Today, Congresswoman Maxine Waters (D-CA), Chairwoman of the House Financial Services Committee and Congressman Bill Foster (D-IL), Chair of the Task Force on Artificial Intelligence, sent a letter to Jerome Powell, Chairman of the Federal Reserve System; Todd Harper, Chairman of the National Credit Union Administration; Rohit Chopra, Director of the Consumer Financial Protection Bureau; Jelena McWilliams, Chairman of the Federal Deposit Insurance Corporation; and Michael Hsu, Acting Comptroller of the Currency of the Office of the Comptroller of the Currency, all members of Federal Financial Institutions Examination Council, urging them to keep up with the rapid development of technology and ensure algorithmic bias does not occur.
The letter highlights the need for Artificial Intelligence (AI) technologies, including machine learning (ML), to be used in an ethical manner, and help low- and moderate-income communities of color that have been underserved for far too long. Additionally, the letter highlights the risks and possible benefits that emerging technologies pose in the financial services and housing industry.
“As you assess the use of AI from financial institutions you oversee, you must prioritize principles of transparency, enforceability, privacy, and fairness and equity,” wrote the lawmakers. “This will ensure that AI regulation and rulemaking can meaningfully address appropriate governance, risk management, and controls over AI. They will be essential when confronting challenges in the development, adoption, and management of AI. They must be relied upon when considering clarification that financial institutions will need to move forward with AI in a just and equitable manner… Financial institutions using AI must be encouraged to do more to promote racial and gender equity. Fintechs and others using these technologies should play their part in building a more just and fair financial system in the 21st century. Historically marginalized communities, including and especially low- and moderate-income communities of color, have long been subject to discrimination in the financial services and housing space, and forced into using subpar products and services. Through their use of new technologies, financial institutions using AI have the potential to play a role in offerings to communities who have been neglected in the past...”
See the full text of the letter below.
The Honorable Jerome H. Powell
Board of Governors of the Federal Reserve System
20th Street and Constitution Avenue N.W.
Washington, D.C. 20551
The Honorable Todd M. Harper
National Credit Union Administration
1775 Duke Street
Alexandria, VA 22314
The Honorable Rohit Chopra
Consumer Financial Protection Bureau
1700 G Street NW
The Honorable Jelena McWilliams
Federal Deposit Insurance Corporation
550 17th Street NW
Washington, DC 20429
The Honorable Michael Hsu
Acting Comptroller of the Currency
Office of the Comptroller of the Currency
400 7th Street, SW
Washington, DC 20219
Chairman Powell, Chairman Harper, Director Chopra, Chairman McWilliams, Acting Comptroller Hsu:
As the Chairwoman of the U.S. House Committee on Financial Services (“Committee”) and the Chairman of the Task Force on Artificial Intelligence within the Committee, we are writing to respond to your request for information on the use of artificial intelligence (AI) by financial institutions you oversee. Due to the unprecedented COVID-19 pandemic, we are at a singularly challenging time for most Americans. This is especially true for the low- and moderate-income communities and communities of color, who have been disproportionately affected by the pandemic. The use of Artificial Intelligence (AI), including Machine Learning (ML), and other emerging technologies in the financial services and housing industry both poses risks and offers possible benefits, but only if regulators are able to keep up with its rapid development and ensure algorithmic bias does not occur.
The Financial Services Committee’s Activities on AI Oversight
House rules require the Committee to continually review and study “the organization and operation of Federal agencies and entities having responsibilities for the administration and execution of laws and programs addressing subjects within its jurisdiction.” As a “formal interagency body empowered to prescribe uniform principles, standards, and report forms for the federal examination of financial institutions,” the FFIEC sits squarely at the nexus of the Committee’s jurisdiction over both financial regulators and financial institutions. As the Committee outlined in our Oversight Plan for the 117th Congress, any use of AI in the financial services industry must emphasize principles of transparency, enforceability, privacy and security, and fairness and equity, with strict scrutiny on financial institutions that exhibit algorithmic bias or engage in technological redlining. AI must be used in a way that serves the American public, its consumers, investors, and labor workforce, first and foremost.
Last Congress, as Chairwoman of the Committee, I convened the Committee’s first ever Task Force on Artificial Intelligence, led by Representative Bill Foster (D-IL), which has examined how to reduce AI bias, the impact of AI on capital markets and jobs in the financial services industry, AI usage by cloud computing providers, and other related topics. This Congress, we re-established the AI Task Force so that it can continue its investigation on whether emerging technologies such as AI are serving the needs of consumers, investors, small businesses, and the American public, which is needed especially as we seek build back better after the COVID-19 pandemic. Our first AI hearing this Congress focused on the use of AI and Machine Learning, and explored how Human-Centered AI can build equitable algorithms and address systemic racism and in Housing and Financial Services. In this hearing, our Committee learned that routine reviews of AI/ML models could work to promote inclusiveness and help overcome historic disparities experienced by protected classes. Our second AI hearing this year explored how financial institutions have increasingly relied on AI to create and authenticate digital identities of their customers, and our most recent AI hearing focused on how governments, industry, and civil society must build better AI ethical frameworks.
Racial Bias Concerns in AI and ML Technology
Historical data used as inputs for AI and ML can reveal longstanding biases, potentially creating models that discriminate against protected classes, such as race or sex, or proxies of these variables. The use of ML models on new data sources, often called alternative data, has also been shown in some instances to result in discriminatory decisions. For example, lenders using alternative data to offer private student loans or income share agreements have been accused of violating federal fair lending laws by penalizing borrowers of color who attended Historically Black Colleges or Universities, or other minority-serving institutions (MSIs).
Racial discrimination in ML housing models have also been found to be discriminatory. For example, research from UC Berkeley indicated that, with the emergence of online marketplace lending, lending discrimination had shifted from human bias to algorithmic bias, leading to Black and Latinx borrowers being charged higher rates. Technologically-enabled discrimination has led to legal action from regulators. For instance, on August 13, 2018, the U.S. Department of Housing and Urban Development (HUD) filed a complaint against Facebook alleging that its algorithms were using personal data characteristics (e.g., race, color, national origin, religion, familial status, sex, and disability) to exclude certain users from viewing housing advertisements in violation of the Fair Housing Act.
Rather than helping to take human biases out of decision-making, new types of algorithmic underwriting technologies may continue to result in, or exacerbate, disparate impacts on protected groups. Even when the data collected do not explicitly ask for or record protected characteristics, data may still act as a proxy for protected characteristics. For example, zip codes can be used in loan applications and related lending decision processes, and since the populations of residential areas have been found to be highly correlated with race and ethnicity, this variable can lead to disparate racial lending outcomes even though zip codes appear to be neutral.
Guiding Principles to AI Regulation
As you assess the use of AI from financial institutions you oversee, you must prioritize principles of transparency, enforceability, privacy, and fairness and equity. This will ensure that AI regulation and rulemaking can meaningfully address appropriate governance, risk management, and controls over AI. They will be essential when confronting challenges in the development, adoption, and management of AI. They must be relied upon when considering clarification that financial institutions will need to move forward with AI in a just and equitable manner.
Transparency and Explainability. Firstly, any use of AI must be both transparent and explainable, with human review of automated decision systems. As regulators, you must have access to look under the hood of the algorithmic machines underpinning products and services utilizing AI. Otherwise, “black boxes” with little transparency or accountability could proliferate, and algorithms that unfairly reduce opportunities, restrict services to certain demographics, conduct technological redlining, and digitally discriminate could further exacerbate existing inequalities. Therefore, we encourage you to work together to provide clear and robust guidelines, as well as issue rulemaking when appropriate, to make financial institutions disclose pertinent information on their AI modeling, data sets, and methodologies. We are very concerned about regulatory sandboxes and No-Action-Letters from regulators, as they have led to discrimination. No product or service for the American people should be allowed to be available without it being clearly explainable and tested.
Oversight and Enforceability. Secondly, regulators must use existing authority to enforce all applicable laws and regulations, conducting robust oversight of financial institutions using artificial intelligence. A dynamic “regtech” monitoring and compliance system is essential to ensuring that AI can assist, rather than harm, consumers, and investors. It must be one of your principal goals to ensure that financial institutions are strictly following all consumer, investor, and housing laws, including the Dodd–Frank Wall Street Reform and Consumer Protection Act, Fair Credit Reporting Act, Fair Debt Collection Practices Act, Equal Credit Opportunity Act, Fair Housing Act, Gramm-Leach Bliley, and other relevant statutes.
Financial institutions must fully understand that they are bound by these civil rights protections when building and manipulating datasets. Regulators must subject financial institutions using AI to comprehensive audits of their algorithmic decision-making processes, and be staffed with enough expertise as appropriate. In particular, regulators must have enough resources to monitor any use of proxies for protected classes, including race, color, religion, national origin, sex, marital status, age, or being a recipient of public assistance, even when those attributes are not considered explicitly by the AI.
As regulators, your enforcement divisions must be adequately staffed with professionals with subject-matter and technical expertise on these underlying statutes as well as AI, machine learning, and data science. Moreover, your rulemaking offices must actively provide clarity to both financial institutions and the public and strong guardrails for services and products so that industry does not unintentionally harm our financial system that is on a path to recover.
Safeguarding Consumer Privacy. Thirdly, financial institutions using AI and other novel technologies must safeguard consumer and investor data, and do their utmost to prevent cyberattacks from hackers, including foreign adversaries. Data privacy and security challenges exist across the financial services landscape, with consumer data breaches of banks, credit reporting agencies, and technology companies happening all too often. As financial institutions using AI seek to amass ever greater datasets to build their models, regulators must do their utmost to ensure that data is only being collected when necessary, and that all forms of data, including raw, aggregated, and analyzed, is being stored safely. Furthermore, these companies must safeguard any data shared with third parties providing AI services, must only share consumers’ data only with their permission, and must not share pseudonymized data that could be used to impermissibly identify individual consumers. Furthermore, these companies must safeguard any data shared with third parties providing AI services, must only share consumers’ data with their permission, and must not share pseudonymized data that could be used to impermissibly identify individual consumers.
Promoting Fairness and Equity in AI Usage. Finally, any use of AI by financial institutions must be equitable. At a bare minimum, financial institutions using AI must ensure that be extra vigilant in proactively addressing algorithmic bias, in part by ensuring that they retain diverse workforces, especially for their data scientists and computer programmers. Furthermore, any expansion of the use of alternative or additional data must be carefully scrutinized for bias or discrimination, and financial institutions should not use data that is clearly discriminatory or irrelevant, such as whether a possible consumer uses a certain social media service, to price their products.
Financial institutions using AI must be encouraged to do more to promote racial and gender equity. Fintechs and others using these technologies to play their part in building a more just and fair financial system in the 21st century. Historically marginalized communities, including and especially low- and moderate-income communities of color, have long been subject to discrimination in the financial services and housing space, and forced into using subpar products and services. Through their use of new technologies, financial institutions using AI have the potential to play a role in offerings to communities who have been neglected in the past. However, as we seek to recover from this unprecedented COVID-19 pandemic, we must ensure that abusive, predatory, and discriminatory practices do not occur. As our financial services industry transforms due to COVID-19, we have an opportunity to harness technology to build a more equitable and accessible financial system for all consumers and investors. We sincerely hope that regulators can encourage the fintech industry and companies using artificial intelligence to rise to the moment and help build a better future that is more inclusive.
Conclusion: Promoting Responsible Innovation in the usage of AI
The continuation of my Committee’s work through the Task Force on Artificial Intelligence, led by Congressman Foster, will allow Congress to be aware of all cutting-edge developments in our space. The hearings and legislation from my Committee will make sure policy can keep up with the changes to our financial services and help ensure that technology is not being used to discriminate or exacerbate existing biases under the guise of innovation. As regulators, you must do your part to robustly oversee this rapidly changing environment and ensure that strong guardrails are in place on the usage of Artificial Intelligence by financial institutions that benefit all members of the American public first and foremost.
Member of Congress
cc: The Honorable Patrick McHenry, Ranking Member
 Office of the Comptroller of the Currency, et al., A Request for Information and Comment on Financial Institutions’ Use of Artificial Intelligence, Including Machine Learning (Mar. 31, 2021).
 Mary Van Beusekom, Race, income inequality fuel COVID disparities in US counties, CIDRAP (Jan. 20, 2021).
 Rules of the House of Representatives 117th Congress, rule X, clause (2)(b)(B).
 FFIEC, About the FFIEC, (last modified 4/15/2020).
 Oversight Plan of the Committee on Financial Services for the 117th Congress, (last accessed May 7, 2021);Data & Society, Algorithmic Accountability: A Primer (Apr. 18, 2018).
 House Committee on Financial Services, Equitable Algorithms: Examining Ways to Reduce AI Bias in Financial Services, 116th Cong. (Feb. 12, 2020).
 House Committee on Financial Services, Robots on Wall Street: The Impact of AI on Capital Markets and Jobs in the Financial Services Industry, 116th Cong. (Dec. 6, 2019).
 House Committee on Financial Services, AI and the Evolution of Cloud Computing: Evaluating How Financial Data is Stored, Protected, and Maintained by Cloud Providers, 116th Cong. (Oct. 18, 2019).
 House Committee on Financial Services, Equitable Algorithms: How Human-Centered AI can Address Systemic Racism and Racial Justice in Housing and Financial Services (May 7, 2021).
 See Task Force on Artificial Intelligence Testimony of Stephen F. Hayes, Equitable Algorithms: How Human-Centered AI Can Address Systemic Racism and Racial Justice in Housing and Financial Services, 117th Cong. (May 7, 2021).
 House Committee on Financial Services, I am Who I Say I Am: Verifying Identity while Preserving Privacy in the Digital Age (July 16, 2021).
 House Committee on Financial Services, Beyond I, Robot: Ethics, Artificial Intelligence, and the Digital Age (Oct. 13, 2021).
 American Banker, Proceed with Caution on Credit Scoring with Alternative Data (Jun. 11, 2015).
 Student Borrower Protection Center, New Evidence Underscores the Fair Lending Risks Inherent to Income Share Agreements (Mar. 25, 2021).
 UC Berkeley, Mortgage algorithms perpetuate racial bias in lending, study finds (Nov. 13, 2018).
 See USA Department of Housing and Urban Development v. Facebook, (Mar. 27, 2019).
 See Robert Bartlett, Adair Morse, Richard Stanton, et al., Consumer Lending Discrimination in the Era of Fintech, University of California-Berkley (Oct. 2018).
 See Talia Gillis, False Dreams Of Algorithmic Fairness: The Case Of Credit Pricing, Harvard University (Nov. 1, 2019); see also Ninareh Mehrabi, et al., “A Survey on Bias and Fairness in Machine Learning,” (Sep. 17, 2019).
 Lu Zhang, Yongkai Wu and Xintao Wu, A Causal Framework for Discovering and Removing Direct and Indirect Discrimination, Association for the Advancement of Artificial Intelligence (2017).
 Upstart was the recipient of the CFPB’s first No-Action-Letter, leading to relaxed supervision or enforcement activity, but investigations by advocates in the space led the company to agreeing to having its underwriting model be examined by a neutral fair lending monitor discriminatory lending to borrowers who had attended Historically Black Colleges or Universities. See CFPB, Consumer Financial Protection Bureau Issues No Action Letter to Facilitate the Use of Artificial Intelligence for Pricing and Underwriting Loans (Nov. 30, 2020); see also NAACP Legal Defense and Educational Fund and Student Borrower Protection Center Announce Fair Lending Testing Agreement with Upstart Network (Dec. 1, 2020).
 Carnegie Endowment for International Peace, Timeline of Cyber Incidents Involving Financial Institutions (last visited Apr. 22, 2021).
 Federal Trade Commission, Equifax Data Breach Settlement (Jan. 2020).
 Isabella Jibilian & Katie Canales, The US is readying sanctions against Russia over the SolarWinds cyber attack. Here's a simple explanation of how the massive hack happened and why it's such a big deal, Business Insider (Apr. 15, 2021).
 Nicol Turner Lee, Paul Resnick, & Genie Barton, Algorithmic bias detection and mitigation: Best practices and policies to reduce consumer harms, Brookings Institution (May 22, 2019).
 See e.g. International Finance Corporation (IFC), Artificial Intelligence Innovation in Financial Services (Jun. 2020).