Avoiding Risk and Finding Reward in Utilizing Artificial Intelligence in Financial Services
Financial institutions can gain plenty by utilizing artificial intelligence, but those same tools can be used by external parties to generate cybersecurity hazards. Alex Koskey and Matt White of Baker Donelson explain how AI is improving the financial services industry while also creating new risks and considerations.
Artificial intelligence is a modernday double-edged sword. For all its benefits in risk management, fraud prevention and credit decisioning, it creates plenty of its own risks for financial intuitions, especially when used by external parties. Alex Koskey and Matt White, attorneys in Baker Donelson’s data privacy and cybersecurity practice, discuss how AI can help financing companies operate more effectively while providing guidance on how to protect against threats created through tools such as deepfakes and AI-generated phishing attacks.
How can AI help financial services companies improve operations?
Artificial intelligence is increasingly being utilized in a variety of ways in the financial services industry. For example, on the customer-facing side, AI utilization in credit decisions can provide a faster and more accurate assessment of a potential borrower’s creditworthiness. It’s also used frequently to assist in trading decisions (i.e., algorithmic trading). An AI’s ability to quickly and accurately predict optimal investment opportunities based on real time reviews of market conditions and stock performance explains the dramatic increase in the deployment and use of these tools over the past few years.
There are also many uses for AI that allow financial institutions to better handle back of the house operations, including risk management and fraud prevention. There are a variety of other operational processes in which the financial services industry may begin to utilize AI’s capabilities. The ability to process enormous amounts of data, along with its ability to contextualize and digest that data, will likely be an invaluable valuable tool to augment human efforts to come up with solutions to vexing problems and time intensive requirements.
How can AI improve fraud detection and risk management?
Due to its enormous processing power, AI can analyze real-time conditions across various markets and environments to provide accurate identification and prediction of risks. Often, these risks are identified significantly more quickly and earlier in time than any human could do. Further increasing AI’s effectiveness, both structured and unstructured data can be analyzed, which cannot be efficiently done by human reviewers.
Similarly, AI is increasingly being deployed in the context of fraud prevention. For several years, financial institutions have utilized AI to combat credit card fraud. An AI’s ability to quickly analyze numerous variables, including spending habits, location and other behaviors, has assisted financial companies in quickly identifying fraudulent transactions. Some financial institutions are also deploying AI to combat money laundering. Because AI can analyze complex transactions to identify variables and patterns that can be updated in real time, it has proven effective in this context as well. Other examples of uses include customer service focused “AI chatbots” and targeted marketing/ promotions/rewards based on users’ purchase histories, and credit scores and utilization.
What has driven the recent increase in the use of these kinds of tools in the last few years?
While language-based AI applications have drawn much of the attention recently, AI tools have been being developed for years. As technology progresses, processing power increases and the cost of implementation decreases, more companies have jumped on the bandwagon to utilize AI. AI’s ability to quickly and accurately analyze and contextualize huge volumes of data, as well as offerings such as machine learning, natural language processing and text/data mining, are providing companies with the opportunity to more efficiently operate and offer new and exciting services.
What are some of the leading AI implementations within the financial services industry? What are some advancements you expect to be made and incorporated in the near future?
One significant example of an opportunity AI presents for financial institutions is to increase the effectiveness of current KYC/AML programs that are, at best, having only a limited impact on preventing money laundering. In addition to allowing for more efficient operations, AI’s ability to quickly and accurately analyze and contextualize huge volumes of data can assist financial institutions in optimizing effective controls for customer due diligence, screening and transaction monitoring. However, such implementations would require not only the financial investment in AI tools, but also the updating of current risk management frameworks. Additional thought must be given to how financial institutions can protect the data privacy rights of their customers as well.
If a company is considering implementing some form of AI technology, what do they need to do to prepare themselves, whether that be updating risk management frameworks or back-office data or something else?
Obviously, considerations for preparing to use AI-based offerings are going to differ dramatically based on the specific implementation. However, one common thing that financial institutions need to monitor is regulatory developments and commentary concerning AI. Most of the major financial institution regulators have issued guidance on various aspects of AI, including regulatory obligations and risks. Financial institutions need to monitor these pronouncements to ensure compliant implementations of new products, services and/or tools.
What are some of the risks of utilizing AI in a financial services/ lending situation and how can lenders protect themselves?
An important risk associated with AI that cannot be overlooked is its impact on cybersecurity. While AI presents a host of new tools and abilities to combat cyber fraud, bad actors are also increasingly using AI to launch better, more sophisticated cyberattacks. Some of the positive implementations of AI in this context include using advanced learning techniques to analyze volumes of potentially malicious data. Similarly, AI can detect and identify vulnerable patterns within company networks and can monitor for them in real time. Moreover, with AI, many threat detection activities can be automated without any human intervention required, thus increasing both effectiveness and efficiency of threat mitigation and detection efforts. However, for all the positives, threat actors are also leveraging the unique abilities of AI. For example, studies have shown that phishing attacks crafted by AI are more successful than those crafted by humans. Similarly, an AI can generate new malware or craft previously unknown malicious code that can be easily deployed by threat actors. We are also seeing more attacks utilizing deepfakes and language-based AI tools preparing spoofed voicemails. This is creating significant risks for companies and essentially an arms race for cyber warfare superiority.
Do you think AI is the future of financial services, especially when it comes to credit decisioning and operations? Why or why not?
I suspect over the next few years we will see financial institutions adopt wide-spread usage of AI tools. As these tools become increasingly available and cost-efficient, we will see financial institutions adopt them for fraud monitoring, security, and to more efficiently process data driven efforts.
We will also see increased usage of AI for credit decisions. AI models can quickly incorporate internal data, information from credit bureaus, credit attributes, demographics, spending habits and more to give financial institutions a more accurate view of an individual’s creditworthiness. Importantly, while it is theoretically possible for human efforts to produce similar results, such an endeavor would be incredibly time intensive, could not adapt and react to real-time changes, and would necessarily be driven by fewer data sources, leading to less effective credit decisions. AI’s ability to provide this information more quickly, based on more data, and in many instances automatically, can lead to not only better and more accurate credit decisions, but also credit decisions for individuals who under traditional models don’t qualify for a credit score. As a result, we’ll likely see increased utilization of AI in credit decision-making to both provide better and more accurate credit decisions, but also to expand the ability to provide credit to segments of the population that currently struggle to access credit.