This article in Forbes identifies potential problems based on AI and Open Banking impact on consumer data. If an institution releases AI-based solutions without addressing these and other issues then you face the risk of regulatory problems with UDAAP, 3rd Part Management Rules, or other regulations.
For example:
“As AI algorithms are built by humans, they too are exposed to our imperfect judgment, errors and implicit biases. While some AI fails are humorous — such as Facebook AI chatbots Alice and Bob, who developed their own secret language, other examples, like the varying accuracy of facial recognition technology based on race, can be far more damaging. To ensure that we are building AI algorithms, data-driven decision making tools and marketing campaigns that don’t inadvertently stereotype or decide unfairly, marketers need to acknowledge the limitations of their AI tools and implement the necessary checks and balances to maintain human accountability for machine-driven decisions.”
Sure Nikon and HP fell for this very same facial recognition problem (see Biometrics-A New Wrinkle Changes the Authentication Landscape, Page 21), but such a problem in banking would certainly not make regulators happy!
The article indicates that consumers are becoming increasingly concerned and indicates how banks can respond:
“Recent headlines about the misuse of customer data by tech companies and the increased frequency of large-scale data breaches have banking customers worried about the security and misappropriation of their information. The U.K.’s move to an open banking regime has some consumers worried about their data, too. A survey of U.K. bank account holders by Ipsos revealed that two-thirds were concerned about how their personal financial data might be used.
Amid these feelings of insecurity and skepticism, financial marketers have a captive opportunity to reassure customers about the safety of their data and proactively educate the market about how to protect their data and seek recourse should their information be misused or compromised. The steady stream of step-by-step guides on how to freeze your credit score following the 2017 Equifax breach is an example of how financial marketers can earn the trust of their institution’s customers.”
In conclusion, triple check all consumer facing Machine Learning solutions for bias. Pilot test carefully and open the throttle slowly while watching the results closely. Consider the changes that technology is driving in the market and the backdrop in which this takes place. Change is one thing, but the relentless flood of personal data into the wild by supposedly reliable financial institutions makes that change even more discomforting. Take the extra time required to communicate all the ways you are testing the technology and protecting consumer data. Consider the cost of this increased transparency part of your UDAAP compliance, because if the customer was unaware they were participating in a pilot and things go sideways, your institution will be uncomfortably exposed.
Overview by Tim Sloane, VP, Payments Innovation at Mercator Advisory Group