The Paper Trail is Dead

Introduction White collar investigations have traditionally begun with people. A whistleblower provides a document to prosecutors; an accountant uncovers an irregular transaction; a regulatory agency receives a complaint; and investigators pursue the evidence until they can reach a conclusion about whether a crime has occurred.

That paradigm may be changing. The federal government is increasingly using data analytics to detect potentially fraudulent financial and healthcare activity before traditional investigations begin.

In April 2026, the Justice Department launched its Fraud Oversight through Careful Use of Statistics (FOCUS) initiative to collaborate with "data miners" who analyze government data for the purpose of detecting fraud, and in August, the Department created the National Fraud Detection Center, a prosecutor-driven, multi-agency initiative designed to use analytical tools to identify criminal leads in federal programs. ¹ The shift raises a difficult issue for modern white-collar defense: when an algorithm detects a suspicious pattern, how much weight should that pattern carry as evidence of a crime?

From Whistleblower to Algorithm

The False Claims Act has long provided a private cause of action against federal contractors who commit fraud against the United States. But the government has an additional weapon in its arsenal: the information it possesses about companies and individuals seeking to do business with it.

Healthcare programs generate reams of information about billing, prescriptions, diagnoses and payments. Federal procurement creates a paper trail of contracts, invoices and expenditures. Financial markets generate data about transactions and trades. Patterns can emerge from those individual data points.

The Justice Department's Fraud Division is increasingly using data-driven investigative techniques as a key element of its enforcement strategy, and the Department's National Fraud Enforcement Division describes data analytics as a means for identifying fraud that would otherwise go undiscovered. ²

That approach can fundamentally alter an investigation. If investigators begin to ask not "what happened here," but "why is this different from everyone else," the implications can be profound.

Suspicion Is Not Guilt

A critical element of white-collar crimes is often mens rea. A corporation may make an accounting error, a doctor may bill more than his peers due to the patients he treats, a trader may make an investment based on lawful information, and a contractor may have a higher error rate due to administrative issues rather than fraudulent intent.

Data can determine that a given entity is an outlier. It cannot explain why that entity is an outlier. That is where traditional investigation becomes indispensable. An algorithm can tell prosecutors which physician bills Medicare at an extraordinary rate, but it cannot assess whether the physician acted with intent, misunderstood billing rules or treated a unique patient population.

The Mens Rea Problem

This problem takes on added importance in criminal cases, since many white-collar offenses are defined by elements of mens rea . Fraud, for example, often requires proof of a culpable mental state, which goes beyond showing that a defendant made a false statement or engaged in a suspicious transaction; it requires showing that the defendant possessed the mental state contemplated by the particular statute.

Data analytics can be tremendously powerful in identifying suspects, but the more the government moves toward automated detection and away from pattern analysis, the more the government risks conflate correlation with mens rea .

Imagine a system that flags a company as suspicious because its transactions differ from those of its industry peers. That information may support an investigation, but it is far from conclusive evidence of criminal conduct. A statistical anomaly is a reason to investigate; it is not a reason to convict.

The Government Is Not the Only Data Miner

The evolution also affects corporate compliance programs, as companies increasingly use their own data to identify potentially unlawful payments, purchases or other transactions.

The impetus is understandable: federal prosecutors take compliance programs into account when determining whether and how to resolve corporate criminal conduct. ³ This creates a feedback loop in which the government analyzes data to identify potentially unlawful corporate conduct, and companies analyze data to identify potentially unlawful employee conduct before the government does.

This system may well make white-collar enforcement more efficient, but it also raises a difficult compliance issue: at what point does monitoring become excessive?

A corporation cannot investigate every statistical anomaly, since the process can create vast quantities of false positives, consume enormous resources and create psychological effects that cause employees to view routine transactions as suspect.

Compliance monitoring that is too aggressive can become less of a help and more of a hindrance. A sophisticated monitoring program therefore requires something beyond sophisticated technology: it requires judgment.

The Securities Market Already Works This Way

The securities industry provides an instructive example. The SEC's Market Abuse Unit uses the Consolidated Audit Trail, or CAT, to detect trading irregularities, and in September 2026, the Commission announced a judgment against an individual whose suspicious trading activity had been identified with the aid of CAT data. ⁴

The importance of that development goes beyond that particular case. The rise of the digital economy has created an enormous paper trail of who bought what, when and where.

Regulators can therefore detect relationships and patterns that would have been impossible to identify before. That creates tremendous opportunities for enforcement, but it also creates a fundamental shift in the relationship between regulators and the regulated.

A securities investigator no longer needs to find a pattern by chance; a machine can identify the pattern, leaving the lawyers and investigators to confirm that the pattern is unlawful.

The Due Process Question

This process raises a final issue relating to due process: how much should a defendant know about the government's process for reaching a particular conclusion?

Traditional evidence can typically be scrutinized. A document contains words; a transaction contains records; a witness can be subjected to questioning.

An algorithm is more complex. Its conclusion may turn on a particular statistical model that contains thousands of variables or inputs, which may be difficult for an outsider to understand, which may contain trade secrets, and which may contain other information that makes disclosure undesirable.

Those are not insuperable problems, but the government cannot expect to rely on an algorithm to bring criminal charges and then rely on secrecy to defend those charges.

At the end of the day, a defendant facing significant criminal exposure still needs to know enough to be able to defend himself. The government may have used an algorithm to initiate an investigation, but the government cannot use an algorithm as an unbeatable expert witness.

Why It Matters

The future of white collar enforcement may involve a fundamental change in how investigations are conducted.

As the federal government begins to rely more heavily on databases than on subpoenas, it becomes that much easier for the government to identify fraud, since fraud that has traditionally been hidden in a sea of transactions may no longer be difficult to isolate.

That creates an obvious benefit for law enforcement, but it also creates a responsibility: to ensure that it can distinguish between what is unusual and what is unlawful, between what is merely suspicious and what is manifestly fraudulent.

Conclusion

The future of white collar enforcement will not diminish the importance of traditional investigators.

The database may detect the physician who bills Medicare at an unusual rate, the company that makes improbable payments or the trader who engages in suspicious transactions, but it cannot answer the question that has always been at the heart of the criminal law: what did the defendant do, and did the defendant do it knowingly and unlawfully?

Data can detect a pattern; evidence still has to prove a crime.

1. U.S. Department of Justice, Civil Division, “Civil Division Announces FOCUS Initiative for Data Miners Filing Qui Tam Complaints” (Apr. 30, 2026); U.S. Department of Justice, "Department of Justice Announces Launch of National Fraud Detection Center to Combat Fraud Against Taxpayer-Funded Programs" (Aug. 24, 2026).

2. U.S. Department of Justice, National Fraud Enforcement Division, describing its mission as using "advanced data-driven investigative techniques" to identify and investigate fraud schemes.

3. U.S. Department of Justice, Criminal Division, Corporate Enforcement and Compliance Unit, describing its role in evaluating corporate compliance programs and overseeing data-analytics initiatives.

4. U.S. Securities and Exchange Commission, "Justin Chen," Litigation Release No. 26634 (Sept. 8, 2026).