Blog
What Promises Can Federal Prosecutors Actually Make?
A federal prosecutor’s word is binding only to the extent that the word appears, in writing, inside a document the court has accepted. Everything else is atmosphere.
This is the distinction that separates defendants who understand the federal system from those who learn, at sentencing, that they did not. The promises a prosecutor can make are defined by Rule 11 of the Federal Rules of Criminal Procedure, by the Sentencing Guidelines, and by a body of case law that treats plea agreements as contracts while reminding every defendant that one party to the contract (the court) never signed it.
The Categories Under Rule 11
Federal plea agreements fall into three categories, and the categories are not of equal weight.
Under Rule 11(c)(1)(A), the government agrees to dismiss certain charges or to refrain from bringing new ones. This is the most direct form of prosecutorial commitment. If the court accepts the agreement, those charges are gone. The promise, once accepted by the court, is enforceable in the traditional sense.
Under Rule 11(c)(1)(B), the government agrees to recommend a particular sentence or sentencing range, or to refrain from opposing the defendant’s request. The promise is real but thin. The prosecutor will say the words. The court is not required to listen. A defendant who pleads guilty under a (B) agreement has no right to withdraw the plea if the judge imposes a harsher sentence than the one the prosecutor recommended. Most plea agreements in federal court are (B) agreements. Most defendants do not appreciate what that means until the sentencing hearing has concluded.
Under Rule 11(c)(1)(C), the parties agree to a specific sentence or range that becomes binding if the court accepts it. This is the closest the federal system comes to a guaranteed outcome, and even here the guarantee is conditional: the judge retains discretion to accept the agreement, reject it, or defer a decision until after reviewing the presentence report. If the court rejects a (C) agreement, the defendant may withdraw the plea. In practice, some judges refuse to entertain them at all.
What a Recommendation Is
The distance between a recommendation and a guarantee is the distance most defendants misunderstand.
When a prosecutor promises to recommend a sentence of sixty months, the prosecutor has promised to stand before the court and say the words. That is the full extent of the obligation. The judge, who has reviewed the presentence report, who has heard from probation, who has weighed the Section 3553(a) factors, is free to impose a sentence of one hundred and twenty months or a sentence of thirty-six. The prosecutor’s recommendation serves as evidence of the government’s position. It is not a constraint on judicial authority.
This matters because defendants process the plea agreement as a contract for a specific outcome. They hear “sixty months” and the number becomes the architecture of their expectations: how long they will be absent from their families, when they will return to ordinary life, what they are willing to surrender for that certainty, and whether the certainty is worth the rights they are relinquishing in exchange for a figure that the agreement itself, usually on the second or third page, in a paragraph defense counsel reviewed but the defendant absorbed as procedural formality, expressly identifies as nonbinding. The agreement said so. It was never hidden. It was not read with the gravity it required.
We approach this moment in the representation differently. Before any substantive discussion of an offer, we ask the client to read the agreement alone, without counsel present, and to mark every sentence they do not understand. The marks cluster, almost without exception, around the paragraphs that disclaim the court’s obligation to follow the recommendation. Those are the paragraphs that determine whether the agreement delivers what the client believes it delivers. Starting the conversation there, rather than with the government’s concessions, changes the quality of the decision.
Cooperation and the Gate
Six months after the plea agreement is signed, the question changes. For defendants who cooperate with the government, the promise that matters is whether the prosecutor will file a motion for a downward departure under Section 5K1.1 of the Sentencing Guidelines.
Only the government can file that motion.
The defense cannot request it. The court, absent circumstances so extraordinary they occupy a vanishing fraction of the case law, will not compel it. The cooperation agreement is where the promise either exists or does not. If the agreement states that the government will file a 5K1.1 motion upon the defendant’s satisfactory cooperation, the commitment is firmer than most defendants realize, though the word “satisfactory” introduces its own ambiguity. If the agreement states that the government may file, in its sole and unfettered discretion, the defendant has acquired something closer to a possibility than a right.
The distinction is not theoretical. In the Eighth Circuit, the court in United States v. Barresse examined a cooperation agreement that omitted the discretionary language and concluded that the government’s refusal to file could constitute a breach. In the Second Circuit, agreements employing the “sole and unfettered discretion” formulation have been challenged only on the narrow ground of prosecutorial misconduct or bad faith. The same act of cooperation, under language that differs by perhaps six words, produces a different legal result.
Whether courts will continue to maintain that distinction as cooperation agreements grow more standardized across districts is a question that remains open.
But the deeper problem is what occurs before the agreement is signed, when law enforcement officers (who are not prosecutors and whose promises carry no binding authority) tell a defendant in an interview room that cooperation will “help” them. The Supreme Court’s holding in Frazier v. Cupp acknowledged that deceptive interrogation tactics, including assurances of leniency, do not render a confession involuntary. An officer can suggest outcomes that no prosecutor is obligated to deliver. And defendants encountering the federal system for the first time cannot perceive the difference between a suggestion and a commitment. The officer’s assurance that cooperation will “make things better” is not a term of any agreement. It is a statement with no binding force, offered in a room where the power differential makes careful parsing difficult.
I am less certain than the preceding paragraph implies about whether the current framework for evaluating these pre-agreement representations is adequate. The case law treats them as non-binding, and the law is clear on this point; but the clarity of the law does not eliminate the confusion it permits.
In something like three of every five cooperation matters we have handled, the client arrived at the first meeting believing the agreement contained protections it did not contain. The cooperation agreement had been signed, in some instances weeks earlier, and the client’s understanding of its terms was shaped more by the investigating agent’s characterization than by the document. The standard practice in most offices appears to involve a single review of the cooperation terms at signing. We schedule a second review after the first debriefing session, to address the provisions the client now has reason to care about. The first review is legal; the second is practical.
A promise the government can make is not the same as a promise the government has made. The distinction lives in the language of the agreement, not in the conversation that preceded it.
The 5K1.1 motion, even when filed, does not determine the sentence. It authorizes the court to depart below the guideline range. The extent of the departure is the court’s decision, guided by factors that include the significance of the assistance, the risk to the defendant, and the timeliness of cooperation. The prosecutor recommends; the court decides.
Immunity and Its Limits
Immunity in the federal system is not what the word suggests to the person receiving it.
Federal law provides use immunity under 18 U.S.C. Sections 6002 through 6004, not transactional immunity. Transactional immunity would prohibit prosecution for any conduct related to the testimony; use immunity prohibits only the direct or derivative use of the testimony itself. A witness who testifies under a grant of use immunity can still face prosecution for the same conduct if the government demonstrates that its evidence was obtained through independent sources. The Supreme Court in Kastigar v. United States established this framework and placed the burden on the government to prove independent sourcing. The protection is real but not complete.
Proffer agreements, which govern the initial meetings between a cooperating defendant and the government, offer less protection than most defendants expect. The government agrees not to use the defendant’s statements in its case in chief. It reserves the right to use the statements for impeachment if the defendant later offers contradictory testimony. It reserves the right to pursue leads derived from the statements. A proffer agreement governs only the statements made during the meeting itself, and the protections it offers do not extend as far as most defendants assume.
The word “promise” appears in this area of practice with a frequency that exceeds its accuracy. Prosecutors can promise to dismiss charges, and that promise, once the court accepts it, holds. They can promise to recommend a sentence, and that promise binds the government to the recommendation but does not bind the court to the result. They can promise to consider filing a departure motion, and that promise means they will consider it. They can promise use immunity, and that promise protects against one category of prosecutorial action while leaving others available.
When the Promise Breaks
Santobello v. New York remains the foundational case. The prosecutor agreed to make no sentencing recommendation. A different prosecutor, unaware of the commitment, recommended the maximum sentence. The Supreme Court vacated the judgment and held that when a plea rests on a prosecutorial promise, that promise must be honored.
The principle is straightforward. Its application has generated decades of litigation. The remedy for a broken promise was left to the lower courts to determine: either specific performance of the original agreement or withdrawal of the guilty plea. That binary has produced disagreement across circuits on which remedy applies under what circumstances, what constitutes a breach, and the degree of prejudice a defendant must show. The cases do not produce uniform answers.
What arrives too late for most defendants is the realization that the enforceability of a broken promise depends on the character of the promise itself. A broken (A) promise, where the government fails to dismiss agreed charges, is different in kind from a broken (B) promise, where the government fails to recommend the agreed sentence. And if the prosecutor does make the recommendation and the court disregards it, no promise has been broken at all. The system operated as designed.
There is a silence in the conference room after this is explained. Not confusion; recognition.
The Larger Architecture
The federal plea system is constructed on a principle that no one states and everyone assumes: the prosecutor’s power is significant because much of what the prosecutor offers cannot be guaranteed. The power to recommend is valuable because judges do, in the ordinary course, give weight to the government’s position. The power to file a 5K1.1 motion is valuable because most courts, most of the time, grant the departure. The system functions on expectations that are accurate on average and unreliable in the individual case.
A defendant contemplating a plea agreement is not purchasing a result. One is purchasing a probability, and the terms of the purchase are set by the government, reviewed by the court, and constrained by a sentencing framework that reserves final authority to a judge who was not party to the negotiation.
Consultation is where this conversation begins: an evaluation of what has been offered, what has been implied, and the distance between the two. A first call costs nothing and assumes nothing. It is a reading of the document, not the atmosphere.