The cost of the United States’ financial industry regulatory compliance is substantial, with precise figures fluctuating annually based on changes in regulatory policies (such as found in the Dodd-Frank Act), market dynamics, and specific institutional requirements.
Federal (and state) regulators oversee financial institutions, markets and products using licensing, registration, rulemaking, supervisory, enforcement and resolution powers, not to mention the industry’s specific rules and each organization’s internal ethical standards, accountability protocols, reporting mechanisms, and documentation.
Moreover, U.S. financial institutions are vulnerable to risks, including reputation damage, loss of trust, and licensure suspension, and legal penalties that can reach $10 billion.
According to a 2006 survey sponsored by the Securities Industry Association (now the Securities Industry and Financial Markets Association), 93.9 percent of compliance costs in the U.S. financial industry are labor-related and 3.3 percent are equipment-related. Recently, Forbes reported that the cost of U.S. government regulatory compliance for financial institutions since the 2007-09 global financial crisis has skyrocketed, with major financial firms reporting an average cost of up to $10,000 per employee to maintain regulatory compliance requirements.
Financial regulatory compliance forms a foundation of trust, integrity and law that supports every monetary-related transaction in the financial system. Given the high cost of financial industry regulatory compliance, how can such compliance be improved and cost savings enhanced? The use of evolving artificial intelligence technologies in improving the efficiency, accuracy and scalability of financial institution operations is emerging as a viable industry choice.
Financial regulatory compliance is presently reliant on human oversight and manual processes, a continuing challenge with “baked-in” human error limitations.
AI technologies offer a variety of tools that can increase accuracy and precision, improve efficiency, and enhance customer experience for financial institutions and their regulatory compliance requirements. AI application tools address improved transaction monitoring, customer verification, risk assessment and mitigation, fraud detection and prevention, regulatory text analysis, robotic process automation, automated regulatory compliance, and monitoring and reporting, among other operational areas.
An example of a “Reg-Tech” business that has been incubated in Fidelity Labs — an in-house software incubator and digital studio for Fidelity Investments — is Saifr, which has recently announced it is releasing four AI models that target industry-specific, financial services regulatory compliance requirements, including one that can help suggest compliant language for content, such as marketing materials and emails.
There are several challenges to implementing AI regulatory compliance tools in the financial business.
The first challenge concerns financial data and privacy, which, if the AI system is not secured, is susceptible to data breaches and unauthorized access.
The second challenge concerns bias and fairness, whereby existing biases in the data that the AI technologies are trained on could potentially lead to unfair outcomes and discriminatory company practices.
The third challenge involves transparency and explainability, since AI systems often operate as technological “black boxes,” making it difficult to understand and explain how they arrive at specific decisions. This situation can be problematic for regulatory compliance, which often requires clear justifications for company decisions.
The fourth challenge concerns regulatory adaptability, as financial regulations change frequently. AI systems need to be continually updated to remain compliant. Failing to keep the AI updated with the latest regulatory changes can lead to regulatory non-compliance. Also, generalizing and overfitting to specific regulatory scenarios can limit the AI system’s effectiveness and adaptability.
The fifth challenge — poor AI model management — can lead to inaccurate predictions and compliance failures, requiring robust mechanisms to monitor and effectively manage the risk associated with AI models.
The sixth challenge — integrating AI technologies with existing legacy systems in financial institutions — can be complex and costly, requiring significant changes to current processes and technology infrastructure.
The seventh challenge — the quality of the input data, if inaccurate, incomplete or biased — can impair the system’s performance and lead to incorrect compliance assessments.
The eighth challenge — inadequate human oversight — is risky and, thus, important to have human experts that review and validate the AI’s decisions to ensure accuracy and compliance.
The ninth challenge involves debates and uncertainties regarding the ethical implications of using AI technologies in sensitive areas like finance.
The tenth challenge — resource intensity — requires significant investment in time, money and expertise to reap the benefits from AI technologies employment.
Addressing these regulatory financial compliance challenges will require an integrated approach involving robust system design, continuous monitoring and adaptability, effective data governance practices, and collaboration among AI technologists, regulatory experts, and public policymakers.
