Blog
Deferred Prosecution Agreements in Federal Cases
A deferred prosecution agreement is not an acquittal. It is the government telling a defendant, almost always a corporation, that the evidence is sufficient, the charges are prepared, and the decision to prosecute has been made, but that execution will be suspended if the defendant performs. The performance is not optional. The suspension is not permanent. And the leverage, from the moment the agreement is signed until the day it expires, belongs entirely to the prosecution.
The instrument has existed in federal practice for decades, though its prominence is recent. What began as a mechanism for diverting low-level individual offenders from the criminal docket has become the Department of Justice’s preferred resolution tool for corporate misconduct. The transformation was neither accidental nor, in the estimation of most practitioners, complete. The DPA remains a creature of prosecutorial discretion, shaped more by internal Department memoranda than by statute, governed by case law thinner than the stakes would suggest.
The Statutory Basis
The Speedy Trial Act of 1974, codified at 18 U.S.C. § 3161, provides the procedural scaffolding. Subsection (h)(2) excludes from the seventy-day trial clock any period during which prosecution is deferred pursuant to a written agreement with the defendant, with the approval of the court. The original legislative intent, as the Senate report accompanying the Act specified, was to permit deferral for individual offenders in circumstances where rehabilitation, not punishment, served the public interest. Congress has had over fifty years to narrow the language and has declined to do so, which tells us something about legislative acquiescence, if not legislative endorsement.
The Justice Manual, at Sections 9-28.000 and following, provides the operational framework for corporate DPAs. These are not statutes. They do not carry the force of law. They are internal guidelines, subject to revision by any sitting Deputy Attorney General, and they have been revised with increasing frequency.
What Gets Deferred, and What Does Not
In practice, the corporate DPA operates as follows. The government files a criminal information with the court. It then simultaneously files the deferred prosecution agreement, which holds prosecution in abeyance for a specified period, typically two to five years. The defendant admits to a set of stipulated facts, agrees to pay monetary penalties, implements compliance reforms, and in many cases submits to the oversight of an independent monitor. If the defendant satisfies every condition, the charges are dismissed. If the defendant breaches, the government possesses a signed confession of facts and a charging document already on the court’s docket.
The non-prosecution agreement, the DPA’s quieter counterpart, operates outside the court entirely. No charges are filed. No judicial approval is required. The agreement exists between the prosecutor and the defendant as a private contract, enforceable only through the threat of future prosecution. The distinction matters because a DPA carries at least the architecture of judicial involvement. An NPA does not.
For individuals, deferred prosecution in the federal system takes a different form. Pretrial diversion programs, administered under Justice Manual Section 9-22.000, permit prosecutors to redirect certain defendants toward supervision, community service, restitution, and rehabilitation in place of conviction. The individual must waive the right to a speedy trial. The individual must, in most districts, submit to a proffer interview. The individual must acknowledge responsibility, though this acknowledgment is not a guilty plea and is not, absent a breach, admissible at trial.
The categories of eligible defendants have expanded. Health care fraud, bank fraud, and false statement cases now account for a significant portion of individual DPAs in the federal system. The expansion is worth noting, though I am less certain than the policy language implies about whether it reflects a principled commitment to rehabilitation or a resource-driven need to clear federal dockets.
The Department’s Self-Disclosure Framework
On March 10, 2026, the Department of Justice announced a department-wide Corporate Enforcement and Voluntary Self-Disclosure Policy. The CEP, as the Department refers to it internally, represents the culmination of a decade-long effort to standardize the incentives offered to corporations that report their own misconduct. It supersedes every prior component-specific and U.S. Attorney’s Office-specific enforcement policy, with the sole exception of the Antitrust Division’s longstanding leniency program.
The architecture is a three-tiered framework. At the first tier, a corporation that voluntarily self-discloses misconduct, fully cooperates with the investigation, and timely remediates the underlying conduct will receive a declination. The Department will not prosecute. The corporation must still pay disgorgement and restitution, but the criminal case ends before it begins. The second tier addresses what the Department calls “near miss” disclosures: cases where the corporation attempted in good faith to self-report but either failed to satisfy the technical criteria for voluntary disclosure or presented aggravating circumstances. In those cases, the resolution takes the form of a non-prosecution agreement with a term of fewer than three years and a penalty reduction of fifty to seventy-five percent off the low end of the applicable Sentencing Guidelines fine range. The third tier covers everything else.
The shift from “presumption of declination” to the declarative statement that the Department “will” decline prosecution is not merely semantic (prosecutors retain discretion, and the definition of “aggravating circumstances” is elastic enough to absorb a great deal of prosecutorial judgment, so whether this functions as a binding commitment or a strong presumption dressed in declarative clothing is a question the next several years of enforcement will answer). The language first appeared in the Criminal Division’s May 2025 revision and has now been extended across the Department.
The CEP also creates a race. Its benefits are available only to companies that self-disclose before the government learns of the misconduct through other channels, including whistleblowers. In an environment where the Corporate Whistleblower Awards Pilot Program incentivizes individuals to report directly to the Department, the window for a qualifying voluntary disclosure is narrower than it has ever been. A company that discovers internal misconduct and deliberates for six months may find that a former employee has already made the call. The CEP does not reward the second reporter.
The practical consequence is a compression of the decision-making timeline. We have seen this compression produce its own distortions: incomplete internal investigations disclosed prematurely to preserve the option of a declination, and overcorrections where companies report conduct that, on closer examination, may not have been criminal at all. The incentive structure is clear. Whether it produces the intended outcomes is a different question.
Judicial Oversight and Its Limits
The role of the federal judge in a deferred prosecution agreement is, by design, limited. Two circuits have held that the “approval of the court” language in the Speedy Trial Act requires only that the judge determine the DPA is bona fide and not a mechanism to circumvent the trial clock. The judge does not evaluate the fairness of the agreement’s terms. The judge does not assess proportionality. The judge does not supervise implementation.
Judge Emmet Sullivan, approving a corporate DPA in 2015, observed in dicta that the use of deferred prosecution agreements for corporations departs from what Congress originally contemplated. The observation was not binding, and it has not altered the practice. But it identified a structural problem that the years since have only confirmed: the DPA functions as a sentencing device, imposing penalties and conditions and oversight on an entity that has never been convicted, and the judiciary’s role in that process is, if we are being precise, ceremonial.
Whether this arrangement serves the public interest depends on which public one is asking about. For the shareholders and employees of a corporation facing potential indictment, the DPA is a lifeline. A criminal conviction can trigger debarment, loss of licenses, and collateral consequences severe enough to end a going concern. Arthur Andersen dissolved not because it was guilty but because it was indicted. The DPA was designed, in part, to prevent that species of destruction: to impose accountability without rendering the entity insolvent and its employees unemployed.
The counterargument does not require elaboration for anyone who has followed the last two decades of corporate enforcement. A corporation that pays a fine calibrated as a fraction of annual revenue and resumes operations under a compliance plan it helped design has not, in any sense that the word ordinarily carries, been punished. The individuals who authorized the conduct often face no charges. The Justice Manual requires prosecutors to pursue provable individual charges even when the corporate case resolves through a DPA. The requirement exists because the practice of doing so was, for years, inconsistent.
The Compliance Monitor
The appointment of an independent compliance monitor, once a standard feature of corporate DPAs, has become less frequent. The trend away from monitorships accelerated after 2018. It reversed in 2021, when the Department rescinded the earlier guidance and declared that prosecutors were free to impose monitors where circumstances warranted. The oscillation tells its own story about how little consensus exists on whether monitors accomplish their stated purpose.
The monitor is paid by the corporation. The monitor reports to the government but operates within the corporation’s offices. The corporation’s counsel reviews the monitor’s findings for privilege before they reach anyone outside the building. There is a particular silence in conference rooms during the first week of a monitorship, before anyone has decided how candid to be. What the monitor discovers is less important than what the monitor is permitted to report, a distinction that sits at the center of every monitorship and is seldom discussed in the agreements that create them.
The McKinsey Resolution
In December 2024, McKinsey and Company entered into a five-year deferred prosecution agreement to resolve a federal criminal investigation into the firm’s consulting work for Purdue Pharma. The agreement required McKinsey to pay $650 million, accept responsibility for conspiring with Purdue to aid in the misbranding of prescription drugs, and submit to oversight by both the Department of Justice and the HHS Inspector General. A former senior partner, Martin Elling, separately agreed to plead guilty to obstruction of justice for destroying documents during the investigation.
The resolution was, in the government’s framing, unprecedented: the first criminal prosecution of a management consulting firm for its role in the opioid crisis. The framing is accurate as far as it goes. The consulting engagement produced strategies to increase sales of a drug that was killing people, and the firm’s internal documents confirmed that McKinsey understood what it was doing.
And yet the resolution illustrates the tension that sits at the center of the instrument. McKinsey admitted the conduct. McKinsey paid a figure that, while substantial, represents a manageable portion of the firm’s global revenue. McKinsey was barred from future work involving the marketing of controlled substances, a restriction that carries less weight when one considers that such work had already become reputationally impossible. The individual prosecution was limited to one partner on an obstruction charge for destroying evidence. The partners who designed the strategy, who composed the presentations and attended the meetings and signed the deliverables, were not charged.
Whether this constitutes justice depends on what one believes a deferred prosecution agreement is for. If the purpose is to impose consequences sufficient to deter future misconduct while preserving the entity’s capacity to operate, the McKinsey DPA achieved that purpose. If the purpose is to hold accountable every person whose decisions contributed to a public health catastrophe, it did not come close. The two purposes are not easily reconciled. Those cases tend to recur in the literature, and in conversation, longer than the ones where the instrument performed as designed.
What the Framework Assumes
The Department’s March 2026 policy has made the self-disclosure landscape more uniform than at any prior point in the history of federal enforcement. The incentives are explicit. The framework is designed to produce cooperation, and it will.
But a system built on incentivizing corporations to report their own crimes depends, at its foundation, on the accuracy and completeness of the reports it receives. The corporation controls the investigation. The corporation defines the scope. The corporation decides which emails to produce, which custodians to search, and which threads to characterize as outside the scope of the disclosure. The prosecutor receives what the corporation provides, supplemented by whatever independent investigative resources the Department can deploy. The self-disclosure framework operates, in this sense, on a form of trust that the criminal justice system is not ordinarily designed to extend.
For any corporation evaluating whether to disclose, and for any individual who has received a target letter or who perceives that their conduct may fall within the scope of a federal investigation, a preliminary assessment of the facts and the exposure is where the conversation begins. It costs nothing and assumes nothing. It establishes the frame within which every subsequent decision will be made. The Department’s policies reward speed, and the statute of limitations does not pause for deliberation.