PRECEDENTIAL
UNITED STATES COURT OF APPEALS FOR THE THIRD CIRCUIT
Nos. 24-2990 and 24-3198
ADOLPH MICHELIN
v.
WARDEN MOSHANNON VALLEY CORRECTIONAL CENTER; DIRECTOR PHILADELPHIA FIELD OFFICE IMMIGRATION AND CUSTOMS ENFORCEMENT; DIRECTOR UNITED STATES IMMIGRATION AND CUSTOMS ENFORCEMENT; SECRETARY UNITED STATES DEPARTMENT OF HOMELAND SECURITY; ATTORNEY GENERAL UNITED STATES OF AMERICA,
Appellants in case 24-2990
ADEWUMI ABIOYE
WARDEN MOSHANNON VALLEY PROCESSING CENTER; ACTING FIELD OFFICE DIRECTOR OF THE IMMIGRATION and CUSTOMS ENFORCEMENT and REMOVAL OPERATIONS PHILADELPHIA FIELD OFFICE; SECRETARY UNITED STATES DEPARTMENT OF HOMELAND SECURITY; ATTORNEY GENERAL UNITED STATES, Appellants in case 24-3198
Appeal from the United States District Court for the Western District of Pennsylvania (District Court Nos. 3:23-cv-00022 and 3:23-cv-00251) Magistrate Judges: Honorable Patricia L. Dodge and Honorable Christopher B. Brown
Argued on November 10, 2025 Before: RESTREPO, McKEE, and AMBRO, Circuit Judges (Opinion filed: February 2, 2026) Laura S. Irwin Office of United States Attorney 700 Grant Street Suite 4000 Pittsburgh, PA 15219
Joseph A. McCarter [Argued] Sarah S. Wilson United States Department of Justice Office of Immigration Litigation P.O. Box 878 Ben Franklin Station Washington, DC 20044
2 Counsel for Appellants Warden Moshannon Valley Correctional Center, Directory Philadelphia Field Office Immigration and Customs Enforcement, Director United States Immigration and Customs Enforcement, Secretary United States Department of Homeland Security, Attorney General United States of America
Casey Corcoran Robert S. Jones Andrew Tutt Kathleen Weng Arnold & Porter Kaye Scholer 601 Massachusetts Avenue NW Suite 1121 Washington, DC 20001
Jonah B. Eaton Nationalities Service Center 1216 Arch Street 4th Floor Philadelphia, PA 19107
Nicole L. Masiello William Sharon [Argued] Arnold & Porter Kaye Scholer 250 W 55th Street New York, NY 10019
Counsel for Appellees Adolph Michelin and Adewumi Abioye
3 OPINION OF THE COURT
AMBRO, Circuit Judge
A petition for a writ of habeas corpus has been a civil action since before our law was our law. In the mid-eighteenth century, just before our Founding, William Blackstone observed that the law we were about to inherit had two “primary objects”: “the establishment of rights, and the prohibition of wrongs.” 3 William Blackstone, Commentaries on the Laws of England 1 (1772). There were two types of wrongs: “private wrongs, and public wrongs.” Id. at 2 (emphasis omitted). Public wrongs, he explained, “are a breach and violation of public rights and duties, which affect the whole community considered as a community; and are distinguished by the harsher appellation of crimes and misdemeanors.” Id. (emphasis omitted). Private wrongs, by contrast, “are an infringement or privation of the private or civil rights belonging to individuals, considered as individuals; and are thereupon frequently termed civil injuries.” Id. (emphasis omitted). For public wrongs, the remedy came by a criminal action. For private wrongs, the remedy required the “application to . . . courts of justice . . . by civil suit or action.” Id. at 3.
One civil right, secured by civil action, was that of “personal liberty.” Id. at 119. When this right was violated, the prisoner could petition a court to “remov[e] the injury of unjust and illegal confinement,” id. at 137 (emphasis omitted), by
4 “commanding” whomever held him in custody “to produce the body” and explain the “cause of his caption and detention,” id. at 131. And so, in listing the civil actions available for “the redress of private wrongs,” Blackstone lauded “the most celebrated writ in the English law,” id. at 129—the “great and efficacious writ” that would restore the captive to liberty, id. at 131—the writ of habeas corpus.
After we overthrew the English crown, we retained this framework from English common law. See, e.g., Watson v. Mercer, 33 U.S. (8 Pet.) 88, 110 (1834) (distinguishing “civil proceedings[,] which affect private rights,” from “criminal proceedings[,] which impose punishments”); Bradlie v. Md. Ins. Co., 37 U.S. (12 Pet.) 378, 402 (1838) (defining “proceedings of a mere civil nature” as those “to enforce private rights”). We also retained the recognition that habeas is a civil action. See Holmes v. Jennison, 39 U.S. (14 Pet.) 540, 565–67 (1840) (holding “it is too plain for argument” that a petition for a writ of habeas corpus is a civil action under the Judiciary Act of 1789); Ex parte Tom Tong, 108 U.S. 556, 559– 60 (1883) (“The writ of habeas corpus is the remedy which the law gives for the enforcement of the civil right of personal liberty.”).
Much has changed over the centuries. Not this. See, e.g., Banister v. Davis, 590 U.S. 504, 507 (2020) (“Habeas proceedings, for those new to the area, are civil in nature.”).1
1 On the enduring authority of Blackstone’s account of the writ of habeas corpus, see, e.g., Brown v. Davenport, 596 U.S. 118, 128 (2022); Ex parte Bollman, 8 U.S. (4 Cranch) 75, 97 (1807).
5 Today the Equal Access to Justice Act (EAJA) provides that the prevailing private party “in any civil action (other than cases sounding in tort)” brought by or against the United States is entitled to attorneys’ fees and costs if the Government’s position was not “substantially justified” or “special circumstances make an award unjust.” 28 U.S.C. § 2412(d)(1)(A). Adolph “Lee” Michelin and Adewumi Abioye prevailed in habeas actions under 28 U.S.C. § 2241 challenging their immigration detentions. The District Courts found the Government’s positions were not “substantially justified” and awarded the detainees fees and costs. The main question these consolidated cases present is whether a petition for a writ of habeas corpus from immigration detention under § 2241 is an EAJA “civil action.”
We answer yes and thus affirm.
I. BACKGROUND
We begin with the facts of the two cases.
A. Abioye
Abioye, a Nigerian citizen, entered the United States in April 2018 on a tourist visa. In July 2020, he pled guilty in the U.S. District Court for the District of Maryland to conspiracy to commit wire fraud in violation of 18 U.S.C. §§ 1343 and 1349. In May 2022—the day he completed his sentence— federal Immigration and Customs Enforcement (ICE) agents took him into immigration detention at the Moshannon Valley
6 Processing Center.2 An immigration judge ordered him removed to Nigeria. The next year, the Board of Immigration Appeals (BIA) affirmed. In June 2023, Abioye petitioned for review and a stay of removal from the U.S. Court of Appeals for the Fourth Circuit. It stayed removal pending review of his petition.
In October 2023, after over 16 months in immigration detention, Abioye petitioned the U.S. District Court for the Western District of Pennsylvania for a writ of habeas corpus under § 2241 challenging his detention without an individualized bond hearing as a violation of the Fifth Amendment’s Due Process Clause. The Government opposed his petition. The next month, the District Court granted the writ and ordered a bond hearing. In December an immigration judge released him on a $5,000 bond.
Abioye moved for attorneys’ fees and costs under § 2412(d)(1)(A) of the EAJA. The District Court found the Government’s position was not substantially justified and awarded him $18,224.58.
B. Michelin
ICE agents arrested Michelin, a Jamaican citizen, in January 2022 and detained him at Moshannon. A year later, he was still detained—and had no bond hearing. Although he had petitioned the BIA to reopen his immigration case, it had not responded. So he too petitioned for a writ of habeas corpus
2 From March to August 2023, Abioye was held at the Pike County Correctional Facility; otherwise, he was held at Moshannon.
7 under § 2241 from the U.S. District Court for the Western District of Pennsylvania. It granted his petition, holding his prolonged detention without an individualized bond hearing violated the Fifth Amendment’s Due Process Clause. An immigration judge released Michelin on a $10,000 bond, and he rejoined his family in Philadelphia.
Michelin sought an award of attorneys’ fees and costs per the EAJA. The District Court found the Government’s position not substantially justified and awarded him $15,841.60.
* * *
The United States appealed both decisions. We consolidated the appeals.
II. JURISDICTION AND STANDARDS OF REVIEW
We have appellate jurisdiction to review fee awards under 28 U.S.C. § 1291. Whether we have subject matter jurisdiction depends on the question at the heart of this appeal: whether the EAJA waives federal sovereign immunity to fee awards for habeas actions challenging civil immigration detention. Our review of legal questions is plenary. Newmark v. Principi, 283 F.3d 172, 174 (3d Cir. 2002). We review for abuse of discretion a district court’s finding that the Government’s position was not substantially justified. Cruz v. Comm’r of Soc. Sec., 630 F.3d 321, 324 (3d Cir. 2010).
III. DISCUSSION
The EAJA provides:
8 [A] court shall award to a prevailing party other than the United States fees and other expenses . . . incurred by that party in any civil action (other than cases sounding in tort) . . . brought by or against the United States . . . unless the court finds that the position of the United States was substantially justified or that special circumstances make an award unjust.
28 U.S.C. § 2412(d)(1)(A). These appeals concern whether a habeas challenge to immigration detention pursuant to § 2241 is a “civil action” under the EAJA. It is. The Abioye appeal also asks whether the Government’s position in that case was substantially justified. It was not.
A. The phrase “any civil action (other than cases sounding tort)” in the EAJA unambiguously encompasses habeas challenges to immigration detention under § 2241.
The “cardinal canon” of statutory interpretation is the “presumption that a legislature says in a statute what it means and means in a statute what it says there.” Conn. Nat’l Bank v. Germain, 503 U.S. 249, 253–54 (1992) (cleaned up). On Abioye’s and Michelin’s reading, the EAJA means what it says: save torts, it covers all civil actions; hence, habeas is included. To the Government, it does not: the statute could mean only “purely,” “wholly,” or “garden-variety” civil actions, Abioye Opening Br. 25; habeas is excluded.
We agree with Abioye and Michelin. Habeas actions are civil actions. And the complete phrase—“any civil action (other than cases sounding in tort)”—clearly encompasses them. We grant the Government this: in some instances, the
9 phrase “civil action” does not cover habeas actions. But those situations are the exceptions, not the rule. And the EAJA cannot plausibly be read that way.
1. The sovereign immunity canon applies only if a statute is ambiguous after deploying the ordinary tools of statutory interpretation.
Before we go further, we pause to clarify what we are looking for where, as here, a statute implicates the sovereign immunity of the federal Government.
Generally, the United States enjoys sovereign immunity from awards of monetary relief. FDIC v. Meyer, 510 U.S. 471, 475 (1994). However, the Government may waive sovereign immunity by statute. Kirtz v. Trans Union LLC, 46 F.4th 159, 164 (3d Cir. 2022). “[A] waiver of sovereign immunity must be ‘unequivocally expressed’ in statutory text.” FAA v. Cooper, 566 U.S. 284, 290 (2012) (quoting Lane v. Peña, 518 U.S. 187, 192 (1996)). Any ambiguity is “to be construed in favor of immunity,” id. (citing United States v. Williams, 514 U.S. 527, 531 (1995)), whether it concerns the existence of a waiver or the scope of one, id. at 291. The EAJA’s fee award provision is a statutory waiver of sovereign immunity. Ardestani v. INS, 502 U.S. 129, 137 (1991). Accordingly, if the phrase “any civil action” is ambiguous between an interpretation that reaches habeas actions and one that does not, then we must construe the scope of the waiver to exclude them.
Still, “the clarity of each statute must be evaluated on its own terms.” Dep’t of Agric. Rural Dev. Rural Hous. Serv. v. Kirtz, 601 U.S. 42, 52 (2024) (cleaned up). What matters is whether the statute is ambiguous after exhausting the
10 traditional tools of statutory interpretation, not before. United States v. Rutherford, 120 F.4th 360, 380 n.28 (3d Cir. 2024). Cf. Sebelius v. Cloer, 569 U.S. 369, 376–81 (2013) (holding provision unambiguous and declining to apply sovereign immunity canon after considering text, structure, and purpose).3 If there is “no ambiguity left,” “[t]here is no need for us to resort to the sovereign immunity canon.” Richlin Sec. Serv. Co. v. Chertoff, 553 U.S. 571, 590 (2008). Only if, after exhausting these tools, “there is a plausible interpretation of the statute that would not authorize money damages against the Government” do we “take the interpretation most favorable” to it. Cooper, 566 U.S. at 290–91.
2. Habeas challenges to immigration detention under 28 U.S.C. § 2241 are civil actions.
“When interpreting a statute, we begin with the text.” Lackey v. Stinnie, 604 U.S. 192, 199 (2025). To repeat, the EAJA permits awards of fees and costs in “any civil action
3 We leave two tools—purpose and legislative history—in the toolbox. The Supreme Court has admonished that “no amount of legislative history can supply a waiver that is not clearly evident from the language of the statute.” Kirtz, 601 U.S. at 49 (cleaned up). And we take its directive to focus on “the language of the statute” in discerning the existence and scope of a waiver to caution against considering purpose as well. Id.; see also Ardestani, 502 U.S. at 138 (finding that “the broad purposes of the EAJA would be served by making the statute applicable to deportation proceedings,” but declining to “extend the EAJA to [them] when the plain language of the statute, coupled with the strict construction of waivers of sovereign immunity, constrain us to do otherwise”).
11 (other than cases sounding in tort).” 28 U.S.C. § 2412(d)(1)(A). The statute does not define “civil action.” In the absence of a statutory definition, we “start with the phrase’s plain meaning” in 1980, when Congress enacted it in 1980. Clean Air Council v. U.S. Steel Corp., 4 F.4th 204, 209 (3d Cir. 2021). It was “well settled” then that habeas corpus is a civil action. Browder v. Dir., Dep’t of Corr. of Ill., 434 U.S. 257, 269 (1978); see also Stafford v. Briggs, 444 U.S. 527, 543 (1980) (“Habeas corpus is a civil action.” (cleaned up)). Thus, we have acknowledged that the “plain meaning” and “literal scope” of the phrase “civil action” include habeas actions. See Santana v. United States, 98 F.3d 752, 754 (3d Cir. 1996).
The same is true when we consider “civil action” as a legal term of art. When Congress enacted the EAJA, legal dictionaries defined “civil action” on broad terms that reached habeas proceedings. See, e.g., Black’s Law Dictionary 222 (5th ed. 1979) (defining civil actions as “[a]ction[s] brought to enforce, redress, or protect private rights” or as “all types of actions other than criminal proceedings”); Radin Law Dictionary 55 (2d ed. 1970) (defining a civil action as a “legal proceeding brought to enforce a civil right or obtain redress for its violation”); Ballentine’s Law Dictionary 202 (3d ed. 1969) (the phrase “comprehend[s] every conceivable cause of action, whether legal or equitable, except such as are criminal in the usual sense”).
No wonder then that, at the time of the EAJA’s enactment, the Federal Rules of Civil Procedure treated habeas proceedings as civil actions. Rule 1 provided that the Rules applied to “all suits of a civil nature . . . with the exceptions stated in Rule 81.” Fed. R. Civ. P. 1 (1980). And Rule 81 specified the Rules “are applicable to proceedings for . . .
12 habeas corpus . . . to the extent that the practice in such proceedings is not set forth in statutes of the United States and has heretofore conformed to the practice in civil actions.” Fed. R. Civ. P. 81(a) (1980). We know of no relevant statute or practice that would have negated the civil status of immigration challenges like this one. In any event, what matters most is that the Rules generally presumed habeas actions were civil actions, even if they were also distinctive ones. Harris v. Nelson, 394 U.S. 286, 293–94 (1969).4
The common law history confirms habeas proceedings are civil actions. “When Congress uses a term with origins in the common law, we generally presume that the term ‘brings the old soil with it.’” Kousisis v. United States, 605 U.S. 114, 124 (2025) (quoting Sekhar v. United States, 570 U.S. 729, 733 (2013)); see also Lackey, 604 U.S. at 200. “Civil action” is one such term. As we described at the outset, the writ of habeas corpus has deep roots in that rich soil. See 3 Blackstone at 115– 38; see also Ex parte Yerger, 75 U.S. (8 Wall.) 85, 96 (1868)
4 Today, Rule 81 continues to provide that the Rules “apply to proceedings for habeas corpus” under almost identical conditions. See Rule 81(a)(4). That said, our holding is that a habeas petition under § 2241 challenging immigration detention is a “civil action” for purposes of the EAJA, not that such an action is a “civil action” for purposes of the Federal Rules of Civil Procedure. We have no occasion to address the latter topic. In any event, some provisions would caution against reading every use of “civil action” that way. For instance, Rule 3 provides that “[a] civil action is commenced by filing a complaint with the court,” Fed. R. Civ. P. 3, yet a habeas action begins with what the statute dubs “an application,” 28 U.S.C. § 2241. So we do not decide this issue.
13 (holding the Supreme Court had the power to issue writs of habeas corpus, one of “the most important powers in civil cases of all the highest courts of England”); Kurtz v. Moffitt, 115 U.S. 487, 494 (1885) (“A writ of habeas corpus . . . is a civil suit or proceeding, brought by him to assert the civil right of personal liberty, against those who are holding him in custody[.]”).
The Government strains to deny habeas actions are civil. It acknowledges—as it must—that the Supreme Court and our Court have recognized habeas proceedings are at least “technically” civil actions. See Abioye Opening Br. 21 (quoting Schlanger v. Seamans, 401 U.S. 487, 490 n.4 (1971)), 27 (quoting Santana, 98 F.3d at 754). Nonetheless, the Government insists habeas actions are “not wholly civil,” Abioye Opening Br. 31, but rather “unique, hybrid actions in a category of their own,” Abioye Opening Br. 11.
The “hybrid” theory raises a question: A hybrid of what and what? The Government avoids saying in its brief, because the answer exposes this argument as a dead end. To the extent habeas actions are hybrids, they are hybrids of civil actions and criminal ones. See, e.g., Santana, 98 F.3d at 754; O’Brien v. Moore, 395 F.3d 499, 505 (4th Cir. 2005); see also Abioye Opening Br. 8 (recounting that the Government told the District Court habeas actions were “hybrid criminal/civil proceedings”). The idea is that “to the extent that a habeas proceeding reviews a criminal punishment with the potential of overturning it, the habeas proceeding necessarily assumes part of the underlying case’s criminal nature.” O’Brien, 395 F.3d at 505.
But we are not reviewing habeas petitions for release from criminal detention. We are reviewing them for release
14 from immigration detention. In that context, every element is civil. See Zadvydas v. Davis, 533 U.S. 678, 690 (2001) (recognizing immigration detention is “civil detention”); INS v. Lopez-Mendoza, 468 U.S. 1032, 1038 (1984) (recognizing removal proceedings are civil). Even immigration actions subject to sui generis procedures, like removal proceedings before an immigration judge, are “purely civil.” Lopez- Mendoza, 468 U.S. at 1038. A hybrid of a civil action and a civil action is a civil action.
Resisting that conclusion, the Government offers two additional arguments that all habeas actions are in a category of their own. The first says habeas is “unique” because it seeks release from confinement. Abioye Opening Br. 31. But what settles its status is the right it seeks to vindicate, not the remedy. Habeas actions are civil because they protect the civil right to personal liberty. See 3 Blackstone at 115–38; Santana, 98 F.3d at 754 (citing Ex parte Tom Tong, 108 U.S. at 559). Second, the Government suggests habeas is not civil because it does not share every feature of archetypal civil actions. True enough. But it does not follow that habeas is not a civil action. Many civil actions—including ones expressly within the scope of the EAJA’s “civil action” provision—have unique rules of practice or procedure, like veterans’ benefits actions. See Henderson v. Shinseki, 562 U.S. 428, 440 (2011) (observing “[t]he contrast between ordinary civil litigation . . . and the system that Congress created for the adjudication of veterans’ benefits claims could hardly be more dramatic”); 28 U.S.C. §§ 2412(d)(1)(A) (awarding fees in “any civil action” brought “in any court”), (d)(2)(F) (defining “court” to include “the United States Court of Appeals for Veterans Claims”).
15 Finally, the Government argues habeas is something other than a civil action because, on rare occasion, courts have held habeas actions fall outside the scope of particular statutory references to “civil actions.” However, none of those decisions repudiated the centuries-long doctrine that habeas actions are civil. And none undermined the presumption Congress used the term in the EAJA with that history in mind. See Bruesewitz v. Wyeth LLC, 562 U.S. 223, 243 (2011) (noting that when “nearly all of the relevant judicial decisions have given a term or concept a consistent judicial gloss, we presume Congress intended the term or concept to have that meaning when it incorporated it into a later-enacted statute” (cleaned up)). All these decisions did was acknowledge reasons specific to those statutes to treat habeas differently there. They do not create any ambiguity here, much less change the nature of habeas.
Consider each of the cases. The Supreme Court once said in a footnote that Congress’s provision for nationwide service of process in “a civil action” in 28 U.S.C. § 1391(e) did not apply to habeas actions. Schlanger, 401 U.S. at 490 n.4. It once criticized “the label [‘civil action’] [a]s gross and inexact.” Harris, 394 U.S. at 293–94. And we have described habeas as “somewhat of a hybrid,” Callwood v. Enow, 230 F.3d 627, 632 (3d Cir. 2000), and held that “a civil action” in one provision of the Prison Litigation Reform Act (PLRA) does not cover habeas actions, Santana, 98 F.3d at 754–56.
None of these decisions held habeas actions are not civil. In fact, each one reaffirmed that they are. See Schlanger, 401 U.S. at 490 n.4 (acknowledging “habeas corpus is technically ‘civil,’” even though it “is not automatically subject to all the rules governing ordinary civil actions”); Harris, 394 U.S. at 293–94 (affirming habeas corpus is “characterized as
16 ‘civil’” even though “the proceeding is unique” in that it “has conformed with civil practice only in a general sense”); Callwood, 230 F.3d at 632 (“A suit seeking a writ of habeas corpus, although admittedly somewhat of a hybrid, is considered civil in nature.”); Santana, 98 F.3d at 754–55 (describing habeas proceedings as, “in effect, hybrid actions,” but reiterating that “[h]abeas corpus proceedings are technically civil actions,” “independent civil dispositions of completed criminal proceedings”). That is why in Callwood, after calling habeas a hybrid, we reaffirmed it was a civil action. 230 F.3d at 632. We have just done so again, “hold[ing] that the phrase ‘civil action’ in [28 U.S.C.] § 1631”—a statute enacted just two years after the EAJA—“encompasses habeas proceedings,” even “though they are hybrid actions.” Khalil v. President, United States of America et al., Nos. 25-2162 & 25- 2357, 2026 WL 111933, at *5 (3d Cir. 2026). Even if (or when) habeas is a hybrid, that is a distinctive type of civil action, not a third category of proceeding beyond civil and criminal.
All Schlanger, Harris, and Santana held was that the particular provisions before them used “civil action” in a way that did not cover habeas. Each reached that conclusion by doing something we may not: overriding the plain meaning of statutory text with legislative history. Schlanger acknowledged “habeas corpus is technically ‘civil’” but reasoned the use of “civil action” in § 1391(e) did not cover it because “the legislative history of that section is barren of any indication that Congress extended habeas corpus jurisdiction.” 401 U.S. at 490 n.4; see also Stafford, 444 U.S. at 543 (observing Schlanger “recogniz[ed] that habeas corpus is a ‘civil action,’” but “turn[ed] to the legislative history to determine which ‘civil actions’ § 1391(e) governed”). Harris declined to apply Federal Rule of Civil Procedure 33 to habeas actions despite
17 the fact “that habeas corpus proceedings are characterized as ‘civil’” only after divining “the intent of the draftsmen.” 394 U.S. at 293–95. In Santana, we disregarded what we acknowledged was “the plain meaning” of the text and the PLRA’s express exclusion of habeas elsewhere, id. at 754–55, in significant part because of that statute’s legislative history, id. at 755, and precedents that took the same approach, like Schlanger and Harris, id. at 754–55. We do not read statutes this way anymore, Food Mktg. Inst. v. Argus Leader Media, 588 U.S. 427, 436 (2019)—especially not waivers of sovereign immunity, Kirtz, 601 U.S. at 49.5
These decisions do not make it plausible that, in the EAJA, “civil action” excludes habeas actions. Each reaffirmed that they are civil. No surprise, for our law long has recognized that habeas “is a civil, and not a criminal, proceeding,” see Fisher v. Baker, 203 U.S. 174, 181 (1906)—a line Harris cited, 394 U.S. at 293; see also, e.g., In re Nwanze, 242 F.3d 521, 526 n.2 (3d Cir. 2001) (“[H]abeas corpus petitions are technically civil actions[.]”); Parrott v. Gov’t of Virgin Islands, 230 F.3d 615, 620 (3d Cir. 2000) (“Because habeas proceedings are generally considered civil in nature, the term ‘civil action’ includes habeas petitions.” (citation omitted)); Henderson v. Frank, 155 F.3d 159, 167 (3d Cir. 1998) (“[T]he writ of habeas corpus is a civil proceeding[.]”).
5 A panel of our Court once opined that a habeas challenge to criminal detention is not an EAJA “civil action.” Daley v. Fed. Bureau of Prisons, 199 F. App’x 119, 121 (3d Cir. 2006). We are unpersuaded by this unexplained dictum in a not precedential opinion—not least because the case was an appeal by a pro se litigant who could not recover attorneys’ fees in the first place.
18 All told, we cannot put it better than the Supreme Court did when Congress enacted the EAJA: “It is well settled that habeas corpus is a civil proceeding.” Browder, 434 U.S. at 269; see also Stafford, 444 U.S. at 543 (“Habeas corpus is a civil action.” (cleaned up)).6 Accordingly, we hold “civil action” in the EAJA clearly encompasses habeas challenges to immigration detention under 28 U.S.C. § 2241.
3. Even if “civil action” were ambiguous, “any civil action (other than cases sounding in tort)” clearly covers habeas challenges to immigration detention.
Unable to dislodge the centuries of authority establishing a petition for a writ of habeas corpus is a civil action, the Government retreats to the position that the EAJA is ambiguous between a meaning that includes habeas actions and one that does not. As we have discussed, even though the phrase “civil action” almost always covers habeas, there are rare exceptions when it does not. Emphasizing that fact, the Government contends that in isolation the phrase “civil action” might mean either all civil actions or only some subset of them, excluding the ones that are “unique,” Harris, 394 U.S. at 294, or “hybrid,” Santana, 98 F.3d at 754. And that, the
6 The Government also contends habeas is not an EAJA “civil action” because the enacting Congress would have had in mind the Supreme Court’s relatively recent decisions in Harris and Schlanger. Even if those decisions held habeas were not civil—which they did not—this argument would fail by its own terms, because Browder and Stafford—which expressly recognized habeas is civil—are years closer to the EAJA’s enactment.
19 Government would have us conclude, means the EAJA’s “civil action” provision is ambiguous.
However, we do not read statutes in isolation. “A statutory provision is not ambiguous simply because, by itself, it is susceptible to differing constructions.” Hayes v. Harvey, 903 F.3d 32, 41 (3d Cir. 2018) (en banc) (cleaned up). “Rather, in examining statutory language, we take account of the specific context in which that language is used.” Id. (cleaned up).7
Here, the most immediate context—the wording on either side of “civil action”—dispels any doubt the EAJA encompasses habeas actions. Zoom out one word, and we find the provision does not just say “civil action.” It says “any civil action.” 28 U.S.C. § 2412(d)(1)(A). On Abioye’s and Michelin’s interpretation, the word “any” matters: it conveys that the statute reaches civil actions “of whatever kind”—even variations like habeas proceedings. See Cazun v. Att’y Gen., 856 F.3d 249, 255 (3d Cir. 2017) (quoting Webster’s Third New International Dictionary 97 (1976)); see also Lac du Flambeau Band of Lake Superior Chippewa Indians v. Coughlin, 599 U.S. 382, 390 (2023) (emphasizing the significance of the Bankruptcy Code’s categorical abrogation of the sovereign immunity of “any governmental unit that might attempt to assert it” (emphasis in text)).
Their interpretation commands strong support. The Supreme Court has held the words “any civil action”
7 For good reason: In isolation, almost anything can be ambiguous. “Mary had a little lamb” seems clear. But did Mary own a baby sheep or eat a piece of one for dinner?
20 unambiguously cover every civil action not expressly excepted. Ex parte Collett, 337 U.S. 55, 58 (1949) (“The reach of ‘any civil action’ is unmistakable.” (footnote omitted)).
On the Government’s interpretation, the word “any” plays no role. The statute would have had the same scope if Congress had written “civil action” alone, applying only to “garden-variety civil actions.” Abioye Opening Br. 25. That will not do. “We are reluctant to treat statutory terms as surplusage in any setting.” Bufkin v. Collins, 604 U.S. 369, 386 (2025) (quoting TRW Inc. v. Andrews, 534 U.S. 19, 31 (2001)). If “any” means anything, it means “any civil action” encompasses habeas actions.
The Government responds that “[t]he adjective ‘any’ is indeed a broad term, but it cannot expand the reach of the noun it modifies.” Abioye Reply Br. 12–13 (quoting San Francisco v. EPA, 604 U.S. 334, 348 (2025)). True enough. But the point is not that “any” expands the scope of “civil action”: the word gives “civil action” its complete scope, including civil proceedings with unusual features. Even if habeas is a distinctive civil action, reading “any civil action” to encompass a habeas proceeding is not like reading “any mammal” to “encompass a bird or fish,” see San Francisco, 604 U.S. at 348, for the Government acknowledges “habeas corpus is technically ‘civil’”, Abioye Opening Br. 21 (quoting Schlanger, 401 U.S. at 490 n.4); see also Abioye Reply Br. 10 (“habeas proceedings may ‘technically’ be civil actions”), 13 (“they are ‘technically’ civil actions”). At most, it is like reading “any mammal” to encompass a platypus—technically a mammal, even though it lays eggs. See Pugin v. Garland, 599 U.S. 600, 623 (2023) (Sotomayor, J., dissenting).
21 Then there are the words that follow “any civil action.” The EAJA continues with “(other than cases sounding in tort).” 28 U.S.C. § 2412(d)(1)(A). “When Congress provides exceptions in a statute, . . . [t]he proper inference . . . is that Congress considered the issue of exceptions and, in the end, limited the statute to the ones set forth.” United States v. Johnson, 529 U.S. 53, 58 (2000). Congress explicitly carved out torts. The “proper inference” is that Congress did not implicitly carve out habeas actions. The Government’s sole response is that because habeas is not a civil action, Congress did not have to exclude it expressly. But as we have shown, it is civil.
The wider context confirms our conclusion. Interpreting “any civil action” to include habeas actions is the only way to harmonize the EAJA with the rest of the U.S. Code. Interpreting “any civil action” to exclude them would render the EAJA an anomaly. We strive to construe each word “to contain that permissible meaning which fits most logically and comfortably into the body of both previously and subsequently enacted law.” W. Va. Univ. Hosp., Inc. v. Casey, 499 U.S. 83, 101 (1991), superseded by statute as recognized in Landgraf v. USI Film Products, 511 U.S. 244, 251 (1994); see also Lac du Flambeau Band, 599 U.S. at 392–93 & n.3 (holding Bankruptcy Code’s abrogation of the sovereign immunity of “governmental units” extends to Indian tribes in part because “Congress has repeatedly characterized tribes as governments” in other statutes); Burrage v. United States, 571 U.S. 204, 212 (2014) (interpreting statute in light of how “courts regularly read” comparable language elsewhere in the U.S. Code); Bostock v. Clayton Cnty., 590 U.S. 644, 791–92 (2020) (Kavanaugh, J., dissenting) (observing that “longstanding and widespread congressional practice matters” in interpreting
22 statutes). We do so because “it is our role to make sense rather than nonsense out of the corpus juris.” W. Va. Univ. Hosp., 499 U.S. at 101.
Every time Congress has used “any civil action” or a cognate, the phrase encompasses habeas proceedings, save where Congress expressly excluded them. See, e.g., Boumediene v. Bush, 553 U.S. 723, 796 (2008) (permitting the Government to move for a change of venue for a habeas action under 28 U.S.C. § 1404(a), which applies to “any civil action”); German Santos v. Warden Pike Cnty. Corr. Facility, 965 F.3d 203, 208 (3d Cir. 2020) (holding the district court had jurisdiction over a habeas action under 28 U.S.C. § 1331, which applies to “all civil actions”); Hilborn v. United States, 163 U.S. 342, 345 (1896) (holding a since-repealed statute covering “all civil actions” included habeas actions); Reid v. Covert, 351 U.S. 487, 489–90 (1957) (holding the Supreme Court had jurisdiction over an appeal from a grant of a habeas petition pursuant to a since-repealed statute authorizing direct appeals from “any civil action, suit, or proceeding” meeting certain requirements); 28 U.S.C. § 1914(a) (“[T]he parties instituting any civil action, suit or proceeding in [a district] court . . . [must] pay a filing fee of $350, except that on application for a writ of habeas corpus the filing fee shall be $5.”); 18 U.S.C. § 3626(g) (defining the phrase “civil action with respect to prison conditions” as “any civil proceeding” meeting certain criteria except “habeas corpus proceedings challenging the fact or duration of confinement”).
The Government has not produced a single counterexample. Reading the EAJA consistently with the rest of the U.S. Code therefore requires reading “any civil action”
23 to comprehend habeas actions, as we do everywhere else they are not carved out by name.
On the Government’s reading, the EAJA is one of a kind, and not in a good way: the only statute in the U.S. Code to use “any civil action” to implicitly exclude habeas actions. On Abioye’s and Michelin’s reading, the EAJA is part of a consistent, coherent body of law. That is one more reason to think the statute unambiguously covers habeas proceedings.
In sum, the Government advances a superficially instinctive argument: because “civil action” does not invariably cover habeas actions, the phrase must be ambiguous between an interpretation that does and another that does not. However, in context—preceded by “any,” followed by a torts exception, and situated in a body of law where “any civil action” includes habeas save where explicitly excluded—the EAJA’s “civil action” provision clearly covers habeas actions pursuant to 28 U.S.C. § 2241.
4. We hew closest to the Tenth Circuit’s approach.
Our specific issue has divided other courts. The Fourth and Fifth Circuits have held habeas proceedings are not clearly EAJA “civil actions.” The Second, Ninth, and Tenth Circuits have held habeas challenges to immigration proceedings are. We find the Tenth Circuit’s opinion most persuasive. Still, we chart our own path.
Begin with the Fourth Circuit. In O’Brien v. Moore, it held the EAJA does not waive sovereign immunity to fee awards in habeas proceedings challenging criminal detention. 395 F.3d at 508. The statute’s use of “civil action,” the Court
24 thought, was ambiguous between every civil action, which would include habeas actions, and every “wholly civil action,” which would not. See id. at 504–06. “[B]ecause habeas actions have both a criminal and civil nature,” it reasoned, “when a statutory provision regulates a ‘civil action,’ we can only conclude that it does not necessarily follow that it also regulates a habeas proceeding.” Id. at 505–06.
We do not follow O’Brien for four reasons. First, the decision ignored two crucial pieces of the statutory text: the “any” that precedes “civil action” and the torts exception that follows. Second, the decision largely neglected the deep common–law history of habeas as a civil action. Third, O’Brien’s reasoning does not extend to habeas challenges to civil immigration detention. As we have noted, the Fourth Circuit classified habeas cases as hybrids because “to the extent that a habeas proceeding reviews a criminal punishment with the potential of overturning it, the habeas proceeding necessarily assumes part of the underlying case’s criminal nature.” Id. at 505. The habeas petitions before us review immigration detention—a civil matter—so there is no underlying criminal case whose nature they could assume. Fourth, O’Brien acknowledged this distinction, distinguishing cases where “the habeas corpus proceeding was filed in an immigration context” from the challenge to criminal detention before it. Id. at 507 (citing Sotelo-Aquije v. Slattery, 62 F.3d 54, 56, 59 (2d Cir. 1995)). By its own terms, the case has little to say to us.
Four years ago, a divided panel of the same Court held the EAJA does not unambiguously cover habeas challenges to immigration detention, claiming “[t]his conclusion is required by” O’Brien. Obando-Segura v. Garland, 999 F.3d 190, 193
25 (4th Cir. 2021). The Obando-Segura Court brushed aside the distinction between challenges to criminal detention and civil, asserting—incorrectly—that O’Brien “did not differentiate” between them. Id. at 194. This time, the Fourth Circuit acknowledged that the EAJA covers “any civil action.” Id. at 196. But it disregarded “any” without weighing its consequences: “whatever we think of the possibility that the term ‘any’ can sufficiently disambiguate the term ‘civil action,’ that argument cannot survive O’Brien.” Id. And the Court again failed to take account of the torts exception. So we are not persuaded by Obando-Segura either.
Turn to the Fifth Circuit. Its brief decision in Barco v. Witte, 65 F.4th 782 (5th Cir. 2023), cert. denied, 144 S. Ct. 553 (2024), relies on Obando-Segura and shares its flaws. Barco also held “civil action” is ambiguous between all civil actions, which might include habeas, and “purely civil” actions, id. at 783, which would not because of the “hybrid nature” of habeas, id. at 785. So it, too, held the EAJA does not waive sovereign immunity for habeas challenges to immigration detention. Id. Like O’Brien and Obando-Segura, Barco did not analyze whether the word “any” or the torts exclusion make a difference. Like Obando-Segura, Barco deferred to circuit precedent treating habeas actions as hybrids between criminal and civil actions. And it also did so without explaining why challenges to civil detention have any criminal component or reckoning with the historic civil status of petitions for writs of habeas corpus. Consequently, we find the Fifth Circuit’s decision no more convincing than its Fourth Circuit predecessors.
The Second and Ninth Circuits reached the same result we do, but for reasons we decline to follow. The Ninth Circuit
26 held the EAJA’s fee award provision encompassed habeas petitions because that interpretation would advance the statute’s purposes: providing a financial incentive for private parties to challenge Government conduct and encouraging litigation to improve public policy. In re Hill, 775 F.2d 1037, 1040–41 (9th Cir. 1985). The Court barely considered the text. See id. The Second Circuit held habeas challenges to immigration detention were EAJA “civil actions” because of “the legislative history of the EAJA,” despite “the ambiguity of the term ‘civil action[]’ and the indeterminacy of the existing precedent.” Vacchio v. Ashcroft, 404 F.3d 663, 669–72 (2d Cir. 2005). Unlike these courts, we stick to the text and context read against the backdrop of our legal history.
We have found the Tenth Circuit’s recent decision, Daley v. Ceja, 158 F.4th 1152 (10th Cir. 2025), illuminating. As do we, that Court held the EAJA waives federal sovereign immunity to fee awards for habeas challenges to immigration detention because “any civil action” unambiguously encompasses those actions. See id. at 1155, 1162, 1166. And the Court relied, in part, on an insightful evaluation of the historic civil status of habeas actions. See id. at 1157–60. But we part ways in one analytical respect. The Daley Court considered the EAJA’s purposes. See id. at 1162–64. We do not.
For these reasons, we hold the EAJA unambiguously applies to habeas challenges to immigration detention under 28 U.S.C. § 2241.
27 B. The District Court did not abuse its discretion in finding the Government’s position in Abioye was not substantially justified.
The Government detained Abioye for over 16 months without an individualized bond hearing before he petitioned for a writ of habeas corpus. When he filed his petition, the Government contested his right to a hearing. He prevailed, yet he had spent over 18 months in custody.
Under the EAJA, the Government’s position was substantially justified only if its conduct was “justified to a degree that could satisfy a reasonable person.” Johnson v. Gonzales, 416 F.3d 205, 210 (3d Cir. 2005) (quoting Pierce v. Underwood, 487 U.S. 552, 565 (1988)). The Government bears the burden of proving its position was justified. Hanover Potato Prods., Inc. v. Shalala, 989 F.2d 123, 128 (3d Cir. 1993). “To satisfy this burden and defeat a prevailing party’s application for fees, the government must . . . demonstrat[e] ‘(1) a reasonable basis in truth for the facts alleged; (2) a reasonable basis in law for the theory it propounded; and (3) a reasonable connection between the facts alleged and the legal theory advanced.’” Cruz, 630 F.3d at 324 (quoting Morgan v. Perry, 142 F.3d 670, 684 (3d Cir. 1998)). “[I]n immigration cases, the Government must meet the substantially justified test twice”: once for its underlying conduct and once for its decisions in the ensuing litigation about that conduct. Johnson, 416 F.3d at 210. We do not assume the position of the Government was not substantially justified simply because it lost. William v. Astrue, 600 F.3d 299, 302 (3d Cir. 2009).
8 U.S.C. § 1226(c) requires the Government to detain immigrants with certain criminal convictions pending removal
28 without bond hearings. German Santos, 965 F.3d at 206. The Supreme Court has held § 1226(c) is not facially unconstitutional. Demore v. Kim, 538 U.S. 510, 531 (2003). But the Court “did so because it understood that the detention would last only for a ‘very limited time,’” German Santos, 965 F.3d at 208 (quoting Demore, 538 U.S. at 529 n.12)—“roughly a month and a half in the vast majority of cases” and “about five months” if the immigrant appealed removal, Demore, 538 U.S. at 530. Accordingly, immigrants detained under § 1226(c) may bring as-applied challenges to the constitutionality of their detentions. German Santos, 965 F.3d at 209. The longer they are detained without bond hearings, the more likely their detention abridges the liberty secured by the Due Process Clause. See id. at 209–10; Demore, 538 U.S. at 532 (Kennedy, J., concurring). “When detention becomes unreasonable, the Due Process Clause demands a hearing.” German Santos, 965 F.3d at 210 (cleaned up).
We have identified four “nonexhaustive” factors governing when detention is unreasonable: the length of detention, the likelihood it will continue, the reasons for delay, and how confinement conditions compare to criminal punishment. Id. at 211–12. “The most important factor is the duration of detention.” Id.
The Government had no reasonable basis in law for contesting Abioye’s petition for a bond hearing after over 16 months in detention without one. We have spoken clearly: “Detention becomes more and more suspect after five months.” Id. (cleaned up). To be sure, we have declined to impose a per se rule that any detention longer than a certain duration is unreasonable; the precise time may vary case to case. Id. Nevertheless, we have held “detention became unreasonable
29 sometime between six months and one year” after it began. Id. (citing Chavez-Alvarez v. Warden York Cnty. Prison, 783 F.3d 469 (3d Cir. 2015), abrogated in part on other grounds by Jennings v. Rodriguez, 583 U.S. 281 (2018)). Nothing in the facts here gave the Government a reasonable basis for arguing against a bond hearing after roughly double that length of time—and more than triple the length of time the Supreme Court assumed in Demore and we emphasized in German Santos. That is especially true because Abioye’s detention was likely to continue much longer because of his appeal to the Fourth Circuit. See id. at 212 (finding second factor “strongly supports a finding of unreasonableness” when detainee has pending appeal). That appeal continues to this day.
To be sure, the Government could reasonably have thought the third and fourth factors did not weigh against its position. The reasons for delay favored neither side. Although Abioye’s appeal of the order of his removal prolonged his detention, we do “not hold an alien’s good-faith challenge to his removal against him, even if his appeals or applications for relief have drawn out the proceedings.” Id. at 211. Likewise, although the Government’s incorrect opposition to Abioye’s petition for a bond hearing compounded the delay, we do not “hold the agency’s legal errors against the Government, unless there is evidence of carelessness or bad faith.” Id.
The conditions of Abioye’s confinement did not clarify the matter either. On one hand, as the District Court noted, the conditions were not as obviously penal as those we held against the Government in German Santos. There, the detainee was held alongside convicted criminals. Id. at 212–13. Here, Abioye was held with other immigration detainees. The detainee in German Santos was confined to his cell for 23
30 hours a day. Id. at 213. Abioye was afforded several hours of recreation, and the rest of his day was spent in his “pod”—a group of cells—rather than just his own cell. On the other hand, District Courts in our Circuit have recognized the conditions at Moshannon are penal in nature. See Grigoryan v. Jamison, No. 25-1389, 2025 WL 1257693, at *5 (E.D. Pa. Apr. 30, 2025) (collecting cases).8 And Abioye attested the conditions he experienced there were worse than those he experienced in federal criminal custody. For our purposes, what matters is that even if this factor weighed against detaining Abioye any longer without a hearing, it did not weigh so heavily in that direction that the Government had no reasonable basis for contesting it.
With that said, the Government also had no reasonable basis in law for thinking any ambivalence on the third and fourth factors could outweigh its decisive loss on the first two. Our conclusion remains that the District Court did not abuse its discretion in finding the Government was not substantially justified in detaining Abioye for so long, with no end on the horizon, in the absence of any distinctive reason to deny him a hearing. It should have known better.
The Government advances several counterarguments. None proves sound. The first is that the Supreme Court might one day endorse the Government’s position that indefinite detention without a bond hearing under § 1226(c) is constitutional, full stop. Perhaps it will. In the meantime, we have held that indefinite detention without a bond hearing can violate the Due Process Clause and that as-applied challenges
8 We have said the same about the conditions at Pike County Correctional Facility, where Abioye spent a few of his months in confinement. See German Santos, 965 F.3d at 212–13.
31 raising that claim are available. German Santos, 965 F.3d at 209. And this appeal is not a proper vehicle for revisiting that question because the Government did not appeal the District Court’s determination that the statute was unconstitutional as applied to Abioye—only that Court’s subsequent award of fees.
Second, the Government claims the District Court abused its discretion by applying a brightline rule that any detention longer than one year is unreasonable. But the Court did not do that. Instead, it correctly treated the reasonableness of detention as a fact-sensitive inquiry. Even if it had applied a brightline rule, “we may affirm on any ground supported by the record . . . even if the [D]istrict [C]ourt overlooked it or it involves an attack on the [D]istrict [C]ourt’s reasoning.” See Laurel Gardens, LLC v. Mckenna, 948 F.3d 105, 116 (3d Cir. 2020). What’s more, for all the Government’s insistence that the length of detention must be assessed case by case, it offers no special reason to think more than 16 months of detention here—18 months by the time of Abioye’s release—was reasonable in this case.
Third, the Government asserts the District Court abused its discretion because the decision awarding fees did not revisit the German Santos analysis from the merits decision that Abioye’s detention was unreasonable. However, the Court had no obligation to recap its ruling. See Morgan v. Perry, 142 F.3d 670, 683 (3d Cir. 1998) (noting, without issue, that the District Court’s factual findings were recited in the merits decision but not the fee decision).
Fourth, the Government claims the length of detention could not support a finding its position was not substantially
32 justified because the District Court described this factor as “a close call” when ruling on the merits of the habeas petition. Abioye Opening Br. 47–49 (quoting Abioye App’x 29). The Court violated the law of the case doctrine, as the Government sees it, by subsequently finding it had no reasonable basis for detaining Abioye that long. It offers no authority for the proposition that the Court’s offhand remark is subject to the law of the case doctrine. Even if it were, we could affirm on the ground that the record indicates the length of detention was not, in fact, “a close call”—because it was not.
Fifth, the Government argues its position was substantially justified because the District Court did not hold every German Santos factor cut against detention: the Court found the first two weighed in favor of Abioye, but the third was neutral and the fourth favored the Government. However, we have never held that the Government’s position is substantially justified just because a single factor weighs in its favor—and that argument is particularly implausible where, as here, the most important factor (duration) weighed against the Government and the second factor “strongly support[ed]” the same conclusion. German Santos, 965 F.3d at 212.
In sum, the District Court did not abuse its discretion in determining the Government failed to prove it was substantially justified in resisting Abioye’s petition for an individualized bond hearing after detaining him for over 16 months without one, with no end to his detention then in sight.
The EAJA authorizes an award of attorneys’ fees and costs to the prevailing party “in any civil action (other than
33 cases sounding in tort)” brought by or against the United States if the Government’s position was not “substantially justified.” 28 U.S.C. § 2412(d)(1)(A). We hold this provision clearly covers petitions for writs of habeas corpus from immigration detention under 28 U.S.C. § 2241, and so it waives federal sovereign immunity to the fee awards issued here. Our rationale: Habeas actions are civil actions. They have been since before our Nation’s Founding. As the Supreme Court and our Court have recognized, the plain and technical meanings of “civil action” encompass them. Even if “civil action” were ambiguous because it does not always cover habeas proceedings, “any civil action (other than cases sounding in tort)” is as clear as can be. “Any” extends coverage to all civil actions, no matter the type. And by expressly excluding torts, Congress ruled out any implicit exclusions.
We also hold the District Court did not abuse its discretion in finding the Government’s position in Abioye was not substantially justified. After detaining Abioye for over 16 months, with reason to think he would remain in custody for months or years to come, the Government had no reasonable basis in law for fighting to deny him an individualized bond hearing.
We close by echoing the Supreme Court’s recent reflections on the historic role of the writ of habeas corpus. “When English monarchs jailed their subjects summarily and indefinitely, common-law courts employed the writ as a way to compel the crown to explain its actions—and, if necessary, ensure adequate process . . . before allowing any further detention. The Great Writ was, in this way, no less than ‘the instrument by which due process could be insisted upon.’” Brown v. Davenport, 596 U.S. 118, 128 (2022) (citation
34 omitted) (quoting Hamdi v. Rumsfeld, 542 U.S. 507, 555 (2004) (Scalia, J., dissenting)). It remains so today. With this history in mind, we affirm.