Machine learning for anti-money laundering in iGaming

Why iGaming is a money-laundering target

Online gaming moves large volumes of money quickly, across many payment methods and often many accounts. That makes it attractive to anyone trying to make illicit funds look legitimate.

Regulators have noticed. News reports on the Commission's assessment said it wanted operators to lower the winnings threshold for checks below the current €2,000 and to clamp down on multi-accounting.

Deduplication: finding one person behind many accounts

Launderers rarely use a single account. Deduplication uses fuzzy matching to score how similar two records are across names, addresses, devices and payment details, then links accounts that probably belong to the same person.

Without it, every account looks small and innocent. Linked together, the pattern becomes obvious.

Anomaly detection: learning what suspicious looks like

An ML model learns from historical suspicious activity and flags new behaviour that resembles it, in near real time. Two patterns matter most in gaming:

• Money dumping: a player deliberately loses to another account to move funds between them.
• Mule accounts: accounts opened to spread and layer cash so no single balance draws attention.

What makes an AML model fit for gaming

Generic AML tools are built for banks. A gaming AML model should:

• Be built for gaming and sports betting behaviour, not retail banking.
• Follow the journey of money from deposit to encashment alongside play behaviour.
• Flag red flags such as sudden deposit spikes, withdrawals after minimal wagering, and deliberate losses.
• Sort players into high, medium, low and no-risk buckets so compliance teams know where to start.
• Apply country-specific regulations, because thresholds and duties differ by licence.