KUALA LUMPUR, July 8: Artificial intelligence is transforming global banking at unprecedented speed, but unless its risks are properly governed, it could make future financial crises spread faster, amplify market volatility and create new vulnerabilities beyond the control of individual banks.
Rather than simply improving efficiency, AI is increasingly becoming part of the financial system’s core infrastructure, raising concerns over technological concentration, algorithmic bias, cyber threats and the growing dependence of financial institutions on a handful of technology providers.
These were among the key warnings delivered today by Pietro Calice, Senior Financial Sector Expert in the Finance, Competitiveness and Investment Department of the World Bank Group, during his keynote address, The Future of Global Finance – When Speed Outruns Judgement, at the Bank Audit Conference (BAC) held at the Kuala Lumpur Convention Centre (KLCC) today.
Speaking to banking executives, regulators and audit professionals, Calice said AI should no longer be viewed as another digital tool but as infrastructure that increasingly shapes how banks assess borrowers, price risk, monitor compliance and manage liquidity. While these capabilities promise significant gains in productivity and decision-making, they also introduce new forms of systemic risk that conventional banking oversight was never designed to address.
To illustrate how quickly technology could transmit financial stress, Calice described a hypothetical scenario in which a geopolitical announcement triggers an immediate chain reaction across global financial markets. Within minutes, AI systems identify exposed companies, markets rapidly reprice assets, banks reassess lending risks and liquidity positions, while a convincing AI-generated video falsely claims that a major regional bank has suffered a technology failure, prompting customers to move their money before regulators have time to fully assess the situation.
“This is not a prediction. Our institutions are becoming faster, more connected, and more intelligent, but it is much less clear that they are becoming more resilient.”
AI and financial stability
Calice address focused on around three areas where AI could fundamentally reshape the financial system: financial stability, credit markets and competition.
On financial stability, he warned that AI may make individual financial institutions more sophisticated, but that does not necessarily translate into a more stable financial system. As banks, investment funds and trading firms increasingly rely on similar datasets, machine-learning models and risk parameters, they are more likely to respond to the same market signals simultaneously.
Rather than one institution making a mistake, the greater danger lies in many institutions independently making the same rational decision at exactly the same time. That collective behaviour, he said, could trigger rapid asset sell-offs and amplify market volatility.
Calice pointed to the 2023 banking turmoil in the United States as an example of how digital banking and social media have already accelerated financial contagion. AI, he said, adds another layer by processing information faster, generating convincing false content at scale, recommending actions and, in some cases, executing them automatically.
“The operational clock of finance is accelerating. The governance clock is not yet.”
Expanding credit without sacrificing fairness
Turning to credit markets, Calice said AI has the potential to improve financial inclusion, particularly across Asia where micro, small and medium enterprises account for almost all businesses. Machine-learning models using digital footprints can often assess borrowers as effectively as conventional credit bureau data, helping banks extend credit to viable businesses and households with limited formal credit histories.
However, he cautioned that greater predictive accuracy does not necessarily produce fairer outcomes. Research has shown that AI models can improve prediction while widening disparities in lending outcomes across different groups, even after sensitive variables such as gender, ethnicity and location have been removed from the data.
“The governance question is not simply, did we use a prohibited variable? It is what outcome did the model produce, who received credit, who was rejected, who could challenge that decision, and who was accountable when the model was wrong?”
He also challenged the commonly used phrase “human in the loop”, arguing that simply inserting a person into an automated process does not guarantee meaningful oversight.
“A human sometimes has 10 seconds to approve an automated decision that is not exercising judgment. A human who cannot understand the model or reverse the decision is not providing oversight.”
Instead, he said meaningful human involvement requires time, competence, information and authority. AI should complement experienced bankers rather than replace them, with human judgement providing context and insight that machines cannot replicate.
Competition in the AI era
The third area of concern was competition. Calice cited studies showing that independent pricing algorithms can learn to sustain prices above competitive levels without any communication between companies. While this is not evidence of widespread algorithmic collusion in banking, he said it demonstrates how anti-competitive outcomes can emerge without explicit agreements between human decision-makers.
As a result, regulators may increasingly need to examine market interactions and outcomes rather than focusing solely on communications and intent.
“Across these three channels – stability, credit markets and competition – the lesson is the same,” he said. “Institution-level intelligence is not a substitute for system-level design.”
Managing geopolitical and technology risks
Calice also warned that global finance is becoming increasingly interconnected at a time when the global economy is becoming more fragmented through sanctions, export controls, investment restrictions and data localisation requirements.
For ASEAN economies such as Malaysia, whose prosperity depends heavily on cross-border trade, payments, investment and digital infrastructure, disruptions involving technology providers, payment systems or major trading partners could quickly spread across firms, banks and jurisdictions.
He urged financial institutions to treat technology risk and geopolitical risk as interconnected, incorporating scenarios such as cloud outages during periods of market stress, restrictions on technology providers and cyberattacks into routine stress testing.
“Redundancy costs money, but resilience often looks inefficient until the day it is needed.”
To strengthen resilience, Calice proposed five priorities for banks: treating AI as a board-level issue, maintaining comprehensive inventories of AI systems and technology dependencies, validating AI models beyond technical accuracy, improving the ability to substitute critical technology providers and preserving the human expertise needed to challenge automated decisions.
He also called on supervisors to strengthen their own technical capabilities.
“A supervisor cannot challenge a model it cannot understand.”
In closing, Calice said the institutions most likely to succeed in the AI era would not necessarily be those with the largest AI models or the greatest computing power, but those capable of governing AI responsibly while preserving human judgement.
“Machines will always process information faster than people. But the enduring advantage of banking is the institutional capacity to combine information with judgment, to take risk across time, to support customers through uncertainty, and to remain accountable when decisions have consequences.”
He concluded that the future of global finance would ultimately depend not on how quickly AI evolves, but on whether governance can keep pace with technological change.
Media & PR: editor@dailystraits.com. Copyright 2021–Present DailyStraits.com. All rights reserved.