Similar to most industries, the financial services sector is being revolutionized by Artificial Intelligence (AI). AI in the financial services sector promises improvements in efficiency, innovation, and risk management; however, as we all know, there is no such thing as a risk-free investment. The adoption of AI in the critical financial services sector comes with a number of risks, some of which are potentially economically systemic. As AI systems are increasingly deployed in the financial services sector, effective governance is paramount to ensure we limit our collective exposure to these systemic risks.

Risks of AI in the Financial Services Sector

Like any industry, introducing AI to the financial services sector comes with a set of known risks, such as algorithmic bias and hallucinations. However, as we saw just 15 years ago with the sub-prime mortgage crisis, the financial services sector is unique in that it permeates every other sector of the economy. This characteristic of the financial services sector mandates we constantly evolve our governance and risk management best practices, particularly when new technology is widely deployed, to ensure, to the maximum extent possible, we avoid a systemic problem that ripples through the economy.

Algorithmic Bias

We all know that AI systems are only as good as their underlying training data. Leveraging AI, for example, to determine creditworthiness or underwrite loans is susceptible to biased outcomes if the underlying training data includes unknown biases. There are established laws, such as the Equal Opportunity Credit Act and the Fair Housing Act, that govern discrimination when making credit and loan decisions. Financial institutions that deploy AI for these purposes must ensure their systems are compliant and non-discriminatory or risk opening themselves to potential lawsuits and regulatory fines.

Market Stability

Any person that has invested in the stock market understands the market is subject to volatile movements when new information is introduced. AI-powered trading platforms have the potential to amplify this volatility. As these systems become more widely adopted, this potential risk increases. When a growing number of market participants make decisions based on AI-powered insights that are based on a common data set, volatile reactions to new information can be magnified. This is not a new phenomenon unique to AI – we call this groupthink or herding behavior. However, with AI systems that can make calculations and decisions in milliseconds, the speed at which groupthink behavior can move markets is greatly accelerated.

Systemic Risks

It is safe to say that no one wants to go through another economic crisis created by the financial services sector. Widespread adoption of AI technology creates new opportunities for systemic risks. As we just saw with CrowdStrike, businesses that rely on the same technology create a unique, single point of failure concentration risk. If the financial services sector develops to rely on a small set of AI technology vendors, the risk that biases, hallucinations, or unintended outcomes can rapidly disseminate through our banks and financial institutions increases. As Gary Gensler, Chair of the Securities and Exchange Commission, stated, “I would be quite surprised if in the next 10 or 20 years a financial crisis happens and there wasn’t somewhere in the mix some overreliance on one single data set or single base model somewhere.” Furthermore, the “black box” nature of certain AI decisions complicates how we deploy updated risk management and oversight best practices.

Recommendations for AI Governance in Financial Services

While we have existing regulations for the financial services industry, they may not be adequate to address the new challenges of AI. Given the systemic nature of the financial services sector, we must be diligent in implementing robust AI governance and oversight to maintain stability.

Collaborative Approach

All stakeholders, including AI developers, banks, financial institutions, and regulators, must work together to develop best practices for deploying and governing AI systems in the financial services sector. We are still in the early stages of AI adoption and governance is only now becoming a major theme for the new technology. Collaboration between industry and government will increase accountability and transparency and build public trust that there is real oversight of AI in the financial services sector.

Bias Mitigation

Bias mitigation must become a priority from the very start of AI model development and continue throughout the entire life cycle of AI systems. AI systems should be trained on diverse data sets and regularly audited to monitor and test for bias and discriminatory outcomes. We already have regulations in place that govern discrimination in the financial services industry. AI systems must be held to the same standards. To increase transparency, AI model development should also be documented in detail, including identifying data sources being used for training.

Continuous Monitoring of AI Systems

The cornerstone of effective AI governance is continuous monitoring of AI systems for accuracy, bias, and fairness. The systemic nature of the financial services sector requires banks and other financial institutions to have real-time insight into the performance of their AI systems to identify potential risks before they become unmanageable. For the financial services sector, it is not sufficient to identify problems with AI systems after they have permeated through the broader economy. It is the responsibility of all stakeholders, from banks and financial institutions to AI developers and regulators, to be proactive in monitoring and testing their AI systems to ensure, to the maximum extent possible, they are functioning as intended.

Conclusion

The benefits of AI systems in the financial services sector are manifest. We understand the potential efficiencies and innovation this new technology could foster. However, the financial services sector’s unique ability to impact the broader economy necessitates an emphasis on AI governance and oversight. Through collaboration between industry and regulators to develop best practices and an unfaltering commitment to bias mitigation and continuous testing of AI models, we can maximize the benefits of AI technology in the financial services sector while mitigating the potential risks.

Written by Chris Dougherty, Chief Financial Officer of Synergist Technology.