Is AI the ultimate weapon in the fight against Financial Crime? Well, Sebastian Drave, Harbr’s Chief Data Scientist, recently participated in a panel discussion on this topic. Reflecting on the session, the audience engagement, and the wider Fincrime World Forum event, it was incredibly encouraging to sense that the conversation around data sharing as a key pillar of the effort in the detection and prevention of financial crime is becoming much more mature and nuanced. 

Financial Crime is quite a unique problem space for analytics and data science. This is due to a number of purely technical factors, such as the availability of accurate training data and the operational implications of generating ‘alerts’ within regulated institutions, but also due to the nature of the activity itself. These are 4 key takeaways from the discussion:

  • Effective Financial Criminals hide in plain sight, displaying consistent behavior for extended periods of time
  • Criminal activities will often sit within the mid-range of activity distributions of a client book as a whole
  • Criminal enterprises purposefully set up their operations across multiple banks, in multiple jurisdictions to create obfuscation as each organization only has a very limited viewpoint on its activities
  • Key lead intelligence is mastered by a small number of organizations

As a result of these, and other, factors, no technology, capability or organization is able to effectively combat this issue in isolation. Rather it requires a coordinated effort across multiple parties to make a step-change in the combative capabilities that the financial services industry, law enforcement, and governments can bring to bear against sophisticated financial criminals. 

In the case of the utilization of AI in this framework, the consideration is the establishment of the wider ecosystem required to allow the capabilities offered by AI to make an impact. In recent times the term AI is increasingly being used in isolation to Data Science. Whilst this is in some parts driven by the enveloping of AI / Machine Learning capabilities into products with wider user bases, ultimately these two areas can’t be separated as the development of a new AI capability is at its core a research and development task which should be approached with a scientific methodology. Once this R&D phase is complete/mature, however, there is a much wider operationalization piece that needs to be considered to allow an algorithmic capability to be used more widely. A great parallel example of this concept has recently been shown in the development and distribution of COVID-19 vaccines. Whilst obviously driven in urgency by the wider unprecedented situation, the vaccine programs were not starting from scratch in terms of the integration of the pure R&D development of the vaccines themselves and the trials, regulatory approvals, and manufacturing capabilities required to take them safely into the wider world. Ahead of the emergence of the Global Pandemic, the Oxford University Jenner Institute team lead by Professor Sarah Gilbert had already established the ‘Disease X’ program, which laid the foundational pathway to take the proven scientific concept of a vaccine for a novel and dangerous disease into industrial-scale production and deployment. Considering AI from such a perspective it is clear that this foundational pathway is also required to accelerate the capabilities provided by trained AI outputs into production financial crime detection and risk management systems which have similar, though distinct, regulatory demands, oversights, and monitoring. 

Where the above analogy diverges, however, is the availability of the necessary data to perform this initial stage. In the case of vaccine development, it was just the genetic sequence of the COVID-19 virus that was required (though within that statement is hidden a myriad of groundbreaking scientific discoveries and capabilities, from both past and present) whereas no such single source of truth exists for the development of AI capabilities to combat financial crime. Furthermore, the factors outlined above amplify the challenge as developing such a viewpoint within a single organization would be challenging, doing so when the actors are purposefully executing their operations across many institutions is significantly more so. 

This was a realization I came to while working in previous financial crime data science roles and was a big driver in my decision to join Harbr. The vision of secure, multi-organization data exchanges that enabled the effective collaboration across multiple organizations data, whilst ensuring all parties maintain full custody of their content, to derive new value and actionable insights was so highly aligned to the challenges we were facing. Through joining Harbr it’s also become apparent that this same requirement exists across the vast majority of industry verticals, for a multitude of reasons. As the adoption and deployment of AI capabilities have progressed, ever more organizations have begun to realize that their data is only ever part of the picture and that a wider, context fulfilling view – accessible in a manner with which free form analytics and R&D can be safely applied, is required to truly realize the power that these capabilities can provide. 

While my current role as Head of Customer Solutions sees my work across multiple different areas, financial crime is an area that I am still very passionate about. It was a pleasure to be involved in the FinCrime World Forum event and it was great to see that the industry as a whole is in much greater alignment on the core challenges and need for greater data sharing to combat the challenge and societal damage created by financial criminals. Many thanks to the FinCrime World Forum team and Patrick Craig (EY) for coordinating and chairing the ‘Is AI the ultimate weapon in the fight against Financial Crime?’ panel session. You can now watch the FinCrime World Forum on-demand.