Photo of Steven Boranian

The Supreme Court recently issued a unanimous decision resolving a circuit split on an important question of federal subject matter jurisdiction: Whether a district court’s erroneous dismissal of a nondiverse party before final judgment can cure a jurisdictional defect that existed when a case was removed to federal court.  You might think this is a question that only jurisdiction wonks like us would care about. 

But you would be wrong.  When the Supreme Court granted review in this case, Bexis wrote a post called “Will the Supreme Court Turn Fraudulent Joinder into a Free Shot for Plaintiffs?”  The answer, unfortunately, is yes.  The Supreme Court has ruled that a district cannot cure a jurisdictional defect by involuntarily dismissing a nondiverse defendant, even if the case goes all the way through trial to judgment.  This of course matters a great deal to us because plaintiffs suing in product liability cases often join nondiverse or local defendants to prevent removal—distributors, retailers, pharmacies, healthcare providers, etc.  We often resist these efforts with arguments like fraudulent joinder, and we often succeed.  The Supreme Court’s opinion, however, has amplified the risk that an erroneous jurisdictional ruling in the district court will render moot everything that comes after. 

The case is Hain Celestial Group, Inc. v. Palmquist, No. 24-724, 2026 WL 501733 (U.S. Feb. 24, 2026) (to be published), which arose from a child born in 2014 to the plaintiffs.  For the first two years, the plaintiffs (Texas citizens) fed their child baby food manufactured by one defendant and purchased from another (also a Texas citizen).  When, however, the child began exhibiting developmental disorders, some doctors attributed these conditions to heavy-metal poisoning.  Id. at *3. 

Fast forward to 2021, when a subcommittee of the U.S. House of Representatives released a staff report purporting to find that certain baby foods contained elevated levels of toxic heavy metals.  Following the report’s release, the plaintiffs sued both the manufacturer and the retailer in Texas state court.  

The manufacturer (a citizen of Delaware and New York) promptly removed the case to federal court on the basis of diversity jurisdiction.  The retailer defendant, however, was a Texas citizen, like the plaintiffs, so the manufacturer argued that the retailer had been improperly joined and should be dismissed.  That would eliminate the only nondiverse defendant.  Id. at *3-*4. 

The district court agreed that there was complete diversity, applying Fifth Circuit precedent holding that a defendant is improperly joined if the plaintiff fails to state a facially plausible claim against that defendant.  The court concluded that the plaintiffs had not alleged a plausible claim against the Texas retailer defendant, dismissed that defendant, and denied the plaintiffs’ motion to remand.  Id. at *4. 

Let’s pause here.  Fraudulent joinder and/or misjoinder are arguments that we and most everyone who regularly reads this blog have used thousands of times—always with a sound basis in fact and law and often successfully.  If a district court’s order denying remand can be second guessed at any time, even after the case has proceeded through trial to judgment, that order gives the plaintiff an ace up the sleave—a jurisdictional issue built into the case, potentially resulting in a do over if the plaintiff loses at trial.

That is what happened in Hain Celestial.  The case proceeded to trial against the manufacturer alone, and the manufacturer prevailed on a motion at the close of the plaintiffs’ case.  In entering judgment as a matter of law, the district court concluded that the plaintiffs had failed to present sufficient evidence that ingestion of heavy metals could cause their child’s symptoms or at what levels such metals would need to be ingested to produce those symptoms.  Id.

On appeal, the Fifth Circuit reversed the improper-joinder decision, holding that the Texas retailer was properly joined and thus was erroneously dismissed.  Because the district court thus lacked diversity jurisdiction, the Fifth Circuit vacated the judgment.  Id. 

The Supreme Court granted certiorari to resolve a split among the Courts of Appeals on whether vacatur is required under these circumstances, and it affirmed the Fifth Circuit’s decision.  Justice Sotomayor’s opinion began by establishing two foundational principles.  First, no party asked the Court to revisit the Fifth Circuit’s holding that Whole Foods should not have been dismissed.  Second, no party disputed that had the district court correctly performed the joinder analysis at the outset, it would have lacked jurisdiction and been required to remand the case to state court.  

The only remaining question was whether the district court, having erroneously dismissed the Texas retailer, had jurisdiction to enter a final judgment as to the manufacturer.  Id. 

As previewed above, the Supreme Court held that the district court lacked jurisdiction to enter judgment.  The Supreme Court reaffirmed that federal courts are courts of limited jurisdiction and that appellate courts must satisfy themselves of the jurisdiction of lower courts under review.  Generally, appellate courts assess a district court’s jurisdiction based on the facts that existed at the time of filing or removal.  If jurisdiction was lacking at that time, any judgment on the merits must typically be vacated.

There is, however, an exception.  In Caterpillar Inc. v. Lewis, 519 U.S. 61 (1996), the Supreme Court held that if a district court properly cures a jurisdictional defect before final judgment, the appellate court need not vacate that judgment.  In Caterpillar, the nondiverse defendant was fully dismissed with all parties’ consent before trial commenced.  Under those circumstances, the verdict could stand, although the Supreme Court further clarified that if the jurisdictional defect “lingered through judgment” uncured, then vacatur would be required.  Hain Celestial, at *5. 

Applying these principles, the Supreme Court held that the erroneous dismissal of the Texas retailer did not cure the jurisdictional defect.  When the Fifth Circuit reversed the district court’s error, it restored the Texas retailer to the case and thereby destroyed complete diversity.  Consequently, the jurisdictional defect “lingered through judgment” and required vacatur.  Id. *6.  As the Supreme Court observed, the Texas retailer was “only temporarily and erroneously removed from the case; it was not ‘gone for good.’”  Id.

The Supreme Court rejected the manufacturer’s argument that the correctness of the district court’s dismissal was irrelevant because the parties were completely diverse at the time of final judgment.  The Court emphasized that it has never held that a district court can create jurisdiction through its own mistakes.  The Court also rejected considerations of efficiency.  This case went all the way to trial, with considerable time and effort by the parties and the district court alike.  That all turned out to be a waste of time.  The Supreme Court, however, noted that while Caterpillar recognized that considerations of finality and efficiency become “overwhelming” after a diversity case has been tried in federal court, those considerations apply only after a jurisdictional defect has been properly and finally cured.  Id. 

Finally, the Supreme Court rejected the manufacturer’s fallback position that the Texas retailer should be dismissed under Federal Rule of Civil Procedure 21, which permits courts to add or drop parties “on just terms.”  Citing Newman-Green, Inc. v. Alfonzo-Larrain, 490 U.S. 826 (1989), the Court noted that, while Rule 21 may be used to dismiss dispensable nondiverse parties in appropriate circumstances, courts must carefully consider whether such dismissal would prejudice the parties.  Critically, in Newman-Green, the plaintiff sought the dismissal; here, a defendant sought to dismiss another defendant over the plaintiffs’ consistent objections.  The Court emphasized that plaintiffs are the “masters of the complaint” and generally have the right to choose their forum.  The plaintiffs here exercised that right by purposefully and properly joining a nondiverse defendant against whom they could not proceed in federal court.  Rule 21 does not permit a court or defendant to override that choice in these circumstances.  Id. at *7. 

So there you have it.  The Hain Celestial opinion reinforces that removing defendants cannot rely on erroneous interlocutory dismissals of nondiverse parties to establish federal jurisdiction.  The Supreme Court’s analysis makes clear that the jurisdictional cure recognized in Caterpillar requires a proper and final dismissal—not merely the removal of a party through an erroneous ruling that remains subject to appellate review.  We would also be remiss if we did not mention Justice Thomas’s concurring opinion, which wondered whether the doctrine of fraudulent joinder should exist at all, outside of cases of bad faith, actual fraud, or other “extreme abuse.”  Id. at *8-*9. 

Will this prevent us from arguing fraudulent joinder or misjoinder when properly removing cases to federal court?  No, it won’t.  But if the plaintiff has contested complete diversity and lost, we will surely hear about it on appeal.  And, if the district court was wrong, the Supreme Court has now said that any judgment for the defendant is at greater risk. 

Photo of Eric Alexander

Yes, we know our title is rather circular.  In fairness, most of the allegations we have seen in cases attempting to impose liability on the sponsor of a clinical trial are too.  We have been following proposed clinical trial liability in its various proposed forms for quite some time, including winning a jury trial for a sponsor sued by a clinical trial participant longer ago than we would care to recognize.  For some reason, we see fewer of these cases than we used to.  Part of that change may be because plaintiff lawyers tend to prefer drug or device cases that are part of litigations, where the investment needed to get up to speed can be justified by the number of cases.  Part of it may be because sponsors tend to win these cases.  Many theories of liability have been offered, but few have stuck.  This makes sense to us, of course.  The informed consent process for a clinical trial is invariably thorough and well documented.  For medical products that are not yet on the market, clinical trial participants know in advance that the risks and benefits are uncertain.  For a double-blind placebo-controlled trial, the participants know they may be getting a “sugar pill.”  For the study of a disease where an active treatment comparator is required, such as diabetes, they certainly know that they do not get to pick which treatment they will receive or otherwise deviate from the study protocol if they want to remain in the study.  They get free care within the bounds of the study in exchange for participating in the study.  In short, clinical trials have a number of differences from the typical prescription of a drug or device that is on the market.  In another critical way, however, it is the same—the investigator is responsible for the care of the patient with whom she interacts directly and the sponsor is not.  Some of the earliest clinical trial decisions we know about, Tracy v. Merrell Dow Pharms., Inc., 569 N.E.2d 875, 878-79 (Ohio 1991), comes to mind, found that the learned intermediary doctrine applies to claims involving prescription drugs received in connection with a clinical trial.  By and large, that has remained the rule, which also deters some of these cases.  The record on what information about the investigational drug and the study design is relayed to investigators tends to be extensive.

Why all of this introductory hemming and hawing?  We are discussing a clinical trial case brought by pro se plaintiff, not a lawyer.  Graw v. Eli Lilly & Co., No. 4:25 CV 1033 RWS, 2026 WL 496723 (E.D. Mo. Feb. 23, 2026).  We do not usually discuss decisions from pro se cases.  This one, however, appears to have been actively litigated by the plaintiff and, even though she got more deference from the court than a represented party would have received, the decision sets out detailed reasoning for why plaintiffs’ claims were dismissed.  We also find these cases interesting because of the public health implications.  It is good for new prescription medications—Graw is a drug case so we will stick with drugs—to be developed.  Good research is really expensive.  Discouraging it by imposing liability for injuries allegedly suffered by clinical study participants is bad for public health.  Undermining confidence in how clinical trials are conducted is also problematic because, among other things, you need compliant study participants and diligent investigators.  So, we definitely read Graw with some bias as to the merit of plaintiff’s allegations.  In basic terms, plaintiff claimed permanent injuries and massive damages from sustained vomiting while taking a drug, purportedly the study drug, while enrolled in a clinical trial to study diabetes.  (The decision’s details on these issues are sparse, but, from what we can tell, this was probably a double-blind study, so she would have known what she took.)  She claimed that personnel of Study Metrix, an entity that ran the study at her particular site, were generally not very good or responsive in her interactions with them during the roughly four months she was in the study.  By contrast, she had no interactions with the sponsor, but attempted to impose liability on it based on a number of theories predicated largely on the fact of sponsorship and the allegations about Study Metrix.  She had originally sued Study Metrix in the same case, but dismissed the non-diverse defendant to stay in federal court against the sponsor.  Then she sued Study Metrix and a doctor, presumably her own doctor or one she saw in connection with the study, in a separate case in state court.  The sponsor moved to dismiss the federal case for failure to state a claim and lack of service, although the Graw court did not reach the latter ground.

The first alleged cause of action was negligence, which boiled down to allowing plaintiff to sustain injuries (vomiting and a myriad of issues that could be related to being diabetic) while enrolled in the study.  Any negligence claim since Palsgraf requires duty, breach, injury, and proximate cause.  Plaintiff did not get past duty.  She offered “conclusory statements about [the sponsor’s] alleged oversight role” as to the allegedly negligent care provided by Study Metrix to establish a “right or duty to direct her medical care solely because she participated in one of its clinical trials.”  2026 WL 496723, *4.  The complaint lacked factual allegations to establish that Study Metrix was the sponsor’s agent largely because the sponsor did not have the right to control the actions of Study Metrix and its employees in connection with the trial.  The Graw court also rejected that the new allegations in briefing—which the Eighth Circuit requires to be considered when addressing a motion to dismiss a pro se plaintiff’s complaint—about her specific medical course created any duty for the sponsor to step in and direct her care.  Id. at *5.  The court cited two of many cases that have recognized that federal regulations on clinical trials do not impose such a duty or create a cause of action against the sponsor; the latter case happened to be one of the few that set aside Erie restrain to create a new state duty related to clinical trial sponsorship, but Graw did no such thing for Missouri law.  Graw also rejected the facile contention that plaintiff should get discovery to try to establish a duty.  “Graw is still required, however, to ‘[plead] enough fact[s] to raise a reasonable expectation that discovery will reveal evidence of [the claim or element].’”  Id. (quoting Twombly).

Next up was plaintiff’s claim for medical malpractice.  That was easy.  In addition to not having factual allegations that the sponsor ever provided medical care to plaintiff or assumed a doctor-patient relationship with her, plaintiff did not file the affidavit required for med mal plaintiffs under Missouri law.  Id. at *6.

The claim for lack of informed consent, which Missouri recognizes as a cause of action separate from med mal, took more analysis.  As with many of her allegations, plaintiff parroted the legal standard, alleging “the material risk information was withheld or minimized, impairing her ability to make informed decisions about continued participation.”  Id. at *7.  One of the holes here was that plaintiff did not “identify the specific information that was allegedly not disclosed to her prior to or during her participation in the clinical trial.”  Id.  We pause here because we have seen far too many decisions where a failure to warn claim against the manufacturer of a prescription or OTC medical product survives a motion to dismiss without any specifics of what information should have been provided beyond what was already provided.  This plaintiff surely received quite a bit of pertinent information in connection with giving her consent to participate in the trial, so she could not plausibly claim causation from inadequate consent without providing details on what was supposedly missing.  The legal responsibility for obtaining informed consent rested with Study Metrix and its medical staff, not with the sponsor, who was not shown to be the kind of “medical practitioner” on which Missouri law imposes a duty.  Id.

Plaintiff’s claims for intentional and negligent infliction of emotional distress were predicated on Study Metrix’s alleged conduct, so the lack of agency and a direct duty for the sponsor were fatal.  She also could not specify anything that the sponsor did that was “outrageous and extreme.”  Id. at *8-9. 

The product liability claims also did not get off the ground because plaintiff did not allege a design defect, a manufacturing defect, or a failure to warn.  Id. at *9.  (It is unclear if the complaint even identified the particular drug she took.)  She also abandoned the claim during the briefing.

The last and arguably most relevant asserted claim was for “violations of FDA clinical trial regulations.”  Based on a response to a prior court order, which was presumably directed to finding out if there was going to be a Buckman issue, plaintiff cabined her claim as not being based on a violation of the regulations per se, but as “referenc[ing] FDA regulations as evidence of the applicable duty of care and industry standards governing clinical trial sponsors.”  Id. at *10.  Citing the Sixth Circuit decision in Abney v. Amgen (discussed here and here), the court held that “FDA regulations governing clinical trials do not create an independent duty of care” running from sponsors to participants.  Id. at *10.   That was enough.

There is one last thing that we liked about Graw.  The dismissal of the case was with prejudice.  Plaintiff will get no chance to amend in response to a decision that rejected all seven counts she offered.  While plaintiff’s pending case against an entity and individual that arguably owed her the duties she had tried to impose on the sponsor may have played a role, it is good when groundless cases go away without multiple rounds of motions to dismiss.

Photo of Stephen McConnell

We’ve become aware that some clients are using artificial intelligence (AI) to summarize or analyze things like complaints, briefs, internal documents, or even – horror of horrors! – law firm bills.  If the client performing these tasks is an in-house lawyer, such work might be protected by the attorney client privilege or work product doctrine (note that we say “might”).  But what if the AI project was undertaken by a non-lawyer?

Judge Jed Rakoff of the Southern District of New York has some bad news for you. In United States v. Heppner, 2026 WL 436479 (S.D.N.Y. Feb. 17, 2026), he held that a criminal defendant’s written exchanges with a generative AI platform were not protected from government inspection by either the attorney-client privilege or the work product doctrine.  We specify the name of the judge because it matters. When Judge Rakoff speaks, the legal world listens.  Not long ago, we wrote occasional blogposts with the title of “There’ll Always be Posner.”  That title was inspired by The New Yorker, which often ran funny little reportorial pieces called “There’ll Always be England” documenting doings in the land of Dickens, Tolkien, and Led Zeppelin.  Some places or people are so reliably interesting or eccentric that they can always be counted on to supply good copy.  So it was with Judge Posner, a brilliant judge and writer.  His words brimmed with confidence. He eschewed footnotes.  After all, what authority was more authoritative than Posner’s own noggin?  In the taxonomy of judges, it was hard to place Posner. He did not seem a textualist or originalist so much as a very sophisticated consequentialist. In a way, he was the sort of Philosopher King that earlier legal scholars (think of Learned Hand or Herbert Wechsler) decried. When Posner retired from the bench, we lost  a splendid source for our blatherings.  (Posner’s post-retirement career has been … interesting). Similarly, Judge Rakoff is lightning smart. His rulings are often bold and his prose shimmers.  Our first encounter with him was at a CLE long ago.  Rakoff was then a partner at the Mudge Rose law firm.  The subject matter was RICO, and Rakoff was incapable of saying anything uninteresting or uninsightful.  Like Posner, Rakoff exudes originality and confidence.   Like Posner, he is delightful to read (and probably terrifying to appear before).  

Rakoff’s analysis in the Heppner case is easy to follow and hard to fault, but that does not soften the blow to the solar plexus from his ruling that answers “a question of first impression nationwide: whether, when a user communicates with a publicly available AI platform in connection with a pending criminal investigation, are the AI user’s communications protected by attorney-client privilege or the work product doctrine?”  Judge Rakoff’s answer is a resounding No.  So be careful out there. 

Why is Judge Rakoff’s answer No? Facts matter, and some of them here are particularly important in accounting for the result.  Most important is that the criminal defendant, “without any suggestion from counsel that he do so,” used the generative AI platform Claude to prepare “reports that outlined what he might argue with respect to the facts and the law” in anticipation of government charges.  The AI communications were seized from the defendant at the time of his arrest pursuant to a search warrant.  The defendant had not challenged the validity of the search warrant. Thus, in parsing Judge Rakoff’s ruling, you need to ask whether things would have turned out differently if the lawyer had told the defendant to run a specific AI inquiry, and whether a motion to suppress might have gained some traction.

Judge Rakoff held that the defendant’s communications with the AI platform were not embraced by the attorney-client privilege because:

  1. The communications were not between the defendant and his counsel. The privilege requires a “trusting human relationship,” and that was absent here. 
  2. The communications were not confidential. The written privacy policy for the Claude program used by the defendant disclosed that user inputs were being used to “train” the program. Thus, the defendant’s communications with Claude constituted a disclosure to a third party. 
  3. The defendant did not communicate with Claude for the purpose of obtaining legal advice.  Well, maybe the defendant thought that is what he was doing, but Claude itself disclaims that role. Cleverly, the government asked Claude whether it could give legal advice, and the AI platform responded that “I’m not a lawyer and I can’t provide formal legal advice or recommendations.”  Checkmate.  

Similarly, the work product doctrine did not shield the AI communications from disclosure. Those communications might have been done in anticipation of litigation, but they were not “prepared by or at the behest of counsel,” nor did they reflect the defense counsel’s strategy.  Maybe you can conjure up some tweaks to the fact scenario that could have brought the work product doctrine within range, but how risk averse are you?

Judge Rakoff rightly concludes that generative AI “presents a new frontier in the ongoing dialogue between technology and the law.”  The Heppner case adds to that ongoing dialogue, and adds to our worries about whether AI offers more peril than promise. 

Photo of Michelle Yeary

If ever there were a case that reads like a checklist for how not to plead around preemption, it’s Dunham v. Boston Scientific Corp., — F.Supp.3d–, 2026 WL 539533 (W.D. Mich. Feb. 24, 2026). Plaintiff alleged that his spinal cord stimulator—a Class III, premarket approved device—implanted to treat his chronic back pain caused him complications and injuries. He filed suit against the manufacturer alleging manufacturing defect, breach of implied warranty, failure to warn, negligence, breach of express warranty, and fraud. Id. at *1-2. Defendant moved to dismiss all counts primarily on preemption grounds. The court went claim by claim. The result? A clean sweep. Every count dismissed with prejudice.

When you are dealing with a PMA device, you are operating in the land of narrow pathways and narrower pleadings. Plaintiff here tried to navigate that terrain with broad references to federal regulations and a hope that “parallel claim” magic words would do the trick. They did not.

Plaintiff’s first move was familiar–allege violations of Current Good Manufacturing Practices (CGMPs), such as failure to implement adequate design controls, and call it a manufacturing defect. Because he invoked federal requirements, plaintiff argued his claim was not expressly preempted—it was “parallel.” But that only solves half the problem.

The complaint never identified a state-law duty that mirrored those federal obligations. At most, plaintiff generally referenced the Michigan Product Liability Act, but it does not create the type of “regulatory obligations embodied in the manufacturing practices upon which Plaintiff relies.” Id. at *6. Without that state-law hook, the claim depended entirely on federal requirements. And that is where implied preemption—courtesy of Buckman—steps in and shuts the door.

Even better (from a defense perspective), plaintiff did not address implied preemption in his brief. The court treated that silence as forfeiture. Id.  Count I: dismissed.

Next up: breach of implied warranty, again premised on the same alleged CGMP violations. Under Michigan law, implied warranty turns on “fitness” of the product. But if a jury found the device unfit, it would necessarily be saying the device was not as safe and effective as the FDA-approved model. That is a textbook example of express preemption—state law imposing requirements “different from or in addition to” federal approval. Id. at *6-7.

And even if preemption somehow failed, Michigan permits conspicuous disclaimers of implied warranties. Defendant had one. It was clear. Plaintiff did not respond to that argument either. Preempted. Disclaimed. Forfeited. Count II: dismissed.

On to warnings. First and foremost, in the PMA approval, the FDA proscribes the “precise labeling” for the device and the manufacturer cannot deviate from it. Id. 7. So, any claim that the label should have said something different is expressly preempted. But here, plaintiff also alleged that defendant should have warned the general public and the medical community, via means beyond the label, about “emerging adverse event trends” and “known failure modes.” Id. The problem? The FDA does not require that. So, just as with any purported label change, plaintiff’s theory would impose warning obligations beyond federal law. Again, that’s express preemption. Count III: dismissed.

Plaintiff’s negligence claim recycled the alleged CGMP violations. It was, in essence, the manufacturing defect claim wearing a different label. It failed for the same reasons—no viable parallel state-law duty and implied preemption under Buckman. Id. at *8-9. Same song, second verse. Count IV: dismissed.

Plaintiff’s express warranty claims fell into two categories. Those focused on safety, effectiveness, durability, and fitness were expressly preempted because, again, any finding that those warranties were breached would require a finding that the device was not safe and effective—contrary to the FDA’s PMA approval. Id. at *9. On the other hand, an express warranty claim premised on allegedly voluntary and false warranties that go beyond the federally-approved label can sometimes escape preemption. But here, those claims ran into a different wall: privity. Michigan requires it, plaintiff did not plead it. No alleged contractual relationship between plaintiff and manufacturer. And once again, plaintiff did not address the issue in his briefing. Id. at *10. Preempted in part. Barred for lack of privity in the rest. And forfeited besides. Count V: dismissed.

Finally, plaintiff claimed fraud based on allegedly false misrepresentations or omissions concerning the “safety, efficacy, and regulatory compliance” of the device. Id.at *10. In response to defendant’s express preemption challenge, plaintiff argued that his claim was based on allegations that the device was unsafe “despite the PMA approval.” Id. at *11. That theory sounds suspiciously like the FDA was wrong, and a state jury should say so. That is precisely the type of claim the Supreme Court rejected in Buckman. Fraud-on-the-FDA claims are impliedly preempted because they exist solely by virtue of the FDCA’s regulatory scheme. The court was unpersuaded by plaintiff’s attempt to recast the claim as independent of federal law. If your theory depends on the FDA having erred in granting approval, you are squarely in preemption territory. Count VI: dismissed.

Plaintiff did not seek leave to amend, so the court dismissed the entire case with prejudice. Id.

For many of our posts we try to come up with a clever take, a play on words, a theme. Sometimes there is no grand theme. Sometimes the theme is just this: preemption still works. And when plaintiffs ignore half the doctrine—or half the briefing—the results can be swift and decisive. Defense checklist complete. Case closed.

Photo of Bexis

We added a new site to our blogroll recently – “AI Hallucination Cases,” which describes itself as:

This database tracks legal decisions in cases where generative AI produced hallucinated content – typically fake citations, but also other types of AI-generated arguments. . . .  While seeking to be exhaustive (972 cases identified so far), it is a work in progress and will expand as new examples emerge. . . .

Of the (as of this writing) 972 judicial decisions the site identifies, over two-thirds (676) are from the United States.  Not surprisingly, more than half of the cases involve pro se litigants, but in almost 400 (391) of the listed decisions the perpetrators of AI hallucinations were actual lawyers, who, by definition, are supposed to know better.

Unfortunately, the problem does not seem to be getting any better.  The number of judicial opinions involving actual lawyers (non-pro se) accused of passing off AI hallucinations in filed documents for the incomplete month of February 2026 (it’s 2/23/26 as we write this post) is thirty-three (31 of which were in the USA), which is significantly more than a decision a day.  For January 2026, the number is thirty-six (25 of which were in the USA).  For December 2025, the number was fifty-one (36 of which were in the USA).  We didn’t do any detailed counts before then, but anyone who wants to can go to the site and verify our numbers.

A lawyer’s use of AI hallucinated anything in a filed document is unethical.  It’s a fraud on the court and opposing parties.  It’s lack of candor with the court.  It’s unprofessional since at minimum the presence of such hallucinations demonstrates incompetent use of technology.  See ABA Model Rules of Prof. Conduct, Rule 1.1.  In federal court, it’s an open-and-shut violation of Fed. R. Civ. P. 11 for not reading what was cited in signed, filed papers.  The Fifth Circuit, in a recent published opinion, held that “submitting a brief riddled with fabricated quotations” is conduct “unbecoming a member of the bar” that violates F.R.A.P. 46(c).  Fletcher v. Experian Information Solutions, Inc., ___ F.4th ___, 2026 WL 456842, at *2, 6 (5th Cir. Feb. 18, 2026) (citing, inter alia, the same website we mentioned above).

Moreover, it’s stupid as hell.  There’s nothing easier for the other side to identify than legal citations and quotations that don’t in fact exist.  Can any lawyer reasonably believe that the other side isn’t going to review the cases that s/he cites?  There are even AI programs available to detect AI hallucinations, including in legal filings.  Once hallucinations are detected, the party responsible is essentially caught red-handed with virtually no defense.  Then the perpetrator ends up like the attorney in Fletcher, or in this recent prescription medical product liability-adjacent decision:

Plaintiff has submitted dozens of inaccurate factual and legal citations across at least five different filings.  Regardless of any use of generative artificial intelligence, erroneous citations suggest Plaintiff’s counsel submitted these filings without “conduct[ing] a reasonable inquiry into the facts and the law” as required by Rule 11(b).  When confronted with similar situations, courts have ordered the filing attorneys to show cause why sanctions or discipline should not issue.  The Court finds such an order is appropriate here.

Ledoux v. Outliers, Inc., 2026 WL 291023, at *3 (W.D. Wash. Feb. 4, 2026) (citations and quotation marks omitted).

Something needs to be done about this.  We had hoped that embarrassment stemming from the broad public disclosure of the initial instances of attorneys found responsible for AI hallucinations would have been enough of a deterrent.  But our hope in ridicule as a deterrent appears to be misplaced.  Nor do the minor fines (a $2500 sanction in Fletcher, for instance) and remedial legal education requirements (the website lists sanctions, where available) that have been imposed appear to have had much effect.  Stronger medicine is needed.  This is rapidly becoming a major problem for the legal profession.  The current wave of lawyers using AI to make up things that don’t exist exposes the profession to widespread derision and worse.

Reluctantly, we don’t believe that the temptation to use AI to cut legal corners will become resistible unless and until counsel guilty of such behavior suffer significant professional consequences.  After due prior publicity – bar associations should punish attorneys caught with their hand in the AI cookie jar with suspensions from the practice of law.  We think that a first offense should warrant a suspension of one week for each fraudulent citation, quotation, or misrepresentation of a judicial opinion.  Second offenses should be punished at the rate of a month apiece.  If that’s not a sufficient deterrent, then the licensed miscreant is probably not fit to practice law.

That’s our modest proposal to keep AI hallucinations from cannibalizing the legal profession.

Photo of Bexis

To all in-house counsel working in the pharmaceutical, medical device, biotech, and digital health industries: if you can use a few CLE hours, we invite you to attend Reed Smith’s annual Virtual Health Care Week, taking place from March 16 through March 19, 2026.

This four-day all-remote event will explore recent developments, challenges, and opportunities facing U.S. and multinational health care providers, life sciences companies, and industry investors.

The presentations will also be available for viewing on demand for registered attendees. However, CLE credit for on-demand viewing will only be available for those licensed in California, Connecticut, Illinois, New Jersey, New York, Pennsylvania, Texas and West Virginia.

Here are descriptions of the topics and the registration link.

Regulatory Investigations in the Pharmaceutical Industry | Presented by Natasha Tardif, Jieni Ji, and Jeff Layne | This session will discuss how companies can limit their risk of becoming subject to regulatory investigations and how they can best prepare for when such investigations do take place.

Fraud and Abuse Enforcement in Health Care: Evolving Government Priorities and High Risk Areas in 2026 | Presented by Brian Bewley, Michelle Mantine, Katie Dunn, and David Bender | This session will provide an in-depth look at the DOJ, HHS-OIG, CMS, and SEC enforcement priorities for 2026—including the government’s deployment of AI-driven detection tools and novel applications of FCA liability—while providing practical guidance on compliance vulnerabilities across provider, manufacturer, and life sciences organizations.

Government Scrutiny of Gender Affirming Care: Risks and Considerations for Health Care Providers and Life Sciences Companies | Presented by Lesley Reynolds, Sarah Cummings Stewart, and Kristin Parker | This session will assess recent legal, regulatory, enforcement, and litigation developments relating to gender-affirming care in both the pediatric and adult contexts, including the steps taken in recent months to curtail transgender care for minors, even in states where it remains legal. 

Legal Considerations for Users and Suppliers of Health-Specific AI Model Ecosystems | Presented by Wendell Bartnick, Vicki Tankle, and Sung Park | This session will address how tech suppliers are offering scalable and flexible AI ecosystems targeted to health care and digital health companies, as well as sources and ways to mitigate regulatory and commercial risk from FDA, privacy, intellectual property, and other perspectives.

A Rise in Challenges at the NAD: A practical guide to how to keep your competitors honest in a predictable, cost-effective forum | Presented by John Feldman, Sarah Thompson Schick, and Julia Solomon Ensor | This session will explore how the National Advertising Division (NAD) process applies to pharmaceutical and medical device companies, examine recent cases and the types of claims at issue, and highlight strategies companies have used to effectively compete on a level playing field.

PBM Litigation Trends: Where Are We and Where Are We Going | Presented by Selina Coleman, Nick Rodriguez, and Megan Ali | This session will update pending litigation brought by the FTC, states, pharmacies, and putative classes against pharmacies, providers, payors, pharmaceutical manufacturers, and other interested parties that challenges PBM practices. 

Health Care Policy at an Inflection Point | Presented by Scot Hasselman and Matt Loughran | This conversational session will look ahead to cover federal priorities, constraints, and scenarios for 2026.

CLE Information for U.S.-licensed attorneys:

The session “Fraud and Abuse Enforcement in Health Care: Evolving Government Priorities and High Risk Areas in 2026” is presumptively approved for 1.5 CLE credit in California, Connecticut, Illinois, New Jersey, New York, Pennsylvania, Texas, and West Virginia. Applications for CLE credit are being filed in Colorado, Delaware, Florida, Georgia, Ohio, and Virginia. Attendees who are licensed in other jurisdictions will receive a uniform certificate of attendance, but Reed Smith only provides credit for the states listed.

All other sessions are presumptively approved for 1.0 CLE credit in California, Connecticut, Illinois, New Jersey, New York, Pennsylvania, Texas, and West Virginia. Applications for CLE credit will be filed in Colorado, Delaware, Florida, Georgia, Ohio, and Virginia. Attendees who are licensed in other jurisdictions will receive a uniform certificate of attendance, but Reed Smith only provides credit for the states listed.

Please allow 4-6 weeks after the program to receive a certificate of attendance.

CLE Questions? Contact Learning & Development CLE Attendance

Photo of Eric Hudson

We’ve blogged about aspects of the Model Drug Dealer Liability Act before. Over 20 states have enacted versions of the model act in an effort to shift liability for harm caused by illegal drugs to the drug dealers and traffickers who sell and promote illegal drug use.  Those are not normally the types of claims of interest to our readers. But, given that we’ve been blogging this week about fraudulent and harmful medical procedures and junk science in mass torts, maybe it’s no surprise that today’s decision deals with an attempt to use a version of the model act as a  vehicle for finding liability against a major distributor of legitimate, FDA-approved drugs. McKesson Corp. v. Bolton, 2026 WL 394505 (Ga. App. Feb. 12, 2026).

Continue Reading A Drug Dealer? Seriously?
Photo of Lisa Baird

It must be bad plaintiff lawyer week here at the Drug and Device Law Blog.  Yesterday, we brought you a review of The Pain Brokers (One Signal Publishers 2026) and the sordid tale of ginned-up damages in the vaginal mesh litigation. 

Today we bring you The American Tort Reform Association’s 2026 report, The Junk Science Playbook: Exposing Mass Tort Litigation Tactics and its explanation about how mass tort plaintiffs’ lawyers do not simply “find” science; in many instances, they help manufacture it out of whole cloth. 

ATRA’s Report maps out a coordinated ecosystem that produces, launders, and weaponizes questionable scientific claims to drive mass torts and regulatory pressure.  That ecosystem includes those who create junk science (“merchants”) and those who launder and amplify it (“purveyors”).  The Report also points toward solutions—most notably, rigorous Rule 702 gatekeeping—to keep junk science out of court.

Those who don’t have to wade through our neck of the litigation swamp everyday just do not appreciate how bad things are, but this Report has some of the receipts.

The Merchants.

The report identifies “quasi‑academic” organizations whose classifications and studies repeatedly become centerpieces of mass tort advertising and expert reports, yet are regularly at odds with mainstream regulators.  Like the IARC, which only found one agent “probably not” carcinogenic out of nearly 1,000 examined, and which relies on working‑group members who not infrequently have conflicts of interest (like advisory ties with advocacy groups opposed to targeted chemicals). 

Then there are litigation‑aligned testing laboratories like Valisure, which Blog readers know is key to generating what the Report (understandably) labels “faux causation” data in litigations ranging from sunscreen to acetaminophen.  Or Emery Pharma, which markets its ability to design studies for product‑liability cases and then also supply the testifying plaintiff-side experts. 

And of course we have the agenda‑driven nonprofits dedicated to producing and promoting litigation‑friendly science, often funded by money from the plaintiffs’ bar that their work supports. 

The Purveyors.

Once the litigation‑driven “science” exists, there is a laundering pipeline to burnish it and clothe it with apparent legitimacy.  

To begin with, there is the scientific publishing industry, which is suffering an integrity crisis.  Recent years have shown a massive spike in retractions (over 10,000 papers in 2023 alone) and what the Report calls (again, with justification) the “predatory” open‑access journals that have impressive titles, but are pay-to-publish and without meaningful peer review. 

Next, nothing catches eyeballs or clicks like a good scary story, so after “scientific” journal publication, advocacy groups and their PR teams go into overdrive to convert niche papers and negligible results into mainstream health scares, with sometimes disappointingly credulous coverage by major news outlets.

As the media let loose, so does the lawyer advertising and the mass tort litigation.  So the “final step” in the playbook is the conversion of questionable research into courtroom testimony by frequent-flier plaintiffs’ experts.  These experts often may disregard contrary, mainstream science; manipulate or cherry-pick their data, and obscure their financial ties or litigation roles.  Seriously, some of these plaintiffs’ experts receive millions in fees.

Public Health.

We complain about the “litigation industry” regularly on this Blog, and litigation abuses are a problem in their own right.  But what really gets us is when those abuses spill over into hurting people, like the vaginal mesh plaintiffs misled into explant surgeries they didn’t need and their own doctors did not recommend.

The ATRA report identifies another space where litigation incentives are reshaping “science” in a way that will hurt people:  Vaccines.

The plaintiffs’ bar generally does not like the National Vaccine Injury Compensation Program (VICP), a no‑fault program funded by an excise tax on vaccines) that has paid more than $5.5 billion in compensation since 1986.  The VICP requires most vaccine‑injury claims to be brought administratively rather than via tort suits.

Plaintiffs’ lawyers do not rake in as much dough under the VICP as they do through the usual mass torts, so the Report identifies efforts to destabilize the program.  One such effort is by aggressively pushing to add autism as a compensable condition without scientific support, or stripping vaccines from the program in a way that could either bankrupt the fund or route claims back to the tort system, where plaintiffs’ lawyers stand to collect substantial contingency fees.  Another is by pressing for state‑law private rights of action for vaccine side effects (e.g., Texas’ 2025 law, with copycat bills in other states), which the advocates hope will punch holes in the federal preemption that applies to vaccines.  

Anything that uses bad science to mislead the public about vaccines will, inevitably, lead to more illnesses and more deaths from preventable diseases.

Solutions.

The ATRA Report closes with a suite of recommended reforms, most prominently that vigorous judicial gatekeeping based on Federal Rule of Evidence 702 (and state law equivalents) should be used to keep junk science out of the courtroom.  The 2023 Rule 702 amendments and post-amendment case law emphasize that:

  • The proponent bears the burden of proving admissibility by a preponderance of the evidence; there is no presumption of admission.
  • Courts must scrutinize not only an expert’s methodology and data, but also whether each opinion stays within the bounds of what that methodology can reliably support.
  • Reliability in both theory and application is the “hallmark” of admissible expert testimony.
  • Courts must probe more deeply than before; many prior decisions misapplied Rule 702 by re‑labeling reliability as going to “weight” rather than admissibility.
  • Red flags include causation theories that have not been independently validated in peer‑reviewed literature outside of litigation.

The Report also urges more courts not to take either side’s word for it and instead obtain their own, mainstream, credible, insight into the science by using Rule 706 to appoint neutral experts or technical advisors, or convening “science days” in complex MDLs, and requiring disclosure of litigation funding and lawyer involvement for studies cited in expert reports or regulatory proceedings.

Finally, the Report urges defendants to actively counter junk science rather than merely defend individual cases through aggressive discovery into communications and funding records to expose lawyer‑driven study designs or incomplete disclosures or—again, from yesterday’s post—outright illegal and nefarious schemes.

Photo of Stephen McConnell

In truth, this post is not so much a review as a full-throated endorsement.  We are telling colleagues that they need to read Elizabeth Chamblee Burch’s The Pain Brokers (One Signal Publishers 2026). Two things drive our enthusiasm for this book: cynicism and pride. 

First, let’s indulge our cynicism. The Drug and Device Law engages in no pretense of objectivity.  We are entrenched on the defense side and make no bones about it.  We represent product liability defendants and root for them to win even when we have no involvement in the case.  And we are sore losers.  We criticize what we consider to be bad decisions.  Sometimes we kvetch. Sometimes we find ourselves suspecting the other side of cheating.  

At this point we find ourselves making a callback to some recent posts. We quoted Abe Greenwald, an editor of Commentary, who often says that however bad you think things are, they are even worse. And in our post on the disqualification of a plaintiff firm in a talc case because that firm had hired a lawyer who formerly worked for the talc defendant, we grumbled that it was likely the tip of a rotten iceberg.  (Earlier this week the court denied the plaintiff firm’s request to pause the disqualification.) 

It is grindingly obvious that the problem with mass torts is that they are … massive. They are hard to handle. Moreover, mass torts offer opportunities for mass frauds.  For example, too many cases parked in Multidistrict Litigations never really get tested.  Product usage is sometimes faked, and injuries can be exaggerated to get a higher spot on the settlement grid.  

With little fondness, we remember the fen-phen litigation, where traveling echocardiogram doctors inflated the degree of heart valve regurgitation to gin up higher settlements.  Some doctors got in trouble.  But for every phony echocardiogram that was spotted, we bet another ten eluded detection. 

It’s not paranoia if they really are out to get you. 

The subtitle of The Pain Brokers is “How con men, call centers, and rogue doctors fuel America’s lawsuit factory.”  The particular lawsuit factory covered by the book was the vaginal mesh MDL in West Virginia. It was an enormous MDL, with several defendants, thousands of plaintiffs, some bellwether and remand trials, and large inventory settlements.  It ended up being a mass wealth transfer, crafted by a kakistocracy. As the book tells us, “[b]y 2014, pelvic mesh had become the holy grail of mass torts.”

The worse the injury, the higher the settlement would be.  That is the theory, anyway.  A plaintiff who had such serious problems (pain, infection, extrusion) with her mesh that she had to have it removed would be entitled to more money than someone who was living with the mesh intact.  It stands to reason, doesn’t it?

Enter a large cast of litigation villains with a scheme. Some were veterans of the wrong side of the criminal justice system. They (the “con men” in the subtitle) exploited what was then a back door unique to the District of Columbia, which permitted nonlawyers to have an ownership role in law firms.  The aim was to create mesh cases with high values. The intake procedure started with call centers armed with information about a woman’s mesh implantation.  (How they got that information is doubtless an ugly story.) They cold-called women and did everything possible to scare them into removing the mesh.  Following a script, the call centers told the women that the mesh was a ticking time bomb.  Some women thought the calls came from the mesh companies. Apparently nothing was done to disabuse the women of that mistake.  

The next step in the scheme was to connect the women to surgical centers that would arrange to fly the women (often across the country, usually to Florida) to undergo surgeries to “revise” or explant their mesh.  The women were told that they could not use medical insurance.  Instead, they were charged considerably more than going through insurance would have cost them. The women were sent agreements via Docusign, which permitted jumping through the documents to get right to the signature and initial pages. That jumping feature in Docusign can be convenient when you know what you are signing, but awful if you do not. Women ended up taking out high interest loans without even knowing it.  Their surgeries were performed with minimal medical consultation. There really were “rogue doctors” as stated in the subtitle. The women’s lawyers also seemed to offer them little legal consultation. (That fact had consequences for claims of attorney-client privilege.) Having their mesh removed, the women then entered the mesh litigation machine, lined up for settlement, and the subtitle’s “con men, call centers, and rogue doctors” took their cut. After that cut, some plaintiffs ended up with little money. 

Pelvic mesh plaintiff lawyers liked to call their clients victims. Some were. Some were victims because they had mesh removed from them that was performing perfectly well.  Their own, real treating doctors expressed surprise and disappointment that the patients had the pelvic mesh removed.  The problems (such as incontinence or prolapse) the women had that prompted the implants often returned. Often there was post-explant pain. Even if the mesh removal led to a higher settlement amount, the women were not getting much of whatever upside existed.  Instead, their health was harmed.  Also victimized were the defendants and the court system.  It is a depressing tale. 

Second, we arrive at our point of pride.  One of the heroes in the book is a former colleague, Barbara Binis.  She retired from Reed Smith a couple of years ago. She was a splendid trial lawyer and a master of litigation strategy. She had a way of getting after the truth. Her work unearthing the vaginal mesh fraud required sleuthing, persistence, and ingenuity.  Binis tracked down various players in the litigation scheme and took their depositions.  Some of those deponents were more than a little slippery, and some were defended by lawyers who took obstruction to new depths. Binis managed to expose the sleazy reality of the scheme. The wonder of it is that only one doctor was incarcerated.  Unfortunately, due to the statute of limitations, certain rigidities in the system, and some soft-hearted judges, most of the miscreants skated. 

But if the book might induce despair at the bent litigation lottery system (at one point one of the schemers mused over how to make a profit off of earplug, PFAS, and Camp LeJeune litigation), it also shows what a smart, determined lawyer can do to get to the truth.  The book’s documentation of Binis’s dogged pursuit of the schemers is inspiring.  Defense lawyers do not always have to play defense. If there is an opportunity for a defense lawyer to seize the initiative, do it. Meanwhile, the book also introduces us to a plaintiff lawyer who did his best to help some of the victimized women seek redress. It was an uphill run. No law school teaches the skills and strategies needed to nail fraud artists who manipulate the civil justice system.  

We want all our colleagues to read the book to know what is really going on in mass torts.  That knowledge will not merely feed cynicism; it will energize efforts to combat the schemers.  We also want judges and legislators to read the book.  Maybe understanding how the mass tort system lends itself to abuse might prompt decision-makers to try to stem such abuse. 

Meanwhile, some other jurisdictions opened up the same back door letting nonlawyers operate law firms.  Perhaps there were good intentions behind that move, but The Pain Brokers constitutes a cautionary tale. 

Photo of Michelle Yeary

This post is from the non-Butler Snow side of the blog.

Back in November, we told you the court in the In re Paragard IUD Products Liability Litigation wasn’t quite finished with plaintiffs’ effort to impose a post-sale duty to warn on a company that didn’t manufacture, sell, or even own the product when it was implanted. The court had dismissed the design defect claims against the successor company but requested supplemental briefing on causation for failure to warn. We suggested then that, absent wormhole technology, causation was going to be a problem. We recently came upon the warning ruling. Big surprise—summary judgment granted on causation.

As a reminder, the product was originally manufactured and sold by Teva. But in November 2017, the NDA was purchased by Coopersurgical. Coopersurgical moved for summary judgement in cases where plaintiffs’ IUDs were implanted prior to 2017, so manufactured and sold by Teva, but explanted after the change of ownership—at which time it was discovered the IUDs had broken. In re Paragard IUD Products Liability Litigation, 2025 WL 4092375, at *2 (N.D. Ga. Dec. 23, 2025).  

The claims at issue were brought under Florida law, so the court’s analysis started with recognizing that Florida is a learned intermediary state. That means the duty to warn runs to the prescribing physician, not directly to the patient. Plaintiffs attempted to sidestep that rule by arguing for a contraceptive exception—claiming that birth control products are uniquely marketed directly to patients and that physicians defer to patient choice. Id. at *3.  There is just one problem—Florida has not adopted a direct-to-consumer exception to the learned intermediary doctrine. In fact, two Florida courts have expressly rejected the concept. Plaintiffs only cited a Massachusetts case to support their argument. That may be interesting in Boston, but it is not binding in Florida.

So the learned intermediary rule applied. Which meant plaintiffs had to prove that a different warning to the physician would have changed something in a way that would have prevented their injuries.

Plaintiffs argued that by the time the successor company acquired the NDA, before the IUDs were removed, it knew the label did not adequately warn about device breakage. Therefore, say plaintiffs, defendant could and should have provided new warnings, before the IUDs were removed, via sales representatives, literature, dear doctor letters, etc. Then, say plaintiffs, their physicians would have reviewed the warnings, monitored for breakage, warned plaintiffs, attempted to mitigate the risk, and/or recommended early removal. Id. at *4.

But litigation does not run on plaintiffs say so. As it turns out, “[p]laintiffs’ allegations [were] not grounded in any evidence.” Id. Plaintiffs claimed that one of their physicians testified that had she known of the breakage risk she would have “been on the lookout” for it. However, the court actually read the testimony and concluded it did not say what plaintiffs claimed it said. Id. The doctor actually testified there was no resistance during her first attempt to remove the device. Plaintiffs’ spin was more aspirational than record citation. The actual testimony did not support finding causation. Id.

Plaintiffs offered no other factual support for their allegations. There was no testimony that a different warning would have caused a reasonable physician to notify the already-implanted patients of a breakage risk. No testimony that any alternative course of action would have been taken. No testimony that anything would have prevented these injuries. There was no evidence from which a reasonable jury could find that the defendant’s alleged failure to warn was the proximate cause of their injuries.

While the court did not wade into a broad rejection of post-sale successor duty to warn, it didn’t need to. Even assuming arguendo some duty could exist, plaintiffs still had to prove causation. And they couldn’t. Proof of causation requires actual evidence. Not speculation about what doctors might have done differently. On that point, the record was silent.

Drug & Device Resources

Bexis’ Book

FDA-regulated products now account for an estimated one-fifth of overall economic activity in the U.S. They have also been the focus of a litigation explosion. This timely guide covers all aspects of litigation involving drugs, medical devices, vaccines and other FDA-regulated prescription products.
Learn more about this book