Fixing the BOP How to fix the US Federal Bureau of Prisons
Topics

Compassionate Release Is Quietly Dying in Federal Court

16 September 2026

Legal Battles & Legislative Updates

Compassionate Release Is Quietly Dying in Federal Court

The United States Sentencing Commission published its preliminary compassionate release numbers for the third quarter of fiscal 2026 on September 9. In April, May and June the federal courts decided 497 motions under 18 U.S.C. § 3582(c)(1)(A) and granted 63 of them. For the fiscal year through June 30, the Commission counted 1,432 motions decided and 195 grants.

The Bureau of Prisons held 152,844 people on September 12. That works out to about 22 grants a month, one for roughly every 7,000 people in custody. In the same three months of 2023, the courts decided 940 motions. Volume is down by half in three years, and the grant rate sat between 10 and 18 percent in every month of 2026.

Two decisions in May

On May 28, the Supreme Court decided two cases that narrowed the statute further.

Rutherford v. United States involved the stacking rule under 18 U.S.C. § 924(c). Daniel Rutherford faced a 32-year mandatory minimum on two counts, and Johnnie Carter, whose case the Court heard with his, faced 57 years, because each additional § 924(c) count carried a consecutive 25-year term at the time they were sentenced. The First Step Act ended that stacking for first-time offenders, but Congress did not make the change retroactive, so neither man qualified. Both asked the sentencing court for a reduction anyway, arguing that the disparity created by the old law was an extraordinary and compelling reason to release them. The Court disagreed: “When Congress declines to make a sentencing amendment retroactive, the fact that a pre-amendment sentence is longer than it would have been post-amendment is not an ’extraordinary and compelling reaso[n]’ that ‘warrant[s]’ a sentence reduction.”

Fernandez v. United States involved a man convicted of murder for hire who argued that his conviction was invalid. The Court held that a prisoner attacking the validity of a conviction has to go through 28 U.S.C. § 2255, and that “the supposed invalidity of a conviction is not among the ’extraordinary and compelling reasons’ that justify compassionate release.”

Both rulings land on the categories judges were using. The Commission’s tally of the reasons cited in the 195 grants this fiscal year lists “unusually long sentence (10 or more years) and change in law” 24 times and “multiple 18 U.S.C. § 924(c) penalties” 15 times, out of 308 reasons total. Those are the two routes Rutherford shut, and both rested on the same fact pattern: a sentence that was lawful when it was imposed and that Congress or the courts later decided was too long for people sentenced after them. The circuits had split on the question, which is why the Court took the case. That split is over now.

The Commission’s policy statement at USSG § 1B1.13(b)(6) still lists an unusually long sentence as a possible reason for relief. The Court said what it thought of that provision: “Insofar as the Commission’s amended policy statement maintains otherwise, it conflicts with the statute.” What is left after May is a guideline provision that the Supreme Court has now called a conflict and a statute that reaches fewer cases every year.

Where you file decides your odds

The same statute produces wildly different results depending on the courthouse.

  • First Circuit: 51.7 percent granted, 15 of 29 motions.
  • Second Circuit: 37.0 percent, 20 of 54.
  • Ninth Circuit: 28.2 percent, 33 of 117.
  • Fourth Circuit: 16.9 percent, 59 of 350.
  • Third Circuit: 13.5 percent, 10 of 74.
  • Tenth Circuit: 13.0 percent, 6 of 46.
  • Eleventh Circuit: 10.8 percent, 19 of 176.
  • Fifth Circuit: 9.0 percent, 8 of 89.
  • Sixth Circuit: 6.3 percent, 8 of 127.
  • Seventh Circuit: 4.4 percent, 4 of 90.
  • Eighth Circuit: 4.0 percent, 11 of 276.

A motion filed in the First Circuit is worth thirteen times what the same motion is worth in the Eighth. At the district level the spread gets worse. The Eastern District of North Carolina denied all 52 motions it decided. The Northern District of Iowa denied 36 of 36. The Northern District of Ohio denied 30 of 30, the Eastern District of Kentucky 29 of 29, and Wisconsin’s two districts 29 of 29 between them.

This is the disparity the Sentencing Reform Act was written to reduce. The Commission’s inter-district study found that sentencing outcomes “continue to depend at least in part upon the district in which the defendant is sentenced,” and that the differences grew after the Supreme Court made the guidelines advisory in 2005. Once a sentence is imposed, compassionate release is the only mechanism that sends the case back to the sentencing court.

What keeps a motion from being heard

The Commission also counts the reasons courts gave when they denied motions this fiscal year. The largest was 18 U.S.C. § 3553(a), the sentencing factors, cited 661 times, or 22.9 percent of the denial reasons. Rehabilitation judged insufficient came next at 368, and “no extraordinary and compelling reason provided” at 272.

One entry stands out for how mechanical it is. Courts denied 214 motions because the person had not exhausted administrative remedies, which under the statute means either appealing a warden’s refusal or waiting out the 30 days after asking the Bureau to file on their behalf. That is 7.4 percent of the denial reasons given for people who had already served years, sometimes decades, and had already been refused by their warden.

Who the tool serves

The Commission reports the original sentence of everyone who received relief. Of the 195 people granted this fiscal year, 40.5 percent had been sentenced to 20 years or more, and another 25.8 percent to between 10 and 20 years. Two-thirds of the grants went to people serving a decade or longer. This is not a program for short-timers angling for an early exit. For a sentence that long, the only ways to cut an imposed term are clemency from the executive branch and this motion in court. Clemency depends on the president’s mood. The motion is a filing any prisoner can make, and it is the one that is closing.

The people holding those sentences are getting older

The Bureau’s own statistics show a population aging in place. Of the 152,844 people in custody on September 12, 21 percent were 51 or older, including 4,907 past 65. The average age is 42. The Inspector General flagged the trend in 2015, when inmates 50 and older were the fastest-growing segment of the population, up 25 percent from fiscal 2009 to fiscal 2013, and when the institutions with the highest share of older inmates spent five times more per person on medical care than the institutions with the lowest share.

The recidivism data on this population is not close to ambiguous. The Commission’s study of aging found that 13.4 percent of offenders released at 65 or older were rearrested within eight years, against 67.6 percent of those released before 21. In its study of people released in 2010, offenders sentenced for fraud, theft or embezzlement had the lowest rearrest rate of any offense group, 35.5 percent, against 70.6 percent for firearms offenders. Offenders aged 60 and over at release were rearrested at 15.9 percent, against 72.5 percent for those under 21.

Length of sentence carries far less of that weight than people assume. In the Commission’s matched-group study, the odds of recidivism were about 18 percent lower for offenders sentenced to more than 60 months up to 120 months, and about 29 percent lower for those sentenced to more than 120 months, compared with similar offenders who received shorter terms. On raw rates, the Commission has found that recidivism barely moves across sentence lengths at all, running from 50.8 percent for sentences of 6 months to 2 years to 55.5 percent for sentences of 5 to 9 years. A man in his late fifties in his fifteenth year is on the wrong side of both variables. The 35.5 percent figure is an average for an offense group that spans every age and sentence length, and age is the variable that moves it most.

The bill for holding him keeps arriving. The Bureau’s per capita sheet puts the annual cost of incarceration at $42,672, and at $86,798 in the medical referral centers where an aging population ends up. Earlier this year the Commission’s time credits rule and thirteen men facing re-incarceration both showed how narrow the room for relief has become.

What we think should happen

Compassionate release was written to be used. It is the one place in federal sentencing where a court can look at the person in front of it, at the sentence that person is serving, and at what the law and the data now say, then correct a number that no longer matches any of them. The standard is discretionary, so the outcome depends on who is reading the file, and right now that means a 13-fold difference between two circuits applying the same words.

Three changes would move the numbers, and none of them requires new theory.

Congress can make sentencing amendments retroactive when it intends them to reach people already in prison. The disparity the Court refused to treat as extraordinary in Rutherford came out of a drafting choice, not out of anything Rutherford or Carter did. Fix the choice.

The Commission should say in plain language what its own research shows about age and risk. Its studies have found the same pattern for a decade: the oldest release cohorts reoffend least, and fraud offenders reoffend least of all. A policy statement that treats an unusually long sentence as a reason for relief is not an invitation to abuse the statute. It is a response to what the numbers say about housing men in their sixties at $42,672 a year.

The Bureau could stop being the bottleneck. It has to be petitioned before anyone can file, its aging population is growing, the medical bills are concentrated in the same housing units, and warden denials produced 214 more denials this fiscal year.

The Commission publishes the next quarter’s counts in December. Nothing in the statute it measures has changed since May, and nothing in the trend since 2023 suggests December will look any different.

Sources