FCA speaks on global capital market regulation and AMLTech

Megan Butler has spoken on the fundamental importance of co-ordination between national regulators, particularly in dealing with financial crime and cyber-threats. She looked at the particular benefits and risks that machine learning and algorithmic trading have brought. She stressed the importance of firms reporting material breaches, and the need to let FCA know when they are attacked. She said timely reporting was crucial not least so the regulators can help firms to respond. She then moved on to discuss the escalation of money laundering through capital markets, which has become an increasing risk. It is therefore key to get effective AML controls throughout the capital markets. FCA is not criticising firms in the sector as it believes their “tone from the top” is good. But they must become more alert to the risks that go wider than the traditional risks of insider dealing and market manipulation. The next day, Rob Gruppetta spoke on using AI to keep criminal funds out of the financial system. A key theme of his speech was around how machines could be trained to be alert to human suspicions. Increasingly, only a small number of customers and transactions meet a human being, so it is key that software can pick out the suspicions that humans would, and maybe more. There are concerns currently because regulators find it hard to become comfortable that machine learning techniques are effective, and the extent to which machine learning complements, replaces or duplicates other parts of the process.  He noted FSB’s report which addressed these issues, but could not provide an answer. Currently it seems that there are both big potential benefits from using machine learning to tackle money laundering but also limitations – such as unhelpful effects of changes in the law, or an institution’s staff gaming the system. But he said there are non-technological challenges too, such as patchy data quality and lack of visibility of an entire transaction.

%d bloggers like this: