by consulting firm KPMG Which released a research paper on “AI Models for Modern Risk Management” emphasizing the adaptable key pillars of risk management that support the bank’s structure. The report explains in detail how banks An appropriate risk appetite must be identified while establishing such a variety of technology-based models for bespoke assignments.
“With model driven decision making, there has been a wide variation in approach by banks on AI risks around bias, interpretability and other challenges. Hence, banks need to educate all stakeholders, including senior management, on key aspects of AI/ML. Enterprise-wise training programs should be conducted so that they can understand the risks better.” partner and head, financial risk managementKPMG in India.
“It is imperative that a cross functional governance framework be established with a clear definition of roles and accountability. There are key elements that need to be specifically tested during the life cycle of the model, for example design, implementation , during operation and verification. AI and ML models will have risks and challenges that banks and financial institutions do not adequately consider yet model risk management (MRM) has reached. Further increasing regulatory requirements means, they will now have to adapt and extend their existing MRM approach in specific sectors,” Banerjee said.
Banks will need to develop in-house skills or bring in outside experts. External subject matter experts can help them benchmark themselves with peer banks around increased risk management, controls and governance frameworks. This will help in validating the techniques of AI/ML model.
In short, regulatory scrutiny around the AI/ML model is expected to increase in the next few years, according to KMPG, as banks increasingly begin to add more AI/ML models to their inventory.
By 2021, there are detailed proposals from the European Union as well as the Federal Trade Commission (FTC) for stricter AI regulations.
It remains to be seen how this sector develops and regulations take shape over the next few years, but international guidance or standards in this area will be helpful in setting minimum benchmarks for MRM practices in the jurisdiction.