Detection & investigation for banks

Share intelligence.
Not data.

Find money-laundering patterns. Bring the evidence into the investigation. Learn across institutions while each bank keeps control of its data.

Explore the products

AMLR applies from July 2027. Banks that want to collaborate need the infrastructure in place.

SHARE INTELLIGENCE, NOT DATA ARCHITECTURE
Bank ABank-controlled environment

Customer identities
Transaction records
Case evidence

Stays with the bank
Bank BBank-controlled environment

Customer identities
Transaction records
Case evidence

Stays with the bank
Bank CBank-controlled environment

Customer identities
Transaction records
Case evidence

Stays with the bank

What travels
Protected model updates · shared model

A shared model

Learn across institutions. Score activity inside each bank.

Case-specific sharing follows a separate, authorised disclosure path.

Built for a bank's scrutiny

Your data stays in the bank

Customer identities, transactions and case evidence remain in the bank's own environment; only protected model updates travel.

Data protection →

Tested with the regulator

We are working through AMLR, GDPR and the AI Act in the Danish FSA's AMLR Article 75 sandbox.

Regulatory path →

Your analyst decides

The model ranks and prepares evidence. Escalation and reporting stay with your analyst and MLRO.

Investigator →

Ready for bank diligence

ISO/IEC 27001 programme in progress, with security, privacy and deployment material ready for your review teams.

Data protection →

It takes a network to beat a network

One institution sees a fragment.
A network spans many.

Money-laundering networks move between banks. Each investigator works with the part their bank can see, while false alerts compete for their time.

90–95%

Estimated false-positive alerts cited in the BIS's 2023 Aurora report on rules-based monitoring.

BIS Innovation Hub, Project Aurora, p. 19 ↗Industry context from the report, not a Node16.ai performance result.

What the bank gets

Fewer false alerts.
More real cases.
Decisions you can defend.

Real cases, not noise.

In our synthetic experiments, a focused queue was almost entirely relevant alerts.

Patterns across banks. Data within each.

The model learns from patterns no single bank sees whole. Each bank keeps its data.

Evidence first. Every decision recorded.

An audit trail the MLRO can show the regulator.

You choose the balance between capacity and coverage. All results come from our synthetic development environment; real-bank performance is what we test next with partner banks. What this means for your team →

Who does what

The model ranks.
Your analyst decides.

What Sonar and Investigator do

  • Score activity in the bank's own data and rank it for review
  • Show the pattern and the accounts behind it
  • Draft the narrative, with every sentence linked to its evidence

Flagged cases go to your analyst

What your team does

  • Review the evidence and ask the next question
  • Decide: escalate, request more information, or close
  • Report to the FIU, with the decision recorded in the audit trail

How it works

Detect. Investigate.
Then collaborate.

01 / SONAR

Find relevant activity

Graph-based detection on the bank's own activity, evaluated at the team's investigation capacity.

Explore Sonar →
02 / INVESTIGATOR

Put evidence to work

Graph exploration, KYC context, case decisions and reporting preparation in one workflow. Includes Sonar.

Explore Investigator →
03 / COLLABORATOR

Connect the wider picture

Governed information sharing under AMLR Article 75 as the network and legal arrangements develop. Target July 2027.

Explore Collaborator →

Synthetic data · The development corpus

4BTransactions
4MCustomers
187Money-laundering typologies in the corpus

Known patterns at scale let us test detection and investigation before the work with partner banks begins.

The next step for your bank

See the signal.
Follow the evidence.

Request a briefing on Sonar and Investigator, with time for your integration and operating questions.

Request a briefing →Download the white paper · v0.1 ↓