The number that mattered was six billion dollars, and it belonged to a family that never filed for bankruptcy.
That was the arithmetic at the center of Harrington v. Purdue Pharma L.P., 603 U.S. 204 (2024). Purdue, the maker of OxyContin, went into Chapter 11 in September 2019 before Judge Robert Drain in the Southern District of New York. The Sacklers, who owned it, did not. They had spent years pulling money out of the company, roughly eleven billion dollars by the estimates that circulated during the case, and they proposed to hand back about six billion over time. In exchange they wanted something a debtor cannot ordinarily buy: total peace. A release, imposed on every opioid victim in the country, whether that victim agreed or not, extinguishing any claim against the family forever.
I have taught professional responsibility for a long time, and I tell my students that the most interesting questions in law are almost never about whether a defendant did wrong. They are about who gets to decide, and by what authority. Purdue was that kind of case. Nobody serious disputed that the Sacklers had exposure. The question was whether a bankruptcy court, presiding over a company’s reorganization, could reach out and cancel the property rights of people who were strangers to that bankruptcy, who had voted no, or who had never voted at all.
On June 27, 2024, a five-Justice majority said no.
What the Court actually decided, and what it pointedly did not
Justice Gorsuch wrote for the majority, joined by Thomas, Alito, Barrett, and Jackson. It was one of those alignments that reminds you the Court’s ideological map is drawn in pencil. The holding was narrow in its logic and enormous in its consequence. Section 1123(b)(6) of the Bankruptcy Code, the catch-all provision that lets a plan include any appropriate provision not inconsistent with the Code, does not authorize a release that discharges the claims of nonconsenting third parties against a non-debtor.
Gorsuch read the catch-all the way he reads most catch-alls, through the company it keeps. The specific items listed before it in 1123(b) all concern the debtor and the debtor’s property. So the residual clause, he reasoned, cannot be stretched to do something the specific clauses never contemplated, which is to relieve a solvent outsider of liability without the consent of the people holding the claims. A discharge, he reminded everyone, is a benefit the Code extends to the honest but unfortunate debtor who puts his assets on the table. The Sacklers wanted the discharge without being debtors and without putting their assets on the table. The Code, he wrote, does not contain any provision authorizing that outcome.
Here is the part practitioners keep skipping. Gorsuch was careful about the boundary. The opinion does not touch consensual releases. It does not reach the question of what counts as consent, whether an opt-out mechanism is enough, or whether silence can be bought. It does not disturb section 524(g), the asbestos-specific channeling scheme that Congress wrote into the Code in 1994. And it expressly declined to decide whether its holding would unwind a plan that had already been substantially consummated. The Court drew a line and then stood back from the edge of it.
Justice Kavanaugh dissented, joined by Chief Justice Roberts and Justices Sotomayor and Kagan. His opinion ran long and it ran hot. He argued that nonconsensual releases had been a fixture of complex reorganizations for decades, that the majority was substituting a dictionary for practical experience, and that the real losers were the opioid victims, who would now get less and wait longer. He was not wrong about the human cost. He was, I think, wrong about the authority, and the difference between those two propositions is the whole subject of the case.
A doctrine that grew up in the dark
To understand why Purdue landed the way it did, you have to remember that nonconsensual third-party releases were never enacted. They were assembled.
The origin story is asbestos. In 1982 Johns-Manville, then the country’s largest asbestos producer, filed for bankruptcy while still profitable, because the tort liabilities projected out over decades dwarfed anything the balance sheet could absorb. Out of that case came the channeling injunction: a trust funded by the debtor and its insurers, and an order routing all present and future claims to the trust and away from the operating company. Congress liked the mechanism enough to codify it. Section 524(g), added by the Bankruptcy Reform Act of 1994, blessed the Manville approach for asbestos and only for asbestos, and it built in guardrails, including a supermajority requirement that seventy-five percent of voting claimants approve.
That was the tell that the majority in Purdue seized on. When Congress wanted to authorize this extraordinary device, it knew exactly how to do it, it did it once, and it hedged it with conditions. Reading a general grant of the same power into the catch-all would make the whole careful apparatus of 524(g) pointless.
But the lower courts did not wait for Congress. Through the 1980s and 1990s the circuits improvised. The Fourth Circuit blessed releases in the Dalkon Shield litigation, In re A.H. Robins Co., 880 F.2d 694 (4th Cir. 1989), and the factors it used floated free of their facts and became a checklist courts recited elsewhere. The Sixth Circuit followed in the breast-implant mess, In re Dow Corning Corp., 280 F.3d 648 (6th Cir. 2002), enumerating seven considerations that read more like a mood than a rule. The Second Circuit tried to discipline the practice in In re Metromedia Fiber Network, Inc., 416 F.3d 136 (2d Cir. 2005), warning that such releases should be granted only in rare cases and only where truly necessary, which is the kind of language a court uses right before it grants the thing anyway.
Not everyone joined the parade. The Fifth Circuit rejected nonconsensual releases in In re Pacific Lumber Co., 584 F.3d 229 (5th Cir. 2009). The Ninth and Tenth Circuits had long been hostile. The Third Circuit approached them with visible suspicion in In re Continental Airlines, 203 F.3d 203 (3d Cir. 2000). So by the time Purdue reached the Supreme Court there was a genuine, mature circuit split, the kind that had been ripening for a generation while the bar quietly built a whole practice area on top of the majority view.
The Second Circuit had upheld Purdue’s plan in In re Purdue Pharma L.P., 69 F.4th 45 (2d Cir. 2023), reversing District Judge Colleen McMahon, who in December 2021 had vacated confirmation and written, memorably, that the case raised a question that had confounded courts for decades and that she doubted the Bankruptcy Code answered. She turned out to be reading the statute the way five Justices eventually would.
The professional responsibility problem hiding inside the doctrine
I want to sit here for a moment, because this is where my particular corner of the law has something to say that the bankruptcy specialists sometimes miss.
Model Rule 1.8(g), the aggregate settlement rule, is one of the oldest ethical commitments the American bar has. A lawyer who represents multiple clients may not settle their claims as a group unless each client gives informed consent, in writing, after learning the terms of every other client’s deal. The rule exists because the temptation to trade one client’s interest for another’s is not hypothetical. It is the ordinary physics of representing a crowd. The rule insists that consent be individual, informed, and real.
Now look at what a nonconsensual release does. It takes the very thing Rule 1.8(g) protects, the client’s sovereign choice over her own claim, and it removes that choice by court order. The claimant who wanted her day against Richard Sackler personally, who did not care about the trust distribution schedule, who wanted a jury to hear what her son’s addiction cost, was told her claim no longer existed. Not because she settled it. Because a plan she voted against, or never voted on, settled it for her.
There is a real argument on the other side, and I make it to my students every year so they do not mistake indignation for analysis. Mass torts are a collective action problem. If every claimant holds a veto, the holdouts extract a premium and the settlement collapses, and the ordinary victims get nothing while the litigation lawyers bill for another decade. Aggregation is not a betrayal of clients. It is often the only way to deliver them anything at all. That is Kavanaugh’s point, and it deserves respect.
But respect is not the same as authority. The ethical rules bend individual consent toward the group through disclosure and voluntary agreement. The bankruptcy release replaced consent with compulsion and then called the result a settlement. Those are different animals wearing the same coat. Gorsuch, whatever you think of his textualism, saw the coat and looked underneath it.
The Texas two-step and the company that was born to file
While Purdue was climbing toward the Supreme Court, another strategy was running in parallel, and it was more audacious.
Johnson & Johnson faced tens of thousands of claims that its talcum powder caused ovarian cancer and mesothelioma. J&J is not distressed. It is one of the most valuable enterprises on earth. So its lawyers turned to a maneuver that had a nickname before it had a body of law: the Texas two-step. Under the Texas Business Organizations Code, a company can execute a divisional merger, splitting itself into two entities and allocating assets to one and liabilities to the other. J&J did exactly that, creating a subsidiary called LTL Management, loading it with the talc liabilities and a funding agreement, and then marching that subsidiary into Chapter 11 in North Carolina, later moved to New Jersey.
The idea was elegant and, to many observers, offensive. Put the liabilities in a shell, file the shell, and use the bankruptcy court’s power to channel every talc claim into a trust and enjoin suits against the healthy parent. A solvent giant would obtain, through a two-day-old subsidiary, the litigation peace that bankruptcy is supposed to reserve for the failing.
The Third Circuit was not persuaded. In In re LTL Management, LLC, 64 F.4th 84 (3d Cir. 2023), the court dismissed the petition, holding that a debtor must be in financial distress to file in good faith, and LTL, backed by a funding agreement worth tens of billions from its parent, was not. Judge Ambro wrote that good intentions, including to protect the J&J brand and efficiently resolve claims, do not suffice alone. LTL had access to cash on a scale no genuine debtor enjoys, and so it did not belong in the bankruptcy court at all. J&J refiled almost immediately with a slightly reengineered funding structure, and the same court sent it out the door again.
The two-step did not die there. Georgia-Pacific’s Bestwall, Trane’s Aldrich Pump and Murray Boiler, and other divisional-merger debtors remained pending in the Western District of North Carolina, where the bankruptcy bench had been more receptive. But the theory was wounded, and everyone in the mass-tort bar knew it. What Purdue then did was remove the escape route the two-step had been running toward. Even a debtor who could clear the good-faith hurdle would arrive at confirmation to find that the prize at the end, the nonconsensual release of the non-debtor parent and its officers, was no longer available under the Code.
How 2025 rearranged the board
The interesting thing about a decision like Purdue is that it does not end the strategy it forbids. It reprices it. Watch what the sophisticated players did in the year that followed, because that is where the real teaching sits.
Start with Purdue itself. The Sacklers had always said their six billion was contingent on the releases, and after June 2024 those releases were gone. The parties went back into mediation. By January 2025 a reconstructed deal emerged, larger than the one the Court had struck down, on the order of 7.4 billion dollars, with the family’s contribution increased and the timeline compressed. The structural difference was the whole point. This time the architecture was built to secure consent rather than to compel it, giving states and claimant groups a mechanism to sign on to the release rather than have it dropped on them. Attorneys general who had fought the earlier plan, including New York’s, announced support. The money went up precisely because the leverage of compulsion went away, which is the opposite of what the dissent predicted, and I say that gently because the dissent’s fear was reasonable and simply did not come true here.
Then there was J&J’s third attempt, and it is the cleanest illustration of the new order. Having been rebuffed twice in New Jersey, the company moved its filing to Texas, created a new subsidiary called Red River Talc LLC, and this time tried a different theory. Rather than argue for a purely nonconsensual release, J&J ran a prepackaged bankruptcy, soliciting votes from claimants before filing and asserting that it had cleared the seventy-five percent threshold borrowed from the asbestos statute, thereby manufacturing something it could call consent. It filed in the Southern District of Texas, offering upward of ten billion dollars.
Judge Christopher Lopez dismissed it. In In re Red River Talc LLC, decided March 31, 2025, in the Bankruptcy Court for the Southern District of Texas, the court found the vote irreparably tainted. The solicitation had swept in claimants whose votes were procured through questionable methods, counted votes cast by lawyers for clients who had never actually seen the deal, and relied on a voting record the court could not trust. Consent that is engineered is not consent, and a supermajority assembled through a defective process cannot substitute for it. The prepackaged path, the industry’s most creative answer to Purdue, failed on its first real test.
J&J’s response was telling. Rather than appeal or refile a fourth time, the company announced it was abandoning the bankruptcy strategy altogether and would litigate the talc claims in the tort system, defending cases on the merits. After roughly five years and three filings, the company chose the courtroom it had spent all that effort trying to avoid. The Coalition of Counsel for Justice for Talc Claimants, which had opposed every filing, treated it as vindication, and on the narrow question of whether the two-step-into-a-release could deliver forced peace, it was.
Around the same time the bankruptcy bench and the claimants’ bar recalibrated their vocabulary. The phrase you started hearing everywhere was consensual release, and beneath it a genuinely hard question the Court had left open: what does consent mean? Is an opt-out enough, where a claimant who fails to check a box is deemed to have agreed? Some judges accepted opt-out structures as sufficient. Others treated silence as the absence of consent rather than its presence, which is closer to how Rule 1.8(g) would see it. That fight, over whether inaction can bind, is where the doctrine is now being made, one confirmation order at a time, and it will take years to settle.
The other adaptation was migratory. Multidistrict litigation reasserted itself as the default venue for mass torts, because the MDL settlement, unlike the bankruptcy release, has always run on individual consent and inventory deals negotiated firm by firm. The 3M Combat Arms earplug litigation had already pointed the way. 3M put its subsidiary Aearo into bankruptcy in 2022 to corral the claims, an Indiana bankruptcy court threw the case out, and 3M then settled the underlying MDL for roughly six billion dollars in 2023 without any release the Code would have to bless. That sequence, bankruptcy rejected, tort system used, ten-figure settlement reached, became the template J&J effectively adopted in 2025.
What I think the case is really about
I am supposed to be gently contrarian here, so let me be, and against the grain of my own tribe. Most of the professional-responsibility commentary treated Purdue as a morality play, the greedy family denied its bought absolution, and there is satisfaction in that reading. I do not fully trust it.
The uncomfortable truth is that the victims in Purdue may have been better served by the thing the Court struck down. A forced release that delivered money in 2022 might have beaten a voluntary one that delivered more money in 2026, because many opioid claimants are poor and sick and cannot wait, and because the family’s fortune was mobile and the litigation to chase it would have been brutal. Kavanaugh understood that, and anyone who cheers the decision without feeling the weight of his dissent has not read it carefully.
And yet I think the majority was right, for a reason that has less to do with bankruptcy than with the shape of legal power. There is a difference between a court resolving a dispute between the parties before it and a court abolishing the rights of people who are not parties, who did not consent, and who never got their day. That second thing is not adjudication. It is legislation, performed by a bankruptcy judge, funded by a private settlement, and dressed in the language of efficiency. The efficiency is real. So is the danger. Once you accept that a solvent non-debtor can purchase immunity from the claims of nonconsenting strangers because a reorganization plan finds it convenient, you have handed the Chapter 11 process a power that no statute granted and that no election authorized. The next family to use it will be less sympathetic and the claims it extinguishes less compensated, and the precedent will not care about the difference.
My students sometimes ask why the aggregate settlement rule bothers to demand individual written consent when everyone knows the group will settle together anyway. The answer is that the consent is not a formality. It is the thing that makes the settlement the client’s rather than the lawyer’s, and it is the thing that keeps a resolution a resolution instead of a taking. Purdue said the same thing in the register of the Bankruptcy Code. You can settle a claim. You cannot vaporize it over the owner’s objection and call the vapor a deal.
The mass-tort bar heard the message and, being competent, adapted within a year, which is what competent lawyers do. The two-step limps on in the North Carolina cases that were already filed. The prepackaged consent theory took a bad wound in Houston. The MDL machinery absorbed the traffic the bankruptcy courts turned away. And the open question, whether an opt-out is consent or merely its imitation, is the ground on which the next decade of this fight will be waged. I would watch that question closely. It is where the pressure the Court relieved will try, quietly, to come back in.