New CBN rules make manual AML non-compliant.

Read the brief
MaiGuard

Release Notes

Changelog

Significant changes to the MaiGuard platform, API, and documentation.

  1. Latest

    One customer risk model

    • Declared risk profile on Verify — pass what you know about a subject (sector, product, channel) as risk_profile on /v1/verify/identity, and the identity's AML risk report scores those dimensions instead of leaving them blank; a risk level you declare acts as a floor
    • One bounded risk model — the identity risk report and the customer risk score now share the same five signed-off dimensions and the same weights, so the two views of one person can no longer disagree about the model; every dimension keeps a minimum weight nobody can switch off
    • Honest partial views — a dimension with no evidence is excluded and listed as “not assessed” on the report, never scored at a made-up neutral, so a low score can't masquerade as a full assessment
    • Verifications that create customers (upsert linking) now carry the declared risk profile onto the customer record — declare once, both the identity report and the customer risk score hear it
  2. Entity policies

    • Entity policies — apply per-customer rules like VIP limits or enhanced monitoring that take priority over your general rules, so you can handle individual customers differently without changing tenant-wide settings
    • Shadow mode for entity policies — trial new per-customer rules without affecting live decisions until you're confident
  3. MaiGuard Assist

    • MaiGuard Assist — ask questions in plain language to look up transactions, explain anomalies, and get policy answers right in the dashboard, so analysts get answers without manual digging
    • Connect MaiGuard to AI assistants and agents, so your team can work with fraud data in the tools they already use
  4. Machine learning scoring

    • Machine-learning risk scoring alongside your rules — adaptive scores from transaction and behavioural signals catch more fraud than rules alone
    • Compare ML against your rule-based decisions in shadow mode before turning it on, so you can adopt it with confidence
  5. Launch

    Platform launch

    • Real-time fraud scoring API — score transactions and act on the results as they happen
    • Shadow mode, entity graph, and case manager available from day one to test rules safely, surface linked activity, and work alerts
    • Browser SDK to capture device and behavioural signals from your web app for stronger scoring

Was this page helpful?