Starling Bank: A screening engine can run against the wrong population

← All case files

LCF-015 · Sanctions

A screening engine can run against the wrong population

An engine can run on schedule without covering all intended records. Check list ingestion, category selection and customer scope before relying on alert volumes.

Starling Bank · Intermediate · About 5 minutes

The documented case

What the official sources say

Historical event: 2 October 2024
Outcome at that date: FCA financial penalty announced
Jurisdiction: United Kingdom

The FCA announced a £28,959,426 penalty against Starling for sanctions-screening failings and repeated breaches of a restriction on opening high-risk customer accounts. [1]

The FCA said Starling discovered in January 2023 that automated screening had covered only a fraction of the full sanctions list since 2017. It also reported that more than 54,000 accounts were opened for high-risk customers contrary to the restriction. The bank established remediation programmes. [1]

Read the outcome precisely

These are historical regulatory findings, not a declaration that all those customers were sanctioned or criminal. The FCA restriction, financial penalty and any underlying potential sanctions breaches are distinct matters.

What to learn from it

Original LearnAML teaching analysis. The following controls and exercises are not additional findings against the named parties.

  1. Distinguish list-feed availability from screening coverage.
  2. Design a configuration test supported by independent evidence.
  1. 01List source
  2. 02Configuration
  3. 03Screening population
  4. 04Assurance testing

Simplified LearnAML control map; not a reconstruction of every transaction in the case.

Verify coverage before tuning

An engine can run on schedule without covering all intended records. Check list ingestion, category selection and customer scope before relying on alert volumes.

Test restrictions operationally

A restriction must be translated into onboarding rules, escalation routes and exception controls. A policy memo alone does not prove implementation.

Use independent evidence

Compare expected list counts and versions with actual ingested data. Test representative records in a controlled environment and retain evidence of results and sign-off.

Your decision · Fictional practice

What would you do next?

Fictional exercise: A screening vendor dashboard shows successful daily runs and no alerts. During assurance, you discover that only one sanctions-list category is enabled. Operations says this is reassuring because the list feed is technically healthy.

What does the evidence support?

Put the decision into words

Use fictional details only. Do not enter customer information. Optional saves remain on this device, not your LearnAML account.

Compare your reasoning

Use this checklist to improve your answer. It is not an automatically marked essay.

  • Reconcile intended screening coverage against actual configuration and ingested data.
  • Test category and customer scope as well as feed availability.
  • Scope historical exposure without equating a control gap with confirmed matches.

Take one lesson back to work

In your own environment: Reconcile intended screening coverage against actual configuration and ingested data.

Educational analysis, not instructions for a live customer or transaction. Consider your role, authorised procedures and current local requirements.

Sources and editorial record

  1. FCA: Starling financial-crime controls penalty ↗FCA · 2 October 2024Evidence location: Opening findings, January 2023 discovery and remediation paragraphs; Opening findings, January 2023 discovery and remediation paragraphs

Sources accessed and educational analysis prepared: 27 October 2026. Historical snapshot; subsequent proceedings, present custody, current controls and present law are not asserted. Original LearnAML educational analysis. Version 1.0.

About LearnAML