AI adoption spurs credit optimism, yet challenges exists for financial institutions

February 28, 2024 (MLN): The rapid adoption of artificial intelligence by companies, aimed at supporting innovation, growth, and efficiency, is deemed a credit positive in the medium and long term, however, its integration poses significant challenges for financial institutions, Fitch Ratings says.

AI is an emerging risk component to ratings that can be both positive and negative, depending upon a company’s use case, governance, risk management, sector, regulatory oversight and competitive environment.

AI, which encompasses technologies including machine learning (ML), deep learning, natural language processing, robotics, expert systems and fuzzy logic, can improve operational efficiency and reduce costs for financial institutions.

Sub-sectors that benefit the most from AI are typically those with high volumes of clients or transactions, where use cases such as chat bots, fraud detection and process automation can more readily be applied, or with significant software development costs.

The rapid deployment of generative AI, which creates content ranging from text and images to code and music from simple user inputs, has mainstreamed the use of AI across sectors.

However, AI technology can lack transparency in explaining its decision process, leading to the potential for inadvertent discrimination. These challenges will reduce its use case in areas such as consumer lending that have more strict regulatory oversight.

Nevertheless, in fraud detection and risk management AI can help detect, identify, and respond to unusual patterns and risks more quickly.

Automation can be used to reduce risk of cyber attacks, requiring access request to be fully authenticated, authorized, and encrypted, requiring device health verification before connection to a corporate network.Top of Form

Sectors that are less regulated can have higher speeds of adoption and, potentially, increased criminal activity. Criminals may attempt to reverse-engineer or hijack AI models if ransoms are not paid; taint or manipulate data used to train models; or steal algorithms or data.

Cyber criminals may also use AI themselves to attack IT systems. Whether AI may increase or reduce vulnerability to cyber-attacks depends on multiple factors, such as quality of the systems or agile risk management.

This includes working with third-party providers, such as cloud system providers, to assess the security of critical ML and AI systems.

Financial services companies are increasingly relying upon third-party services to support their AI operations. Prioritizing robust operational resilience, including arrangements over third-party providers, is increasingly critical, and an area where policymakers are toughening requirements to mitigate firm-level risks and any systemic implications.

In addition, the U.S. Congress has introduced hundreds of AI-related bills to address issues such as protecting people’s likeness and voice, mitigating the spread of fake audio and video, and limiting risks to customer account security, such as a September 2023 bill to address AI-powered, deep-fake bank scams.

Recently, a bipartisan, 12-member U.S. Congressional task force on AI was launched. Increased regulatory scrutiny could reduce the usage of AI in areas such as underwriting or hiring, but robust regulation is generally considered supportive of credit.

Copyright Mettis Link News