Head of Compliance and Regulatory Strategy @ Silent Eight | AI, Data Privacy & Financial Crime | Fulbright-Schuman Scholar | Translating Regulation to Technology and Back Again
This the inaugural post for my "Questions I wish you would ask your data service" Series. #Definitions "Can you explain what that means?" “How did you define X when the legal meanings vary across jurisdictions?” Labels mean different things to different people and typically don't tell me what is in that specific data bucket. ?? Apples are mixed with ?? oranges and sometimes ?? grapes are thrown in too because 'hey it's data and more is better!' But that introduces ambiguity, skews results, and in some places means that the data cannot be collected or used for certain purposes (e.g. in the EU there’s no legal basis for reputational data collection, but there is for AML/CFT.) ??Fruit basket data is a problem for #AI or #risk models in #AML #CFT #sanctions and a problem for #dataprotection - also focused on accuracy for many of the same reasons as #financialcrime - to protect individuals. Determine the goal and the use case. Identify and define the data needed to achieve it. Research. ·????????International organizations are not non-governmental organizations. ·????????Bank Fraud is not Wire Fraud, nor is it Mortgage Fraud. ·????????Bribery and Corruption are related, but there are distinct differences. ·????????Organized Crime is not Terrorism. These groups use similar methods of violence and coercion, but for different reasons. Without clear definitions can you identify what data should be collected for the goal? How do you know if that data is fit for purpose? How can you determine data quality if you haven’t defined the standards? You can't. Clear definitions are a foundation of good data methodology that shows regulators and your clients (if you are a data vendor) why you chose that data and how you got there. Document it. Process is important. You need subject matter experts, not just data scientists, to tell you how the world works so your data reflects reality. Yes, we are a diverse world of cultures, ideas, government systems, laws etc.?It is possible to find similarities, set inclusion and exclusion ranges, to achieve a broad standard. The definition question is hard for most to answer because they haven't thought about it. That’s a red flag. ?? I know that interoperability is an issue, and the devotion to APIs (also rarely defined) is core to #AML #CFT tech, but it's usually a profit issue. Definitions are tedious, but that’s not a good reason to ignore them. ·????????Do definitions reflect legal and regulatory considerations or are they just copy/paste without context? ·????????If the response is “it’s FATF” that’s not good enough.?Execution of the Recommendations involves interpretation.?What’s their interpretation and how did they arrive at it? ·????????Do international organization or industry standards exist and if so how have they been integrated? ·????????What jurisdictional discrepancies did you find and how did you handle them? Want to learn more??DM to schedule a consultation. Frasher Risk Advisory LLC