Introduction

Good morning. Thank you, Commissioner Goldsmith-Romero, Tony Biagioli, the Technology Advisory Committee (TAC) Designated Federal Officer, and to the members of the TAC.  It’s a pleasure to be here at your second meeting.

The agenda set out for TAC’s Subcommittee workstreams addresses critical issues that will directly influence forthcoming regulation by the Commission as well as future regulation and legislation.

Last year, President Biden announced a Blueprint for an AI (Artificial Intelligence) Bill of Rights.[1] Last month, during a speech in San Francisco, President Biden delivered remarks on the risks and opportunities posed by artificial intelligence.[2] The President echoed reflections and concerns that will shape a values-driven discourse on the integration of AI.  Congress is also casting a spotlight on AI.  Earlier this year, Senate Majority Leader Chuck Schumer from my home state of many years—New York—launched a first-of-its-kind high-level framework and began to outline a new regulatory regime for AI, engaging leading experts to help inform the proposal.[3] For those of us who have spent years thinking, researching, and writing about the potential and concerns surrounding the integration of AI, this is a welcomed approach.

In 2008, I began to research and publish literature examining the integration of AI in financial markets.  Not long after, I began to support federal agencies, including our Commission and the Securities and Exchange Commission (SEC), in the development of regulations addressing concerns emerging from the adoption of technologies that accelerate trading.

Three quick observations:

1. AI technologies tout tremendous promise – a promise to reduce frictions and enhance efficiency, permitting trading at the speed of light; yet, there are many reasons to carefully interrogate these promises to ensure that we mitigate exposure to potential perils that may also follow from adopting AI;
2. We must ensure that responsible AI effectively ensures accountability; accountability requires transparency and visibility; and
3. The integration of AI by our largest and most complex financial institutions, including exchanges, clearinghouses, and those who provide services supporting execution, clearing, and settlement (of, for example, sophisticated derivatives contracts), as well as transactions offering the least complex financial services (consumer payment transfer services or residential mortgages) to the most vulnerable consumers must be subject to sufficiently rigorous evaluation (whether auditing or alternative approaches) and regulations.

Racing at Light-Speed

Even when technologies offer great promise, we must ensure compliance with existing guardrails including the ability to effectively police AI.  In a paper that I published from discussion inspired by a symposium panel with SEC Commissioner Hester Pierce, I explained that “federal statutes [and regulation] regulate risk-taking by financial market intermediaries including the broker-dealers who execute trades and the securities exchange and clearinghouse platforms where trading occurs.  For almost a century, these statutes have enforced norms that encourage disclosure, transparency, and fairness.  In modern markets, innovation, and technology challenge these core principles of regulation.  The engineering of computer-driven automated trade execution, the development of algorithmic trading, and the introduction of high frequency trading strategies accompany a number of important shifts in financial market intermediation.”[4] That academic project traced the evolution of the adoption of computer-based trading and other technologies from the Paperwork Crisis of 1967, the Stock Market Crash of 1987, the Financial Crisis of 2007, and the Flash Crash in 2010.

Coupled with the accelerated pace of algorithmically-driven trading on digital platforms and a global, almost instantaneous, internet-based infrastructure, I raised alarms regarding the need for guardrails that effectively address any integration of high frequency trading, permitting co-location, evolving enforcement challenges and other concerns that arise from nuanced and sometimes impermissible trading strategies such as front-running and spoofing.

Focusing on the Flash Crash: Navigating Innovation in Enforcement Actions

Focusing on the Flash Crash, recall the challenges of disentangling the use of technology for purposes of establishing elements in enforcement cases.  The SEC and CFTC enforcement divisions initially concluded that an automated algorithm deployed trade orders for a single institutional investor (Waddell & Reed), rapidly executing the sale of 75,000 E-Mini S&P 500 future contracts (valued at approximately $4.1 billion) and triggering the ephemeral crash.[5] Several years later, in 2015, the Department of Justice and the CFTC investigations revealed that a rogue London-based futures trader—Navinder Singh Sarao—had manipulated the E-Mini S&P 500 by using an algorithm to flood the Chicago Mercantile Exchange (CME) with sell orders for E-Mini S&P 500 contracts.[6]

Using a high-frequency trading strategy known as “spoofing,” Sarao entered tens of millions of dollars of orders intended to drive down the price of certain futures contracts.[7] After submitting sell orders, he entered orders to buy the same contracts at artificially depressed prices.  Contemporaneously, he cancelled the original sell orders that drove the prices downward before any such orders closed.  In an effort to manipulate the market, he submitted orders intending to withdraw the same orders before an exchange or clearinghouse closed the trade.  Sarao never intended to sell, but his sell orders influenced trading across international financial markets.  Before and after the Flash Crash, Sarao generated $50 million in profits.  In November of 2016, Sarao pled guilty to one count of wire fraud and one count of “spoofing.”

This simplified description does not reach the complexities of the relevant technologies nor address the impact of the incorporation of these technologies in markets that are facing unique and evolving market structure.  The rise of alternative trading systems, private pools of capital, and the panoply of digital asset trading service providers amplifies the challenges for the development of regulation and policy as well as successful enforcement actions.

Understanding the Ethical Implications of AI

I applaud the TAC subcommittee focusing on the implications of AI.  As an Associate Dean and endowed professor at Tulane University, I convened computer scientists, programmers, software engineers, market participants, lawyers, public interest advocates, and academics for a day-long symposium examining the Implications of Artificial Intelligence in a Just Society.  In an essay published in the Journal of International and Comparative Law, my co-author and I explained that “[e]merging technologies promise to play a transformative role in our society, enabling driverless cars, enhanced accuracy and efficiency in disease mapping, greater and less expensive access to certain consumer services, including consumer financial services.  Discussions regarding the role of emerging technologies increasingly center on the development and integration of artificial intelligence technologies or AI-an assemblage of technologies that rely on a variety of computational techniques.”[8]

Beyond the integration of AI in our derivatives markets, we must note the consequences of relying on black box algorithms in other areas of our markets.  While AI may appear neutral in its analysis, for example, in assessing the creditworthiness of residential loan applicants, evidence suggests that proxies may lead to undetected bias. Four years ago almost this week, I testified before the United States House Financial Services Committee as an expert on this very issue.[9] 

Responsible AI in SupTech

A few weeks ago, at the Annual Meeting of the International Organization of Securities Commissions (IOSCO), I joined a panel of experts discussing the adoption of supervisory technology (SupTech).[10] As reported in our CFTC Agency Financial Report,[11]  the Commission’s Division of Data is working to transform our analytics toolkit to leverage new cloud architecture with advancements in AI, machine learning (ML), natural language processing (NLP), and data analytics.  At the core of these efforts is the need for accurate, efficient, and consistent data reporting.  Using NLP will allow the CFTC to convert regulatory reports that come in different formats into structured data for critical analysis.  The data pattern recognition capabilities of ML will enhance enforcement activities through identifying spoofing and other types of market manipulation in ways that were not previously possible.  Through cloud data repositories the Commission can expand or contract our usage based on need, allowing us to more efficiently allocate funding and other resources.  These efforts also allow us to leverage our existing staff expertise and in-house analytics software with automated systems to enhance our surveillance to better achieve our goals of market integrity and transparency.

I have also supported the Administrative Conference of the United States (ACUS) as an academic consultant in the development of reports coordinated across a significant number and diversity of federal regulatory agencies examining the integration of AI.[12]  In the reports, ACUS details the potential and perils of administrative agencies subject to constitutional protections integrating artificial intelligence.  SupTech is a possibility, but not without careful oversight to ensure the integrity of the technologies that we adopt as well as preservations of the rights that citizens have long enjoyed under our Constitution.

I am looking forward to hearing from all of today’s panelists.  I am hopeful that there will be thoughtful dialogue about concerns related to enforcement actions, the potential for bias and responses to these concerns, as well as the privacy implications of integrating AI.  

Digital Assets 

Market conditions and recent developments demonstrate the necessity of continuing to vigilantly enforce our regulations. Last week, the CFTC, SEC, the Federal Trade Commission, and the Department of Justice introduced important litigation against Celsius Networks and its Chief Executive Officer, Alex Mashinsky.[13] Earlier this spring, the CFTC and SEC introduced actions against Binance.[14] In each instance, the CFTC seeks to enforce existing regulations.  Notably, however, our regulation is incomplete.  I am heartened that Congress is actively considering proposed legislation.  I am hopeful to learn how TAC seeks to influence CFTC Commission regulations now and as a result of pending legislation, particularly as relates to decentralized finance (DeFi).

Cyber Resilience 

Finally, cybersecurity and cyber resilience should be top of mind for many of us at the Commission.  For a decade, I have also presented as an expert and published on these issues for over a decade as well.  The systemic nature of this problem, as well as the increasing centrality of technology to our markets and our economy, is such that it is incumbent upon us to explore multiple approaches to solving it—some may focus on governance and regulatory policy, while others look to identify vulnerabilities in software and hardware.  I look forward to hearing TAC member perspectives on this topic today.

Conclusion

The TAC is a critical forum for discussion of these critical issues, and I am hopeful that it will generate many recommendations for the Commission and areas for further study.  Thank you all for your participation.