Financial crime compliance is struggling to keep pace with the threats it was built to catch, according to new analysis from Napier AI.
The Financial Action Task Force’s April Ministerial Declaration named AI misuse, virtual assets and complex financial structures as top priorities for 2026-2028, a signal that these are already live risks rather than future concerns.
Napier AI points to a small number of typologies exposing weaknesses in anti-money laundering (AML) systems, including stablecoin and digital asset layering, synthetic identity fraud, and trade-based money laundering.
In Australia, the Napier AI / AML Index highlights human trafficking financed through money muling, drug-related flows via unregistered money service bureaus, and smurfing at scale, increasingly automated by criminals using AI to generate transaction patterns and dodge detection.
The figures are stark. Australia lost an estimated $87.39bn to money laundering in 2024-25, and while AI could potentially recover $2.65bn of that, Napier AI notes the wider trend is more concerning: laundering losses are rising 3% year-on-year, while compliance costs are climbing 9%, well above global averages. Australia is currently seen as a leader in balancing compliance costs against outcomes, but Napier AI warns that position is fragile without genuine AI adoption that improves effectiveness rather than simply adding cost.
Regulators are increasingly explicit that AI is not a substitute for compliance judgement. AUSTRAC has said institutions cannot hand over compliance responsibility to AI, and a recent Federal Court of Australia judgement cautioned against using large language models to summarise complex material without human validation.
For Napier AI, this underscores that human-in-the-loop oversight is non-negotiable, particularly for high-risk decisions, and that model outputs must be continuously validated against the knowledge of experienced analysts, not owned solely by data scientists.
Napier AI argues that many of the available gains do not require advanced AI at all. Multi-configuration screening, risk-based segmentation and more granular customer profiling can meaningfully improve outcomes when institutions properly understand their own risk exposure. The bigger issue is legacy infrastructure: most platforms still rely on batch processing and static rules, meaning AI layered on top adds complexity rather than transformation.
The recommendation from Napier AI is to reconnect AI adoption with foundational AML capabilities, starting with a robust risk-based assessment and measuring effectiveness by detection quality rather than false-positive reduction alone. Only once real-time, adaptive infrastructure is in place, Napier AI suggests, can AI deliver on its promise of identifying and stopping the financial crime that matters.
For more, read the full story here.
Copyright © 2026 FinTech Global
