture

Request path from the surface the customer or staff member touches, through the governance gateway, to the model and its wired systems

Entry surfaces

Retail Loan Origination

 

Business application · In-house (LOS)

Governance gateway

Truint policy enforcement point

9-stage filter chain · p99 54 ms

Evidence ledger writer

Hash-chained, RFC-3161 anchored

Agent runtime

AML Investigation Agent

AGT-1021 · v4.3 · trust 85

Azure OpenAI · gpt-4.1

 

LLM / model endpoint · Central India

Tools & data

core.accounts.close

 

Tool / function · Restricted

Policy & Product Knowledge Base

 

Data source · Internal

Enforcement chain on every request

  1. 1Identity & entitlementCaller identity resolved; agent entitlement checked against the registry record.
  2. 2Kill switchEstate and per-agent revocation state for AGT-1021 evaluated before any spend.
  3. 3Manifest checkOnly integrations on the signed manifest below are reachable.
  4. 4Budget & rateToken, cost and concurrency budgets for Financial Crime.
  5. 5Injection screenPrompt and retrieved context screened for untrusted-content patterns.
  6. 6DLP & maskingRestricted payloads masked or tokenised inline.
  7. 7Policy decisionCompiled policy bundle evaluated; allow, warn or block.
  8. 8Human in the loopRestricted actions parked for a named approver before execution.
  9. 9Ledger barrierDecision hash-chained into the audit ledger before the response is released.

Registry record

Source of truth for this agent

Registry ID
AGT-1021
Business owner
T. Sengupta
Department
Financial Crime
Model endpoint
Azure OpenAI · gpt-4.1
Version
v4.3
Data classification
Restricted
Human-in-the-loop
Enforced on restricted actions

Trust composition

Grounding & citations89
Fairness79
Security posture87
Oversight coverage96

Wired integrations

Signed manifest — nothing else is reachable

Azure OpenAI · gpt-4.1connected

LLM / model endpoint · Microsoft Azure · scopes chat.completions, embeddings

core.accounts.closeconnected

Tool / function · Core banking (Finacle) · scopes close

Retail Loan Originationconnected

Business application · In-house (LOS) · scopes application.read, decision.write

Policy & Product Knowledge Baseconnected

Data source · In-house vector store · scopes retrieve

Policies in force

POL-002 · No account closure without approvalPOL-003 · No loan rejection without explanationPOL-004 · No customer data sent to public LLMPOL-005 · High-value transaction requires human approvalPOL-006 · PII masking requiredPOL-008 · Prompt injection protectionPOL-011 · Narrative grounding required

Recent decisions

Live gateway stream filtered to this agent

11:39:01emp-20663allow

Why did alert AML-77120 trigger?

POL-011 · Narrative grounding required · risk 25 · 961 ms

11:37:07emp-20029allow

Why did alert AML-77120 trigger?

POL-011 · Narrative grounding required · risk 10 · 338 ms

11:34:23emp-20847allow

Why did alert AML-77120 trigger?

POL-011 · Narrative grounding required · risk 7 · 480 ms

Open risks

Drill into the investigation

No open risks on this agent.

Ledger entries

Hash-chained decision records

Escalated to FIU review

AUD-918438 · 2026-07-26 11:40:12.885

Governance obligations

  • Quarterly attestation by T. Sengupta under MB-04
  • Restricted actions parked for a named human approver
  • Kill switch tested; revocation propagates in under 30 seconds
  • Evidence retained ten years for regulator inspection