“Today’s report makes for sober reading highlighting the many interconnected global risks firms face.
“The Bank of England is right to treat AI resilience as a financial stability issue and not simply a technology risk for individual firms - there is a real danger that banks are becoming reliant on AI faster than they are preparing for large-scale disruption. More than nine in ten banking executives say they could withstand a large-scale AI disruption, yet around half have only conducted one scenario test for such a situation and more than a quarter haven’t done any testing at all. This mismatch of over-confidence and potential under-preparedness is concerning given the speed at which frontier AI could exploit vulnerabilities in new and legacy systems. Without regular, robust testing, firms aren’t going to build the muscle memory to enable a rapid response to incidents. As the backbone of the UK economy, the banking sector is particularly exposed to this new threat landscape. Firms must have cast-iron confidence, and be able to demonstrate credibly, that their use of AI is effective, safe and ethical.
“We often think about threats as external to an organisation, but the rise of autonomous systems can breed danger from within at an unprecedented speed and level of sophistication that could have potentially catastrophic market consequences. So, it’s no surprise how rapidly the policy conversation is moving towards understanding the impact of AI on customer behaviour and practical safeguards to prevent market meltdowns. Autonomous agents challenge assumptions about human oversight and accountability. Firms should expect much sharper scrutiny of how AI behaves under stress, who is accountable and what would happen in the event of a business-wide system failure."
Recent KPMG research of 150 UK banking executives about their AI resilience finds:
- 75% say AI is vital or very important to their core business operations.
- 93% believe they could maintain business continuity in the event of a large-scale AI disruption.
- However, 47% have carried out only one scenario test for the failure of AI-enabled systems, while 26% have not conducted any.
- 55% say their organisation’s investment in AI is outpacing its ability to manage and govern the technology appropriately in certain areas.
- 57% say the growing concentration of AI, cloud and data providers is becoming a systemic risk for UK financial services.
- 40% believe AI governance and controls are the greatest risk to their firm’s operational resilience over the next 3-6 months.
