To view a PDF of the full article, please click here.

Artificial intelligence (AI) is a key emerging technology that is poised to see vastly expanded use in many areas in which white-collar criminal practitioners work. AI currently is playing a growing role in helping white-collar lawyers and their clients analyze vast amounts of data to uncover insights, connections, and patterns that would be impossible to detect through manual reviews. As AI begins playing a more important role in compliance, fraud detection, and governmental investigations, regulators in the United States and around the world are adopting rules for how AI is implemented. Courts also are weighing in on the use of AI in litigation where it is used to analyze large amounts of electronically stored information (ESI). This article provides an introduction to AI technology and discusses the key regulatory developments practitioners should be aware of as they advise their clients on AI.

Background on AI

AI refers to computer processes that can mimic human cognitive skills, and computer scientists typically divide AI into Weak AI and Strong AI. Weak AI, also known as Narrow AI, is the type of AI used today to describe programs and algorithms that are designed to perform specific tasks such as playing chess, driving a car, and recognizing patterns in large data sets. Strong AI, also known as Artificial General Intelligence, is a theorized form of AI whereby a computer would have intelligence that is comparable to the human mind and includes generalized abilities to reason and analyze information.

White-collar practitioners should understand that there is a subfield within AI known as machine learning. Machine learning is a process by which computers use algorithms and training datasets to learn in a way that is similar to how humans learn, and it uses a small sample dataset to train a machine learning algorithm. Over time, a machine learning algorithm allows a computer to evaluate how accurately it is performing a specific task and takes corrective measures to eliminate errors in order to get better and better at the task. Once an algorithm is trained on the sample dataset, it can be deployed against much larger datasets to seek out patterns and other information that would otherwise remain hidden.

Early forms of machine learning required data to be organized in a structured database with labels for the various pieces of data. Today data scientists have created deep machine learning, which is a more advanced form of machine learning whereby a computer can analyze and learn from a wide variety of structured and unstructured data. White-collar practitioners can be exposed to deep machine learning in programs that can analyze vast quantities of evidence and data in numerous different formats. Compliance personnel at financial institutions can also deploy deep learning to uncover patterns and identify useful information from large datasets, such as the ultimate beneficial owners of bank accounts, unusual trading activity in the stock market, and unusual flows of funds and payments buried deep in a company’s accounting records that might run afoul of anti-corruption laws. For example, AI has become invaluable in many white-collar investigations related to the Foreign Corrupt Practices Act. White-collar lawyers can work with AI vendors to analyze terabytes of corporate data, including accounting records, vendor records, and emails to uncover falsified invoices through anomalies in invoice numbers and suspicious payment codes and amounts.

Regulators Respond to the Growing Use of AI

Given the potential AI has in proactively identifying financial fraud, regulators in the United States and around the world have begun to promulgate rules and regulations concerning the use of AI. For example the Monetary Authority of Singapore has promulgated Principles To Promote Fairness, Ethics, Accountability and Transparency (FEAT) in the Use of Artificial Intelligence and Data Analytics in Singapore’s Financial Sector; the Hong King Monetary Authority has promulgated High-level Principles on Artificial Intelligence; and the Central Bank of the Netherlands has promulgated General Principles for the use of Artificial Intelligence in the Financial Sector.

In general, these regulatory pronouncements set standards and expectations concerning accountability from senior management for the use of AI in their companies, requirements to test and validate the AI models that are in use, and ethics (including non-discrimination) and transparency. White-collar practitioners will find these regulatory guides very useful as they advise their clients on how to implement AI in their companies.

U.S. financial regulators are also focusing increased attention on AI and new regulations will likely be forthcoming. In March 2021, the Board of Governors of the Federal Reserve System, the Consumer Financial Protection Bureau, the Federal Deposit Insurance Corporation, the National Credit Union Administration, and the Office of the Comptroller of the Currency jointly issued a Request for Information and Comment on Financial Institutions’ Use of Artificial Intelligence, Including Machine Learning, 86 Fed. Reg. at 16837 (March 31, 2020) (AI RFI). The AI RFI included a total of 17 questions seeking information about AI and corporate governance, risk management, and quality controls. By reviewing the AI RFI, white-collar practitioners can get a head start on understanding the issues in AI that are likely to be the subject of forthcoming regulations.

The SEC is also moving forward with its own review of the use of AI in the asset management industry. SEC Chairman Gary Gensler recently announced that the SEC was conducting an early-stage review of how investment advisors are using AI and data analytics for marketing and for offering financial products to customers. Chairman Gensler’s remarks suggest that the SEC is concerned about whether AI was being used in the best interest of advisors’ customers or for other purposes, such as driving up revenue based on unnecessary product recommendations. The SEC has also been transparent about its own use of AI and machine learning to identify emerging market risks and potential misconduct.

The Criminal Division of the U.S. Department of Justice (DOJ) has also been following the growing use of AI in businesses. On June 1, 2020, the DOJ promulgated an update to their Evaluation of Corporate Compliance Programs, which placed a strong emphasis on the DOJ’s heightened expectations for companies to use advanced data analytics and testing as part of an effective compliance program. This is an important update because the DOJ’s conclusions concerning the effectiveness of a company’s compliance program is a major factor in charging decisions as well as in the DOJ’s determination of what it believes to be an appropriate amount of a criminal fine for an organization.

Courts Weigh in on AI and Data Analysis

Due to the vast quantities of ESI that are often involved with complex litigation, courts have been considering the appropriate use of AI and data analytics for years. Da Silva Moore v. Publicis Groupe, 287 F.R.D. 182 (S.D.N.Y. 2012) is a landmark case because it was the first time a court accepted the use of predictive coding generated by a computer to satisfy a party’s discovery obligations concerning the analysis and production of ESI. Since the time the Da Silva case was decided, more and more courts have shown an openness to using technology-assisted review tools to satisfy parties’ discovery obligations.

Courts have also had the opportunity to weigh in on the use of data analytics in the context of claims under the False Claims Act (FCA). In United States ex rel. Integra Med Analytics v. Baylor Scott & White Health, 816 F. App’x 892 (5th Cir. 2020), the plaintiff brought an FCA claim based on its in-depth analysis of publicly available data that it claimed demonstrated the defendant was overbilling Medicare. This was an unusual case because FCA claims are generally brought by insiders who have first-hand knowledge of improper conduct, not by third parties who perform data analytics on publicly available data looking for patterns of Medicare fraud. The Fifth Circuit rejected the plaintiff’s reliance on statistical data to allege fraud. The court held that data analytics do not satisfy a plaintiff’s pleading requirements if the data is also consistent with a lawful alternative explanation. The same plaintiff brought a similar claim under the FCA relying on statistical data analysis that is currently the subject of a pending motion to dismiss in the Southern District of New York. See United States of America, ex rel. Integra Med Analytics v. Isaac Laufer, 17-CV-9424 (S.D.N.Y.). The decision on the motion to dismiss will indicate whether the Southern District of New York will follow the Fifth Circuit’s lead in holding that statistical analysis itself is not sufficient to satisfy a plaintiff’s pleading requirements.

Conclusion

The emerging technology of AI is in its infancy, but regulators and courts are now beginning to write the rules that will govern the use and oversight of AI in a number of important contexts. By becoming familiar with the developing regulation of AI, white-collar practitioners will be in a strong position to advise their clients of the legal issues that arise from the growing use of AI.

Robert G. Heim is partner and co-chair of the white-collar and regulatory enforcement practice at Tarter Krinsky & Drogin.

---

Reprinted with permission from the June 24, 2022 edition of the NEW YORK LAW JOURNAL © 2022 ALM Global Properties, LLC. All rights reserved. Further duplication without permission is prohibited. For information, contact 877-256-2472 or reprints@alm.com. # NYLJ-6282022-552310