AI: an opportunity to define a framework for financial players

3 questions for Anne-Cécile Krieg, Deputy Head of Model Risk Management. The publication, back in July, of a white paper on operationalising the management of artificial intelligence systems’ risks shows, if there were still any need, to what extent the issue of Artificial Intelligence (AI) represents an accelerator of competitiveness and, at the same time, an accelerator of risks for financial institutions.

So how should we address this challenge? We review all this with Anne-Cécile Krieg, Deputy Head of Model Risk Management, who spoke at the ACPR conference of 26 November on this very issue.

Anne-Cécile, what are the key benefits AI offers financial sector players in terms of competitiveness?

I think that AI’s benefits in our universe no longer need to be demonstrated! Processes such as examining documents and classifying data are considerably accelerated by artificial intelligence. It is also a technology that we use and has proven itself, notably in compliance screening. Major flagship projects have also provided benefits for our clients while strengthening our risk management. I’m notably thinking of the revamping of the supervision of approved overdrafts for individual clients or the MOSAIC (More Security With Artificial Intelligence) project enabling processes to be automated with regard to detecting means of payment fraud, with flow analysis and the triggering of alerts as soon as abnormal or suspicious events are detected. 

And, on the other hand, what are the key risks?

Artificial intelligence entails two main types of risks that are two sides of the same coin: the risk of foregoing AI because of fear and the risk of not providing a sufficient framework for using it.

This framework must notably consider technical challenges amplified by more sophisticated approaches (coded bias, for example) or new approaches within the context of AI (such as cybersecurity). It also requires a multidisciplinary analysis of the risks (IT, model or operational).

How is a group like ours organising itself to incorporate AI while controlling risk?

Compared to other industries, we are fortunate to already have a framework on which we can capitalise; it is based on regulatory expectations, for example regarding modelling, data, or control. We are enhancing this framework to counter specific challenges resulting from AI, notably based on a collective approach undertaken with our peers to identify best practice. This fruitful collaboration has led to a white paper on this topic involving BNPP, La Banque Postale and Societe Generale. This document reviews all the risks that have been identified and recommends a sort of “toolbox” for each of them.

In-house training is also a key issue that requires a larger number of channels to involve all hierarchical levels and provide both general and expert training. Learning and knowledge, as we know, act as safeguards.

Sharing best practice, feedback, training – it is assuredly these empirical dimensions that will ensure a well-thought-out deployment of the undeniable benefits of AI.

Read more: the white paper on AI gen