Blog
Negotiating the Best Cooperation Agreement With Federal Prosecutors
The Architecture of a Cooperation Agreement
Cooperation with federal prosecutors is the most consequential decision a defendant will make, and it is almost always made under conditions that guarantee the defendant cannot evaluate it with full clarity. The government presents the agreement as an opportunity. It is, if we are being precise, a transaction in which the defendant trades information for the government’s discretion to recommend leniency, a discretion the government is under no obligation to exercise.
The federal system is constructed to reward cooperation. Under Section 5K1.1 of the United States Sentencing Guidelines, a court may impose a sentence below the guideline range, and below mandatory minimums, if the government files a motion certifying that the defendant provided substantial assistance. The word “substantial” carries the weight of the entire arrangement, and it remains undefined in any way the defendant can rely upon. What qualifies as substantial is, in practice, whatever the prosecutor determines it to be.
One must understand this asymmetry before entering any conversation with the government. The defendant cooperates. The prosecutor evaluates. The judge decides, but only after the prosecutor moves. If the prosecutor does not file the 5K1.1 motion, the court has no authority to grant the reduction. The defendant who has cooperated in good faith, who has told the truth, and who has placed personal safety at risk may receive nothing.
This is the architecture. Everything else is negotiation within it.
The Proffer Session
Before any cooperation agreement is formalized, the government will require a proffer. The proffer session, sometimes called a “queen for a day” meeting, is the mechanism by which the government assesses whether the defendant’s information is worth pursuing. The defendant sits in a room with defense counsel, an assistant United States attorney, and one or more federal agents. The defendant talks. The agents compose memoranda, usually on Form 302, which become part of the permanent record.
The proffer letter promises that the government will not use the defendant’s statements in its case in chief. That promise is narrower than it appears.
The government retains the right to make derivative use of everything disclosed. If the defendant mentions a bank account, the government may subpoena the records. If the defendant names a co-conspirator, the government may secure that person’s testimony through independent means. The proffer letter does not prevent any of this. It prevents only the introduction of the defendant’s own words at trial, and even that protection dissolves if the defendant later takes a position inconsistent with what was said during the proffer. Under Federal Rule of Evidence 410, the inconsistency need not be dramatic. A defendant who testifies at trial and describes an event differently from how it was described in the proffer room has opened the door to impeachment. The entire proffer becomes admissible.
The atmosphere of a proffer room is unlike any other setting in legal practice. The agents are not hostile. The prosecutor is cordial, occasionally warm. The atmosphere is designed to produce candor, and it does. Defendants who enter prepared to be strategic tend to relinquish the strategy within the first forty minutes. The questions are granular. They concern dates, amounts, locations, the names of people the defendant assumed the government did not know about. The government already possesses more information than it discloses, and the proffer functions in part as a credibility test: the agents compare what the defendant says against what they already know, and the comparison determines whether the cooperation proceeds.
The protection a proffer letter offers is real. It is also, in every case we have reviewed, smaller than the client believed it to be.
We prepare clients for proffers through a process that departs from the standard advice. The conventional approach is to review the facts, rehearse the timeline, and instruct the client to tell the truth. That is correct as far as it goes. We conduct what amounts to a simulated debriefing, structured to reveal the points where the client’s memory is uncertain, where the client’s understanding of events diverges from the documentary record, and where the client is most likely to minimize their own involvement (which, it should be noted, is the single most common reason cooperation agreements collapse in this district, more common than outright dishonesty, more common than the failure to provide useful information, and a pattern so consistent that we now treat it as the default risk rather than one risk among several). Federal prosecutors in the Southern District and elsewhere expect cooperating defendants to disclose the full extent of their criminal conduct, not merely the conduct under investigation. A defendant who conceals a prior act, whether from shame or from calculation, has not satisfied the terms of the proffer. The government treats selective disclosure as a failure of truthfulness, which can terminate the cooperation before it begins.
Whether the proffer session is a tool of justice or a mechanism for prosecutorial leverage is a question the system has declined to resolve.
The Terms of the Agreement
A cooperation agreement is a contract. Courts interpret it under ordinary principles of contract law, which means the language controls and the defendant’s expectations, however reasonable, do not override the text. The agreement will contain commitments from the defendant: to provide truthful and complete information, to submit to debriefings at the government’s request, to testify before a grand jury or at trial if called upon, and to refrain from any further criminal conduct. It will also contain a clause preventing the defendant from withdrawing a guilty plea if the government later rescinds the agreement. This creates a situation where a defendant has pleaded guilty and received no sentencing benefit. Practitioners encounter this outcome with some regularity.
The agreement will not specify a particular sentence. It will not guarantee that the government will file a 5K1.1 motion. In most districts, the language is drafted to preserve prosecutorial discretion at every stage. The defendant commits to concrete obligations. The government commits to consider whether those obligations have been met.
What defense counsel can negotiate, and what most counsel neglect, is the specificity of the government’s obligations within the agreement. A well-constructed cooperation agreement does not state that the government “may” file a 5K1.1 motion without further definition. It specifies, with as much precision as the government will accept, the conditions under which the motion will be filed. It identifies who determines whether the cooperation was substantial. It addresses what happens if the cooperation leads to results in a different district, and whether the defendant’s information proving less valuable than anticipated, through no fault of the defendant, constitutes a failure. These provisions are not standard. The government’s template is designed to maximize flexibility for the prosecution. The defendant’s attorney must introduce provisions that constrain that flexibility, and the negotiation over those provisions is where the agreement is either secured or conceded.
Six agreements reviewed in this office over the past year contained language that, read in its plain terms, permitted the government to decline the 5K1.1 motion for any reason or no reason. The clients in those cases had cooperated in good faith. The agreements permitted the government to acknowledge as much and still withhold the motion.
The 5K1.1 Motion and Its Gatekeepers
Only the government can file a Section 5K1.1 motion. The defense cannot file it. The court cannot order it.
Santobello v. New York established that a prosecutor’s promise in a plea agreement must be honored, but the cooperation agreement is constructed to avoid the kind of binding language Santobello would enforce; the terms are permissive where a plea agreement’s terms are mandatory. The First Circuit addressed the enforceability of cooperation-related promises in United States v. Flemmi, holding that FBI agents lack the authority to confer use immunity on a confidential informant. The case turned on a question of actual authority: who within the government has the power to bind the United States to a particular commitment. For purposes of cooperation agreements, the answer is that only the United States Attorney or a specifically authorized designee can make enforceable promises regarding sentencing recommendations. A promise from an agent, however sincere, is not a promise from the government.
The five factors the court considers when evaluating a 5K1.1 motion are the significance of the defendant’s assistance, the truthfulness of the information provided, the nature and extent of the assistance, any injury or danger to the defendant resulting from cooperation, and the timeliness of the assistance. Of these, timeliness is the factor most defendants underestimate. A defendant who cooperates upon arrest occupies a position the defendant who cooperates six months later simply cannot reach.
And even after the government files the motion, the court retains discretion over the extent of the reduction. The Sentencing Guidelines’ application notes direct the court to give substantial weight to the government’s evaluation, and most judges follow that recommendation. But the reduction is not automatic, and a sentencing hearing in which the government’s recommendation is tested by the court remains a proceeding in which defense counsel’s presentation of the cooperation, its risks, and its consequences can alter the outcome.
Timing and Sequence
In federal cooperation, timing is not a factor among several. It is the governing variable.
The first hours after arrest or the first days after receiving a target letter establish the framework for everything that follows. If multiple defendants are involved in the same conspiracy, the first to cooperate receives the most favorable treatment, not because of a formal rule, but because the first cooperator’s information retains the greatest value. By the time the second or third defendant approaches, the government may already possess what they intended to offer. Three defendants in a case we handled last year sought cooperation agreements; the first received a substantial reduction, the second a modest one, and the third, whose information was equivalent in quality, received nothing, because by then the government had what it required.
We counsel clients to make the cooperation decision before it becomes urgent. In cases where investigation is anticipated, we discuss the possibility of pre-indictment cooperation: approaching the government before charges are filed, with a proffer that demonstrates the defendant’s value as a witness. This approach carries risk. The proffer provides the government with information it may not have possessed, and if the cooperation collapses, the derivative use of that information remains available to the prosecution. In cases where indictment is a near certainty, the pre-indictment proffer can result in a non-prosecution agreement rather than a plea and cooperation agreement. Non-prosecution agreements are rare. Prosecutors in most districts reserve them for cooperators whose information is exceptionally valuable and whose own culpability is, in the government’s assessment, limited.
After the cooperation agreement is signed, the sequencing of debriefings, grand jury testimony, and trial testimony follows the government’s schedule. The defendant’s sentencing is deferred until the cooperation is complete, which may mean months or years of sustained uncertainty. During that period, the defendant remains subject to the terms of the agreement. A single instance of dishonesty, a failure to appear for a scheduled debriefing, or the commission of any criminal offense can void the agreement. The guilty plea remains on the record, stripped of whatever sentencing advantage the cooperation was meant to secure.
I am less certain about the frequency with which agreements are rescinded for minor infractions than I would like to be. The data on this point is not collected in any systematic fashion, and anecdotal accounts from defense attorneys vary by district in ways that resist generalization. What I can say is that the risk of rescission is not theoretical, and the cooperation agreement’s language on breach, in every version this office has encountered, favors the government’s interpretation.
Post-sentencing cooperation under Federal Rule of Criminal Procedure 35(b) operates on a different timeline. The government may file a Rule 35 motion within one year of sentencing, or later if the information could not have been provided within that period. Rule 35 reductions tend to be more modest. The sentence has already been imposed. The judge is adjusting it after the fact, and the adjustment reflects what it reflects.
There is a window, in every case, when the defendant’s information holds its highest value. That window opens when the government requires the information and closes when the government obtains it elsewhere. A defense attorney who recognizes that window and positions the client to cooperate within it has done the essential work. An attorney who waits for the government to initiate contact has conceded the timing advantage before the negotiation begins.
Safety and Consequence
The decision to cooperate is not only a legal calculation.
In cases involving organized criminal enterprises or drug trafficking organizations, disclosure of the cooperator’s identity creates risks that the legal system can address but not eliminate. The Sentencing Guidelines recognize this: one of the five factors in the 5K1.1 analysis is the danger or risk of injury to the defendant or the defendant’s family. Defense counsel in the target’s case may file a motion to compel disclosure of the cooperating witness’s identity (a procedural reality that defendants cooperating against violent organizations understand with a clarity that no briefing can produce). The government can recommend placement in a particular facility or, in extreme cases, enrollment in the Witness Security Program.
The risk is not identical across case types. A defendant cooperating against a white-collar co-conspirator confronts something different, in degree and in kind, from a defendant cooperating against a cartel. The agreement should address safety with specificity, though in practice the provisions tend toward the general.
For clients weighing this decision, we describe what the process looks like from the inside: the isolation that begins before sentencing and persists through it, the disruption to family that no provision in the agreement addresses, the indefinite quality of the waiting period during which the defendant has pleaded guilty but does not know what the guilty plea will cost. The phone call from a spouse asking when this will be over. The inability to answer. These are not arguments against cooperation. They are the conditions under which cooperation occurs.
What Cooperation Cannot Repair
A cooperation agreement can reduce a sentence by years, sometimes by decades, but the conviction itself persists. The defendant who cooperates still carries a felony on the record, still confronts the collateral consequences of that conviction, including employment barriers, licensing restrictions, and potential deportation for non-citizens, and still must reckon with the knowledge that their testimony was instrumental in the prosecution of someone they once knew.
The broader pattern in federal criminal practice is one of consolidation. The government gathers information from the periphery of a conspiracy and works inward, using each cooperator’s testimony to reach the next level of culpability. The cooperator is a mechanism in that process. The agreement is designed to ensure the mechanism functions. Counsel’s role is to ensure the mechanism also protects the person inside it, the person who walked into a proffer room on a Tuesday afternoon and began, sentence by sentence, to dismantle the only version of events that had kept them safe.
A first conversation with this firm costs nothing and assumes no commitment. It is the point at which the architecture of a cooperation agreement can be examined before any obligation is created, and before any statement is made that cannot be withdrawn.