Managing AI in banking: are we ready to cooperate?

Keynote speech by Pablo Hernández de Cos, Chair of the Basel Committee on Banking Supervision and Governor of the Bank of Spain, at the Institute of International Finance Global Outlook Forum, Washington DC, 17 April 2024.

Good morning, and thank you for inviting me to speak at this conference today.

In 1986, the historian Melvin Kranzberg published his six laws of technology. At the top of his list was his view that "technology is neither good nor bad; nor is it neutral". Fast forward to today, and this edict could seamlessly apply to the debate about the use of artificial intelligence (AI) and machine learning (ML) in banking.

Discussions about the promises and pitfalls of the use of generative AI and large language models in banking are becoming increasingly common. That talk is also being turned to action, with banks starting to use and invest in AI/ML. More boldly, some are already talking about the potential impact of the yet-to-be-established artificial general intelligence on banking.     

There is now an emerging narrative that lauds the purported benefits of AI in banking – in terms of operational efficiencies and improved risk management – while also cautioning about challenges, ranging from data privacy to model hallucinations to reputational risk.

Yet we're still left with an unanswered question: is the use of AI/ML in banking a net positive or negative to global financial stability, and perhaps society more generally? Have we thought through all of the potential scenarios that could play out in a world where AI/ML plays a prominent role in banking? Are we at risk of our own "consensual hallucinations" about AI/ML if we fail to take a step back and consider the bigger picture?

I shall not attempt to provide a definitive answer to these questions today. Instead, I will follow Kranzberg's framing and try to bring together both the micro- and macroprudential and financial stability considerations. 

My main message is that the use of AI in banking raises important prudential and financial stability challenges. We've yet to see how AI/ML performs across a full financial cycle – and this could be some time off. Left unchecked, such models could potentially amplify future banking crises. But these challenges and limitations are not insurmountable, provided that central banks and supervisory authorities adjust to this new reality and collaborate effectively.