Breached Plea Agreements as Grounds to Withdraw a Federal Guilty Plea

A Plea Agreement Is a Bargain the Government Must Keep

Federal plea bargaining rests on an enforceable exchange. In Santobello v. New York, the Supreme Court held that when a guilty plea rests in any significant degree on a promise or agreement of the prosecutor, so that it is part of the inducement or consideration, that promise must be fulfilled.

The logic is straightforward. A defendant who pleads guilty surrenders the right to a jury trial, the right to confront witnesses, and the privilege against self-incrimination — and gets the government’s promises in return. If those promises go unkept, the foundation of the exchange collapses. The Supreme Court has since described plea bargains as essentially contracts: when one side’s promise is not kept, the contract is broken, and the injured party is entitled to a remedy.

Two features make this doctrine unusually protective. First, good faith is irrelevant. In Santobello itself, the prosecutor who appeared at sentencing apparently never knew that a colleague had promised to make no sentence recommendation; the Court held that the breach’s inadvertence did not lessen its impact, because a prosecutor’s office is treated as a single entity charged with internal communication. Second, the agreement is construed strictly against the government as its drafter. In the Fifth Circuit, ambiguity in a plea agreement cuts in the defendant’s favor, a rule drawn from ordinary contract interpretation and applied with extra force because the defendant has staked his liberty on the deal.

What Counts as a Breach — and What Does Not

The Fifth Circuit asks one controlling question: was the government’s conduct consistent with the defendant’s reasonable understanding of the agreement? In United States v. Purser, the court reviewed that question de novo, construed the agreement strictly against the government, and required the defendant to prove the facts establishing a breach by a preponderance of the evidence.

The doctrine polices substance, not labels. A prosecutor who technically recites the agreed recommendation while urging the court toward a harsher result may still breach, because the test is the defendant’s reasonable understanding of the bargain, not the literal minimum of the words used. At the same time, courts will not add terms the parties never agreed to. In United States v. Benchimol, the Supreme Court held that a government agreement to recommend a sentence obligates it to make that recommendation, but does not carry implied duties to explain its reasons or to advocate enthusiastically. Rule 11 speaks in terms of what the parties in fact agree to.

Because the analysis turns on reasonable understanding, the written agreement’s exact words control the outcome far more often than courtroom argument does. Integration clauses, reservation-of-rights language, and carve-outs for new information all shape what a defendant can reasonably claim he was promised.

Remedies: Specific Performance or Plea Withdrawal

Santobello recognizes two remedies for a government breach: specific performance of the agreement, which ordinarily means resentencing at which the government fully honors its promise, often before a different judge, or withdrawal of the guilty plea. The choice between them belongs to the court, which weighs what the circumstances of the case require.

The majority in Santobello left the remedy to the tribunal closest to the facts; Justice Douglas’s concurrence added that the defendant’s preference deserves considerable weight, and many courts do consider it. In practice, specific performance is the more common cure for a sentencing-recommendation breach, because it restores the benefit of the bargain without unwinding the conviction. Withdrawal — rescission of the whole exchange — tends to fit cases where performance can no longer be meaningful, such as a breach that has already exposed the court to information the agreement was supposed to keep out.

The Fifth Circuit has also recognized that some breaches can be cured before any remedy is needed. In Purser, the government initially urged a higher enhancement than the agreement allowed, then withdrew the objection and advocated the agreed position before and at sentencing; because the district court ultimately acted consistently with the agreement, the breach was cured. Cure is not harmless-error analysis: cure means the government actually performed the contract, while harmless error would excuse nonperformance. For a preserved objection, the Fifth Circuit holds that harmlessness plays no role at all — a standing breach requires a remedy even if the sentencing judge says the breach did not influence the sentence.

Objecting on Time: Preserved vs. Forfeited Breach Claims

Timing changes everything in breach litigation. An objection made when the breach occurs preserves the claim, and a preserved breach is not excused as harmless in the Fifth Circuit. A claim raised for the first time on appeal is reviewed only for plain error under Puckett v. United States, a far heavier burden.

Puckett settled the framework. A forfeited breach claim must clear all four plain-error prongs: an error, that is clear or obvious, that affected the defendant’s substantial rights, and that seriously affected the fairness, integrity, or public reputation of judicial proceedings. Where the broken promise related to sentencing, the substantial-rights showing means demonstrating a reasonable probability that the sentence itself would have been different. Puckett also rejected two shortcuts: a breach is not structural error that excuses the prejudice showing, and a breach does not retroactively make the plea unknowing or involuntary. The plea was valid when entered; the contract was broken afterward, and the law responds with a remedy rather than automatic invalidation.

The practical lesson for the defense is immediate objection. When the prosecutor’s sentencing argument crosses the line, counsel objects on the record, identifies the specific promise breached, and asks for a remedy — specific performance, resentencing before a different judge, or leave to withdraw the plea. Before sentencing, a documented breach can also anchor a motion to withdraw the plea under Rule 11(d)(2)(B), where it supplies a concrete, record-based reason rather than a bare change of heart.

Agreement Types, Court Rejection, and the Right to Withdraw

Not every plea agreement binds the court, and the withdrawal rights differ by type. Rule 11(c)(1) recognizes three kinds of government promises, and the automatic right to withdraw after a judicial rejection exists only for charge agreements under (c)(1)(A) and specific-sentence agreements under (c)(1)(C), not for (c)(1)(B) recommendations.

A second timing trap involves deferred acceptance. Courts routinely accept the guilty plea at the Rule 11 hearing but defer the plea agreement until the presentence report arrives. In United States v. Hyde, the Supreme Court held that a defendant in that posture cannot withdraw the plea as a matter of right; the plea and the agreement are separate events that can be accepted at different times, and withdrawal during the deferral window requires a fair and just reason. Hyde construed the former Rule 32(e), and its standard now lives in Rule 11(d)(2)(B). The window for cost-free withdrawal is therefore narrow: before the court accepts the plea itself, under Rule 11(d)(1), or after the court rejects a binding (A)- or (C)-type agreement.

Withdrawn Offers and the Defendant’s Side of the Bargain

The bargain runs both ways, and not every disappointed expectation is a breach. A prosecutor who withdraws an offer before any plea is entered generally cannot be forced to honor it, and a defendant who fails to keep his own promises risks releasing the government from its obligations under ordinary contract principles.

The withdrawn-offer rule comes from Mabry v. Johnson. There, a deputy prosecutor offered a concurrent-time recommendation, the defendant accepted through counsel, and the office then declared the offer a mistake and substituted a harsher one. The Supreme Court held that a plea agreement standing alone is a mere executory promise without constitutional significance; it is the guilty plea, not the negotiation, that implicates the Constitution. Because the defendant ultimately pleaded with full knowledge of the actual terms, no due-process violation occurred. The moment of no return, in other words, is the plea itself — promises broken after the defendant has performed are constitutional problems, while offers revoked before performance generally are not.

Defendants carry obligations too. Federal agreements routinely condition the government’s promises on truthful cooperation, complete financial disclosure, or abstention from new offenses, and a defendant who violates those conditions can forfeit the benefits of the bargain while the plea remains intact. One protection survives even aggressive drafting: the Fifth Circuit allows an alleged government breach to be raised on appeal despite an appeal-waiver provision, since the waiver itself is part of the very bargain the government is accused of breaking. For anyone weighing a withdrawal motion, a documented breach is among the strongest record-based grounds a defendant can present, because it rests on the government’s conduct rather than the defendant’s change of heart.

Where This Fits

This guide is one of four situations that arise under the Rule 11(d) Motion to Withdraw Guilty Plea. Start with the parent motion for the overall framework, or move to a related fact pattern:

What the Case Law Says

These decisions—verified against primary sources—control how this issue is litigated. Every case still turns on its own facts.

• Santobello v. New York, 404 U.S. 257 (1971) — When a plea rests in any significant degree on a prosecutor's promise, that promise must be fulfilled; the remedy choice rests with the court.
• Mabry v. Johnson, 467 U.S. 504 (1984) — A defendant's acceptance of a plea offer the prosecutor later withdrew creates no constitutional right to specific enforcement of that unexecuted bargain.
• United States v. Benchimol, 471 U.S. 453 (1985) — The government must make the recommendation it promised, but courts may not imply unagreed terms such as enthusiasm or an explanation of reasons.
• United States v. Hyde, 520 U.S. 670 (1997) — A defendant may not withdraw a guilty plea as of right while the court defers decision on the plea agreement; a fair-and-just reason is required.
• Puckett v. United States, 556 U.S. 129 (2009) — A forfeited claim that the government breached a plea agreement is reviewed for plain error; breach does not retroactively make the plea involuntary.
• United States v. Purser, 747 F.3d 284 (5th Cir. 2014) — Breach is judged against the defendant's reasonable understanding, construed strictly against the government; a preserved breach is not subject to harmless-error review.

General summaries of published opinions for information only — not predictions about any specific case.