The long-held belief that regulation and innovation are fundamentally at odds is losing credibility and the Financial Conduct Authority (FCA) may be the clearest proof yet.
Napier AI’s chief data scientist Janet Bastiman recently took to the stage at Money20/20 Europe to share first-hand experience of working collaborating with the FCA.
For years, financial institutions have treated regulators as gatekeepers rather than partners. That dynamic is shifting. The FCA’s innovation initiatives, particularly its sandbox environments, are repositioning the regulator not as an enforcer waiting at the finish line, but as a co-developer present from the outset. Participation is selective and structured: firms must register on the FCA’s innovation platform, apply to specific cohorts, and demonstrate a genuinely novel use case. More than 200 applications were received for just a handful of places in the Supercharged Sandbox.
Those accepted gain access to curated datasets, APIs, and scalable compute infrastructure, resources that are typically difficult to assemble independently. Each project is overseen by both an FCA representative and an industry mentor, keeping innovation anchored to regulatory expectations throughout.
For Napier AI, the sandbox presented an opportunity to advance network-based financial crime detection, an area constrained by two persistent problems: fragmented data and the exponential compute costs of tracing multi-hop transaction paths. Working within the sandbox’s shared data environment and parallel compute infrastructure, the firm tested approaches that would have been impractical under standard research and development conditions.
One method applied information theory and domain expertise to focus directly on high-risk transaction patterns, reducing false positives earlier in the process rather than relying on costly post-processing.
A second approach borrowed from environmental science: just as river pollution can be detected downstream even without knowing its source, Napier AI’s team found they could identify disruptions in otherwise typical transaction flows, flagging illicit activity passing through a network even with only partial visibility of that network.
Both innovations were developed alongside explainability requirements, ensuring outputs could be reviewed by analysts and fed into suspicious activity reporting.
The lessons from the process are candid: the sandbox is resource-intensive, entering with a clearly defined hypothesis significantly improves outcomes, and some of the most valuable insights arrived late, requiring rapid iteration. These are the realities of meaningful innovation, not reasons to avoid it.
For more insights, read the full story here.
Copyright © 2026 FinTech Global
Copyright © 2018 RegTech Analyst
