Bank of England finds two-thirds of industry using machine learning
The Bank of England (BoE) and Financial Conduct Authority (FCA) have conducted a joint survey to get an overview of how UK banks and financial services firms are using machine learning (ML), reports Jane Connolly.
Machine learning use increases in financial services
In order to understand the current deployment of ML, along with the risks and challenges posed by it, the survey was sent to almost 300 firms in the industry.
The organisers acknowledge that the 106 responses received are not likely to be statistically representative of the entire industry, but the survey does reveal that two thirds of respondents are already using ML in some form.
On average, live ML applications are used across two business areas and this is expected to double within the next three years.
ML is most commonly used in anti-money laundering and fraud detection, along with customer-facing applications. Other uses include credit risk management, trade pricing and execution, general insuring pricing and underwriting. The banking and insurance sectors show the most advanced deployment of ML.
Related: Bank of England debuts new £20 note
In terms of the risks associated with ML, the most commonly expressed concern is that ML could exacerbate existing risks, rather than create new ones. This is considered most likely to occur if model validation and governance frameworks fall out of step with technological developments.
Common safeguards against risks include alert systems and ‘human-in-the-loop’ mechanisms.
Regarding the barriers to ML implementation, firms say that internal issues such as legacy IT systems and data limitations are the greatest challenges. However, the respondents did call for additional guidance on how to interpret current ML regulation.
Respondents also stressed that their existing model risk management frameworks might have to evolve as ML applications become more sophisticated.
The BoE and FCA have announced plans to set up a public-private group to explore some of these questions surrounding ML innovation.
