InfoBytes Blog
Filter
Subscribe to our InfoBytes Blog weekly newsletter and other publications for news affecting the financial services industry.
Hsu tells banks to approach AI cautiously
On June 16, Acting Comptroller of the Currency Michael J. Hsu warned that the unpredictability of artificial intelligence (AI) can pose significant risks to the financial system. During remarks presented at the American Bankers Association’s Risk and Compliance Conference, Hsu cautioned that banks must manage risks when adopting technologies such as tokenization and AI. Although Hsu reiterated his skepticism of cryptocurrency (covered by InfoBytes here), he acknowledged that AI and blockchain technology (where most tokenization efforts are currently focused) have the potential to present “significant” benefits to the financial system. He explained that trusted blockchains may improve settlement efficiency through tokenization of real-world assets and liabilities by minimizing lags and thereby reducing related frictions, costs, and risks. However, he warned that legal frameworks and risk and compliance capabilities for tokenizing real-world assets and liabilities at scale require further development, especially considering cross-jurisdictional situations and ownership and property rights.
With respect to banks’ adoption of AI, Hsu flagged AI’s “potential to reduce costs and increase efficiencies; improve products, services and performance; strengthen risk management and controls; and expand access to credit and other bank services.” But there are significant challenges, Hsu said, including bias and discrimination challenges in consumer lending, fraud, and risks created from the use of “generative” AI. Alignment is also the core challenge, Hsu said, explaining that because AI systems are built to learn and may not do what they are programed to do, governance and accountability challenges may become an issue. “Who can and should be held accountable for misaligned, unexpected, and harmful outcomes?” Hsu asked, pointing to banks’ use of third parties to develop and support their AI systems as an area of concern.
Hsu advised banks to approach innovation “responsibly and purposefully” and to proceed cautiously while keeping in mind three principles for managing risks: (i) innovate in stages, expand only when ready, and monitor, adjust and repeat; (ii) “build the brakes while building the engine” and ensure risk and compliance professionals are part of the innovation process; and (iii) engage with regulators early and often during the process and ask for permission, not forgiveness.
Fed governor weighs tokenization and AI
On April 20, Federal Reserve Governor Christopher J. Waller spoke on innovation and the future of finance during remarks at the Global Interdependence Center. Commenting that “[i]nnovation is a double-edged sword, with costs and benefits, and different effects on different groups of people,” Waller stressed the importance of considering whether innovation is creating new efficiencies and helping to mitigate risks and increase financial inclusion or whether it is creating new or exacerbating existing risks. Waller’s remarks focused on two specific areas of innovation that he believes may have the potential to deliver substantial benefits to the banking industry: tokenization and artificial intelligence (AI).
With respect to tokenization and tokenized assets, Waller flagged several advantages to innovations in this space that use blockchain over traditional transaction approaches, including (i) being able to offer faster or “even near-real time transfers,” which can, among other things, give parties precise control over settlement times and reduce liquidity risks; and (ii) “smart contract” functionalities, which can help mitigate settlement and counterparty credit risks by constructing and executing transactions based on the meeting of specified conditions. He acknowledged, however, that both innovations introduce risks, including potential cyber vulnerabilities and other risks.
Waller also addressed the banking industry’s use of AI to increase the range of marketing possibilities, expand customer service applications, monitor fraud, and refine credit underwriting processes and analysis, but cautioned that AI also presents “novel risks,” as these models rely on high volumes of data, which can complicate efforts to detect problems or biases in datasets. There is also the “black box” problem where it becomes difficult to explain how outputs are derived, where even AI developers have difficulty understanding exactly how the AI technology approach works, Waller stated. “All of these innovations will have their champions, who make claims about how their innovation will change the world; and I think it’s important to view such claims critically,” Waller said. “But it’s equally important to challenge the doubters, who insist that these innovations are much ado about nothing, or that they will end in disaster.”