Financial institutions are increasingly aware that artificial intelligence (AI) could transform anti-money laundering (AML) operations. Yet adoption remains cautious.
This hesitancy stands in sharp contrast to other areas of financial services where AI is being embedded at pace. The reason is clear: financial crime compliance is among the most heavily regulated functions across banking, payments, WealthTech, InsurTech and gambling.
Napier AI recently delved into how firms can implement compliance-first AI for AML.
According to the Napier AI / AML Index 2025–2026, global money laundering losses are at least $5.5tn. At the same time, regulated firms could save up to $183bn annually in compliance costs through AI-driven systems, while global economies could recover more than $3.3tn each year by reducing illicit flows.
A recent micro survey of financial crime professionals highlights where the market stands today. Only two respondents said AI for AML was “well embedded” in their organisation. Most were either in early deployment or testing phases, while a small minority had no plans at all. Looking ahead 12–24 months, optimism grows. Seven expect AI to be well embedded, while 12 believe they will still be in early-stage deployment.
Any AI deployment in AML must align closely with regulatory expectations. In the UK, the Financial Conduct Authority (FCA) has taken a proactive stance. Through initiatives such as the Synthetic Data Expert Group, the Supercharged Sandbox in collaboration with NVIDIA, and AI live testing within the FCA’s AI Lab, regulators are encouraging innovation while maintaining safeguards. Public-private partnerships are also emerging. Napier AI has worked with The Alan Turing Institute, Plenitude Consulting and the FCA to build synthetic datasets based on anonymised real transactions, enabling firms to train detection algorithms safely.
Globally, regulators are following similar paths. The Monetary Authority of Singapore (MAS) launched Project MindForge to examine generative AI risks and opportunities. Bank Negara Malaysia (BNM) has issued a discussion paper on AI governance. The European Banking Authority (EBA) is consulting on new AMLA mandates, while the U.S. Department of the Treasury has sought industry input on AI risks and uses.
Despite regulatory support, implementation challenges remain substantial. Internal data restrictions often limit how compliance policies can be surfaced to AI models. Decades of remediation projects mean data is fragmented or poorly understood. Many firms are now conducting “data clinics” to assess usability before feeding information into AML models, Napier AI said.
Policy and process complexity presents another hurdle. Financial crime compliance frameworks are deeply rooted in documentation and manual controls. Automating policy interpretation and embedding updates into operational workflows is still evolving. Some organisations are exploring AI tools capable of querying legacy materials to support analyst decision-making, but this may require rewriting policies to make them machine-readable, Napier stated.
Explainability and transparency are critical. Compliance officers remain accountable to regulators and law enforcement, necessitating human-in-the-loop oversight. Black-box models have previously fallen short of regulatory standards.
For institutions beginning their AI journey, transaction monitoring may appear attractive but can be complex due to grey regulatory areas. Sanctions screening, by contrast, offers clearer decision boundaries and may provide a more manageable starting point. Ultimately, successful AI adoption in AML hinges on designing systems that prioritise transparency, auditability and fairness from the outset.
For more insights, read the story here.
Read the daily RegTech news
Copyright © 2026 RegTech Anlayst
Copyright © 2018 RegTech Analyst
