For years, the prevailing assumption in technology circles has been that regulation and innovation are fundamentally at odds.
Napier AI's experience inside the Financial Conduct Authority's (FCA) Supercharged Sandbox suggests the opposite may be true and the implications for financial crime compliance are significant.
Speaking at Money20/20 Europe, Napier AI shared how its participation in the FCA's flagship innovation programme challenged long-held assumptions about the regulator's role. Rather than acting as a bureaucratic gatekeeper, the FCA is increasingly positioning itself as a co-developer of financial technology.
The FCA's sandbox model is far from permissive. Entry is competitive, with over 200 applications submitted for a limited number of cohort places. Successful participants gain access to curated datasets, application programming interfaces (APIs) and scalable compute infrastructure, alongside oversight from both a regulatory representative and an industry mentor.
For Napier AI, the sandbox unlocked opportunities in network-based financial crime detection that would have been impractical under normal research and development conditions. Financial crime rarely stays within a single institution's data boundaries, it moves across accounts, geographies and entities. Detecting it requires both broader data access and significant processing power, two resources the sandbox directly provided.
One breakthrough involved applying information theory to target high-risk transaction patterns more precisely, cutting through the noise that traditional subgraph analysis generates.
Another drew on an unexpected source of inspiration: river pollution monitoring. By treating financial flows the way environmental scientists model contamination, tracking disruptions downstream without needing to identify the source, Napier AI found it could detect fragments of criminal activity even with incomplete network visibility.
The lessons Napier AI draws from the process are instructive: enter with a defined hypothesis, commit dedicated resource, and expect the most valuable insights to arrive late. Innovation, even within a supportive framework, remains demanding work.