The battle against financial crime is being fought with the wrong architecture, according to Napier AI. While criminal networks collaborate freely and adapt at speed, financial institutions continue to defend themselves in isolation, each building its own models, tuning its own rules and investigating its own alerts.
Napier AI’s Dr Janet Bastiman argues that this institution-centric approach creates a structural weakness. Illicit funds move across firms, jurisdictions and payment systems faster than any single organisation can track them, meaning the industry is only as strong as its weakest link. One vulnerable point in the payment chain can be exploited to shift dirty money through multiple institutions, invisible to each when viewed alone.
Much of this fragmentation stems from legitimate constraints, including technology limitations and regulatory controls designed to protect personally identifiable information, preserve data sovereignty and prevent tipping-off. But the result is an imbalance that favours criminals.
Progress is emerging, however. Initiatives backed by the UK’s Financial Conduct Authority (FCA) are showing how firms can collaborate safely. Napier AI’s work with the FCA, The Alan Turing Institute and Plenitude produced fully synthetic datasets, built from anonymised transaction patterns and layered with realistic typologies, giving institutions a way to train and refine detection strategies without exposing sensitive data.
The next phase, Napier AI believes, is real-time intelligence sharing, where risk signals move securely across a network rather than staying trapped inside institutional walls. Achieving this will demand coordinated regulatory frameworks, trusted data-sharing mechanisms and potentially national or industry-wide utilities, alongside a mindset shift in how firms view their data.
Institutions need not wait for that future to arrive. During a recent project in the FCA’s Supercharged Sandbox, Bastiman modelled transactions as a flowing system, drawing on fluid dynamics. Illicit funds entering the system create subtle ripples that propagate downstream, and by analysing their frequency and amplitude, firms can spot anomalies far from the original injection point, even without full network visibility.
On governance, Napier AI stresses that regulators are focused on outcomes rather than the inner workings of AI models. Firms must consistently demonstrate accuracy, explainability and auditability. Done correctly, AI strengthens all three, enhancing precision, generating human-readable explanations and documenting its own reasoning. Rather than introducing opacity into compliance, it removes it.
For more insights, read the full report here.
Copyright © 2026 FinTech Global
Copyright © 2018 RegTech Analyst
