Alejandro Mijares
Founder and Chief Executive Officer, Mijares Consulting
November 13, 2023
We are excited to present you with Part 2 of our in-depth analysis of the utilization of artificial intelligence (AI) in the financial sector, mainly focusing on bridging the insights gained from the agencies’ Request for Information (RFI) with the findings of the analysis performed by the Firm. In Part 1, we discussed the efforts of federal agencies in collecting information, seeking input on AI within financial institutions, and the structure of the request for information (RFI).
In part 2, our focus shifted to a meticulous review of the responses from 61 diverse organizations and individuals to the 17 critical questions posed in the RFI. This analysis, diligently performed by our team, reveals a nuanced portrait of the financial industry’s stance on AI. Overall, the data reflects a financial industry that is cautiously optimistic about AI’s potential but also mindful of the significant challenges it presents. Explainability and fair Lending are the most pressing concerns, indicating a focus on transparency and equity in AI applications. Meanwhile, while noted, issues like cybersecurity and dynamic updating are less prominently addressed by the organizations and individuals who participated in the RFI. We noted the following key trends:
AI Explainability (Questions 1-3): This topic received a relatively balanced set of responses, with nearly half of the organizations (48%) acknowledging challenges with AI explainability (questions most frequently answered). This percentage indicates a significant concern about understanding and explaining AI decisions, which is crucial for regulatory compliance and maintaining trust with customers. The high frequency of responses to these questions indicates a prioritization within the financial sector to understand and mitigate potential risks associated with AI.
Risks from Data Processing and Usage (Questions 4-5): Fewer organizations (36%-41%) indicated concerns with the risks arising from broader or more intensive data processing and usage. This percentage suggests that while there is awareness of the risks associated with data handling in AI systems, it may not be as pressing a concern as explainability.
Overfitting (Question 6): Overfitting, where AI models perform well on training data but poorly on unseen data, was acknowledged by 39% of the organizations. This percentage reflects a moderate level of concern about the robustness and generalizability of AI models.
Cybersecurity (Question 7): Cybersecurity concerns related to AI were noted by 31% of the organizations. This lower percentage might indicate a belief that:
- The specific risks AI poses still need to be fully understood and experienced (most likely).
- Potential need for awareness or readiness (likely).
- Existing cybersecurity measures are adequate (less likely).
The practices that manage these risks vary, highlighting the complexity and evolving nature of AI-related cybersecurity threats. The identified barriers and challenges suggest that much work still needs to be done in developing adequate security controls specific to AI. This lower percentage underscores the need for ongoing research, collaboration, and knowledge sharing among financial institutions, cybersecurity experts, and AI developers to ensure AI’s secure use and implementation in the financial sector.
Dynamic Updating (Question 8): Dynamic updating in AI models refers to the capability of these systems to adapt and modify their algorithms based on new data without manual intervention. The recognition by 36% of organizations of the challenges posed by AI models that undergo dynamic updates points to a growing awareness of the risks associated with these evolving AI systems. This acknowledgment highlights the complexities and potential pitfalls that can arise as AI models continuously adapt and refine their algorithms based on new data inputs. While this feature can significantly enhance the relevance and accuracy of AI applications, it also introduces risks and challenges, such as model drift, where the model’s performance can degrade if not adequately monitored and managed.
AI Use by Community Institutions (Question 9): Only 28% of organizations, which makes it the least-answered question, indicated challenges with AI use in community institutions, which could imply that the impact of AI is perceived to be less significant in smaller, community-focused institutions, or that these institutions are less engaged with AI. Community institutions face different challenges in AI adoption than larger financial entities. These could range from budget constraints, lack of specialized expertise, or concerns about data privacy and security, among others. Identifying the practices to overcome these challenges would be crucial for policymakers, regulators, and industry leaders to facilitate smoother AI adoption and ensure that community institutions are included in the technological revolution.
Oversight of Third Parties (Question 10): The oversight of third-party AI solutions was a concern for 39% of the organizations, highlighting the importance of managing the risks associated with external AI service providers. This concern reflects the growing reliance on external AI service providers and the complexities of ensuring that these third-party engagements align with an organization’s risk management framework, regulatory compliance, and ethical standards.
Fair Lending (Questions 11-15): Fair lending concerns related to AI varied, with the highest concern (43%) related to the risks of AI-induced discrimination. This percentage indicates a strong awareness of the need to ensure AI does not perpetuate bias or unfair practices in lending. This focus (this topic has the most number of questions in the RFI) is the result of multiple contributing factors:
Firstly, fair lending laws, such as the Equal Credit Opportunity Act (ECOA) and the Fair Housing Act (FHA), are designed to ensure that all consumers are given an equal opportunity to obtain credit without discrimination based on race, color, religion, national origin, sex, marital status, age, or because they receive public assistance.
Secondly, the complexity of compliance with these laws has increased due to the evolving nature of financial products, the advent of digital lending platforms, and the use of advanced algorithms for credit decision-making. These developments necessitate detailed inquiries to ensure that institutions understand the regulations and effectively implement systems and controls to prevent discriminatory practices.
Furthermore, the agencies’ focus on fair lending indicates a broader regulatory trend toward ensuring consumer protection and maintaining public confidence in the banking system. By addressing fair lending comprehensively, the agencies aim to promote a fair and inclusive financial environment that upholds the principles of equality and nondiscrimination.
Additional Considerations (Questions 16-17): Additional considerations regarding the uses and risks of AI were noted by 38% of organizations, while the benefits and risks to customers were acknowledged by 34%. This percentage shows a balanced view of the potential and pitfalls of AI in financial services.
In conclusion, the agencies’ RFI has been a comprehensive effort to engage with various stakeholders in the financial sector to understand the current landscape of AI utilization within financial institutions. The insights gathered underscore the need for a balanced approach to innovation that considers the potential benefits of AI, such as increased efficiency and improved customer service, against the backdrop of risks related to governance, data integrity, and the dynamic nature of AI technologies. The responses also highlight an apparent demand for more explicit regulatory guidance and support to ensure that the deployment of AI tools aligns with consumer protection laws and maintains the integrity and security of the financial system.
As the financial industry continues to evolve with the integration of AI, the information collected through this RFI will be invaluable in shaping the regulatory framework that supports safe, fair, and responsible AI adoption. It is evident that while progress has been made and Financial Institutions are learning to account for and manage these varied challenges and risks through strategic governance, risk management, compliance monitoring, and internal controls, there is a considerable journey ahead to establish robust risk-based governance including risk management frameworks, enhance the explainability of AI systems, test framework for AI models, and ensure that all financial institutions, regardless of size, can participate in the AI-driven transformation of the sector. The collaborative effort between regulatory agencies, financial institutions, and technology experts will foster an environment where innovation thrives while consumer trust and market stability are preserved.
If you find our analysis insightful, please leave a comment with your thoughts. Your engagement helps amplify the conversation around AI’s role in finance, bringing these crucial insights to a broader audience.
Stay ahead in the rapidly evolving world of finance and AI. ‘Follow’ us for continuous updates, in-depth analyses, and thought leadership content that keeps you informed and prepared for the future.
