The Supreme Court’s Compassion Problem in Fernandez v. United States
If a federal judge sees credible evidence that someone may be innocent—or that Congress itself now says his sentence is way too long—should the judge be allowed to take that into account when deciding compassionate release?
On Wednesday, the Supreme Court will hear Fernandez v. United States, with two consolidated “companion” cases right behind it—Rutherford v. United States and Carter v. United States—that tee up the same core fight from a different angle. Collectively, they ask whether the words “extraordinary and compelling reasons” in 18 U.S.C. § 3582(c)(1)(A) mean what they say, or whether the Court will graft onto that statute yet another set of bars, locks, and deadbolts borrowed from the habeas and post‑conviction labyrinth. [1][2][3][4]
The justices should side with Fernandez. But if history is a guide—especially the Court’s near‑reflexive preference for finality over accuracy in Antiterrorism and Effective Death Penalty Act (AEDPA) cases—they likely won’t. And that failure will matter most in the very cases that keep judges up at night: when a person is serving a sentence that Congress has since repudiated, or worse, when he might be innocent.
What Fernandez Is (and Isn’t) About
Joe Fernandez was serving a mandatory life sentence. A federal district judge granted his motion for a sentence reduction under § 3582(c)(1)(A)—the “compassionate release” statute—based on two factors: serious doubts about his guilt and a stark disparity between his sentence and those of co‑defendants. The Second Circuit reversed, holding that concerns about factual innocence and sentencing error belong exclusively in a motion under 28 U.S.C. § 2255, not in compassionate release. [2]
The question before the Court is tight: May “extraordinary and compelling reasons” include reasons that could also support relief under § 2255? [1] Put differently, does § 3582 permit judges, in rare cases, to temper a sentence in light of powerful new information—even if that same information might also form the backbone of a collateral attack?
The companion cases pose the same problem from a different direction. Rutherford and Carter ask whether judges can consider non‑retroactive changes in sentencing law (for example, Congress’s First Step Act reforms that eliminated extreme “stacking” under § 924(c)) as part of the “extraordinary and compelling” calculus, and whether the U.S. Sentencing Commission validly recognized that ground in § 1B1.13’s 2023 amendments. [3][4][8][11]
Why Fernandez Should Win: The Text, the Structure, and the Point of § 3582
Start with the words Congress wrote. Section 3582(c)(1)(A) authorizes a court to “reduce the term of imprisonment” if, “after considering the factors set forth in section 3553(a),” it finds that “extraordinary and compelling reasons warrant such a reduction” and that the reduction is consistent with applicable Commission policy statements. Congress named one limit in § 994(t): “Rehabilitation of the defendant alone shall not be considered an extraordinary and compelling reason.” That’s it. [6][7]
No other carve‑outs appear in the statute. There’s no clause saying “except if the reason might also be raised in a § 2255 motion.” The Second Circuit’s decision demands we read one in.
Structure matters too. Compassionate release is a second‑look sentencing provision. It does not vacate a judgment; it modifies a sentence based on a holistic, equitable assessment that accounts for the § 3553(a) factors (history, characteristics, the need for just punishment, public safety, disparity, etc.). [6] By design, it operates outside the habeas channel. That channel—direct appeal and § 2255—is about legality; § 3582 is about equity and mercy, tempered by public safety.
And the Supreme Court has recognized, in the First Step Act resentencing context, that Congress gave judges broad discretion to consider “intervening changes of law or fact” when exercising their sentence‑modification powers. Concepcion v. United States did not require courts to grant relief, but it confirmed that Congress can and did authorize a wide evidentiary horizon when it hands judges a second look. [10] The same logic applies here, where Congress used capacious language and named only one limit.
If a district judge, armed with a developed record and the § 3553(a) factors, sees (1) credible, newly developed evidence calling guilt into question, and (2) a sentence that Congress now condemns as excessive, what is the judicial virtue in blinders? There is none.
What the Second Circuit Got Wrong
The Second Circuit framed the problem as “circumvention.” Letting judges consider innocence or sentencing disparities through § 3582, the court reasoned, risks “scuttling” post‑conviction review’s gatekeeping framework. [2] But that assumes the purpose of § 3582 is to replicate habeas. It isn’t.
A compassionate‑release reduction neither declares a conviction void nor opens floodgates. A petitioner still must satisfy a high threshold—“extraordinary and compelling reasons”—and then survive the § 3553(a) analysis, including public‑safety concerns. [6] That’s miles away from collateral vacatur.
Congress knew what it was doing when it amended § 3582 in the First Step Act to let defendants, not just the Bureau of Prisons, file motions. It was dissatisfied with BOP’s stinginess and wanted judges—not a carceral bureaucracy—to make the final call. [9] And when the Commission regained a quorum, it modernized § 1B1.13 to reflect what many courts were already doing: treating unusually long sentences and certain non‑retroactive changes in law as legitimate, cabined grounds for relief in truly exceptional cases. [8]
Put bluntly: Saying judges can never consider strong innocence evidence or a congressionally repudiated sentence when deciding whether mercy is warranted is not “faithful to AEDPA.” It’s a category error about § 3582’s basic function.
The Companion Cases: Disparities Congress Chose to Fix
Rutherford and Carter arise from the Third Circuit, which held that courts may not consider non‑retroactive changes in law (like the anti‑stacking fix for § 924(c)) when evaluating compassionate release and went further to question the Commission’s authority to say otherwise. [11] That court and others have invoked a separation‑of‑powers theme: If Congress didn’t make the change retroactive, judges can’t use compassionate release to soften the edges.
But the premise is off. Retroactivity rules tell us who automatically qualifies for resentencing. Compassionate release is not automatic; it’s discretionary and individualized. As other circuits have recognized, that distinction matters. A non‑retroactive change can still be part of a “rare case” mosaic, especially when paired with rehabilitation, age, health, or other unique facts. [7] And the Commission—expressly instructed by Congress to describe “extraordinary and compelling reasons”—did just that, drawing narrow, sensible conditions around when law‑changes may be considered. [7][8]
Nothing in § 3582(c)(1)(A) or § 994(t) authorizes courts to ignore reality. Concepcion says so. And for those who bristle at agency deference: the Sentencing Commission is not an executive agency; it’s a judicial‑branch body acting under a specific congressional directive. (That hasn’t stopped courts from calling its 2023 policy a “power grab,” especially in the post‑Chevron era—but that critique stretches both text and separation‑of‑powers logic. [14])
Why the Court Probably Won’t Do the Right Thing
If you follow the Court’s post‑conviction docket, you’ve seen this movie. Over the last decade, AEDPA doctrine has calcified into a jurisprudence of exhaustion, default, “fair‑minded jurists could disagree,” and record‑freezing rules that elevate finality above accuracy—even in capital cases.
Consider just a few waypoints:
Harrington v. Richter instructs federal courts to deny relief so long as a state decision is not beyond “any possibility for fair‑minded disagreement.” That standard is so forgiving to the state that clear error often isn’t enough. [16]
Cullen v. Pinholster prohibits federal courts from considering new evidence when reviewing claims adjudicated on the merits in state court, even if that state record was truncated or poorly developed. [17]
Brown v. Davenport doubles up hurdles: satisfying the Brecht harmless‑error standard isn’t enough; petitioners must also run AEDPA’s gauntlet. [18]
Shinn v. Martinez Ramirez blocks federal evidentiary development even when state post‑conviction counsel was ineffective, closing a critical safety valve for innocence‑related claims. [15]
Jones v. Hendrix closed the federal “savings clause” to prisoners whose conduct isn’t criminal under current law but who can’t meet § 2255’s second‑or‑successive bar. Translation: you could be legally innocent and still stuck. [19]
Layer those decisions together and ask yourself: Does a majority of this Court look likely to read § 3582 broadly, in a way that mitigates AEDPA’s harshness? Or will it characterize compassionate release as an “end run” around the fortress it has built?
I’d love to be wrong, but the trend line is unmistakable.
The Real‑World Stakes: Actual Innocence Isn’t Abstract
Courts don’t like “freestanding innocence” claims. Herrera v. Collins denied relief for a prisoner who asserted actual innocence; later cases allowed “actual innocence” as a narrow gateway to reach otherwise barred claims, but only under exacting showings. [20][21][22][23] Even when the Supreme Court took the extraordinary step in In re Davis of sending an innocence claim back for a hearing, the district court concluded that Troy Davis had not clearly established his innocence and he was executed. [24][25]
We know, from painful experience, that the system gets it wrong. Cameron Todd Willingham was executed after an arson conviction now widely discredited by modern fire science. [26] Carlos DeLuna almost certainly died for another man’s crime. [28] Post‑execution DNA testing in Ledell Lee’s case found DNA from an unknown male on the murder weapon. [29] Since 1973, roughly 200 people sentenced to death have been exonerated—about one for every eight executions. [30]
None of this means Fernandez is innocent. It does mean that when a sentencing judge, in a case‑specific record, sees reasons that would make any reasonable person pause, our law should not forbid that judge from pausing. Compassionate release is precisely where mercy and institutional humility are supposed to meet.
Answering the Counterarguments
“But Congress chose not to make those changes retroactive.”
Correct—and compassionate release is not automatic retroactivity. It is an individualized second look. Congress knowingly created that safety valve and told the Commission to describe exemplars. [6][7][8]
“But this will flood the courts.”
The flood already happened. After the First Step Act allowed defendant‑filed motions, tens of thousands were filed, and district courts—applying § 3553(a)—denied the large majority. When judges granted relief, it was typically due to exceptional health risks (COVID‑19), extreme sentence length disparities, or unique family circumstances. The statute’s demanding standards and public‑safety backstop are working as designed. [8][9]
“But § 2255 exists for innocence claims and sentencing errors.”
True—and often unavailable. Legal changes are frequently non‑retroactive; new innocence evidence is frequently barred or undeveloped due to AEDPA strictures like Pinholster and Shinn. Section 3582 is not a substitute for vacatur; it is an equitable release valve to correct tragedies without rewriting history. [15][17]
What a Better Rule Looks Like
The Court should hold that § 3582’s “extraordinary and compelling reasons” may include (1) credible evidence undermining the integrity of the conviction (including substantial post‑trial innocence evidence), and (2) profound sentence‑length disparities created by non‑retroactive legal changes—not as automatic grounds, but as factors a district court may weigh alongside rehabilitation, age, health, risk, and the § 3553(a) goals.
That would align with the statutory text, with Congress’s direction to the Commission, and with Concepcion’s common‑sense recognition that second looks should consider reality. [6][7][8][10] It also respects separation of powers by leaving mercy in the hands of the Article III judge who imposed the punishment.
What We’re Likely to Get Instead
Expect a narrowing opinion. The majority may announce a categorical bar: if a reason is “cognizable” under § 2255, it cannot be “extraordinary and compelling” under § 3582. Or it may invalidate (or neuter) the Commission’s 2023 policy on considering non‑retroactive law changes, embracing the Third and Sixth Circuit’s skepticism. [11][12][14]
Either way, the Court will likely emphasize finality and channeling—familiar themes from Richter, Pinholster, Brown, Shinn, and Jones. [15][16][17][18][19] The losers won’t be “technicalities.” They’ll be people serving outsize sentences that Congress tried to fix—and a subset who may, in fact, be innocent.
A Final, Uncomfortable Point
We tell ourselves the justice system is calibrated to prefer ten guilty men go free rather than one innocent be punished. In post‑conviction doctrine, we often behave as if the inverse is acceptable. AEDPA’s deference rules, evidentiary choke points, and serial‑petition barriers deliver predictable results: meritorious claims are lost to procedure; innocence evidence is never heard; and second looks are narrowed to the vanishing point.
Compassionate release is not a jailbreak. It is a humane statutory mechanism that trusts judges to exercise discretion case by case. If the Supreme Court won’t let them do even that, we’ll know exactly what value is driving the law. And it won’t be accuracy.
If you want more on how courts, Congress, and the Department of Justice approach these questions in real life, I explore the rhetorical and strategic side of legal change in Truth & Persuasion and how to navigate federal investigations in Fight the Feds (both on my Amazon author page; see Truth & Persuasion and Fight the Feds there). You can also find deeper dives at ronaldwchapman.com, on YouTube, and on LinkedIn.
Key Takeaways
The text of § 3582(c)(1)(A) and § 994(t) gives judges broad discretion to consider “extraordinary and compelling” reasons, limited only by Congress’s “rehabilitation alone” rule. [6][7]
Compassionate release is a discretionary second look—not a collateral attack—and may properly consider credible innocence evidence and extreme sentence disparities created by non‑retroactive legal changes. [8][10]
The Supreme Court’s AEDPA jurisprudence (Richter, Pinholster, Brown, Shinn, Jones) consistently favors finality over accuracy, suggesting the Court may again narrow relief. [15][16][17][18][19]
Real lives are at stake. We have executed or nearly executed people amid grave doubts (Willingham, DeLuna, Davis), and about 200 death‑sentenced people have been exonerated since 1973. [24][25][26][28][30]
The Court should side with Fernandez. It probably won’t.
Footnotes
[1] Fernandez v. United States, No. 24‑556 (argued Nov. 12, 2025) (Question Presented).
[2] United States v. Fernandez, 104 F.4th 420 (2d Cir. 2024).
[3] Rutherford v. United States, No. 24‑820, and Carter v. United States, No. 24‑860 (consolidated for argument Nov. 12, 2025).
[4] Legal Information Institute (Cornell), Rutherford v. United States, No. 24‑820 (issue overview).
[5] Durbin & Booker, Press Release, “Durbin, Booker Urge Supreme Court to Maintain… in Fernandez v. United States” (Aug. 12, 2025).
[6] 18 U.S.C. § 3582(c)(1)(A).
[7] 18 U.S.C. § 994(t).
[8] U.S. Sentencing Commission, Amendment 814 (effective Nov. 1, 2023) (revising U.S.S.G. § 1B1.13).
[9] United States v. Brooker (Zullo), 976 F.3d 228 (2d Cir. 2020).
[10] Concepcion v. United States, 597 U.S. 481 (2022).
[11] United States v. Rutherford, 120 F.4th 360 (3d Cir. 2024).
[12] United States v. Bricker, ___ F.4th ___ (6th Cir. Apr. 22, 2025).
[13] United States v. Black, 131 F.4th 542 (7th Cir. 2025).
[14] Reuters, “U.S. Sentencing Commission’s expansion of compassionate release was invalid ‘power grab,’ court rules” (Apr. 23, 2025).
[15] Shinn v. Martinez Ramirez, 596 U.S. ___ (2022).
[16] Harrington v. Richter, 562 U.S. 86 (2011).
[17] Cullen v. Pinholster, 563 U.S. 170 (2011).
[18] Brown v. Davenport, 596 U.S. 118 (2022).
[19] Jones v. Hendrix, 599 U.S. ___ (2023).
[20] Herrera v. Collins, 506 U.S. 390 (1993).
[21] Schlup v. Delo, 513 U.S. 298 (1995).
[22] House v. Bell, 547 U.S. 518 (2006).
[23] McQuiggin v. Perkins, 569 U.S. 383 (2013).
[24] In re Davis, 130 S. Ct. 1 (2009) (transferring to district court for innocence hearing).
[25] In re Davis, No. CV409‑130, 2010 WL 3385081 (S.D. Ga. Aug. 24, 2010) (post‑hearing order).
[26] David Grann, “Trial by Fire,” The New Yorker (Sept. 7, 2009).
[27] Texas Forensic Science Commission (Beyler Report on arson investigation in Willingham).
[28] James S. Liebman, The Wrong Carlos: Anatomy of a Wrongful Execution (Columbia Univ. Press 2014).
[29] Death Penalty Information Center, “Forensic Testing Casts New Doubt on Guilt of Ledell Lee” (May 11, 2021).
[30] Death Penalty Information Center, “Innocence” (updated 2025) (approximately 200 death‑row exonerations since 1973).
About Ronald W. Chapman II
Ronald W. Chapman II is a Marine Corps veteran, nationally recognized attorney, and prolific author devoted to helping others understand the complexities of our justice system, civil liberties, and federal investigations intersect. He has successfully represented clients in high‑stakes cases across the country, earning a reputation for strategic thinking, tenacity, and unwavering dedication to justice.
Beyond the courtroom, Ronald’s passion for clarity and fairness drives his writing. His books peel back the layers of legal intricacies, blending real‑life experiences with accessible explanations to equip readers for their own legal challenges. Whether you’re a healthcare professional navigating regulatory waters, a concerned citizen seeking to understand your rights, or simply a reader fascinated by true legal drama, Ronald’s works offer invaluable insights.
Connect with Ronald W. Chapman II on Twitter @RonChapmanII, Instagram @ron_chapman, and LinkedIn Ronald W. Chapman II. Explore his books on Amazon and subscribe to his Substack for exclusive content, case studies, and fresh takes on how law is shaping—and reshaping—our world.