McHUGH, Circuit Judge.

I. INTRODUCTION

After multiple attempts over the course of fourteen years to attack his federal convictions on various drug and firearm charges, Jeffrey Dan Williams filed a motion in the United States District Court for the Northern District of Oklahoma seeking to withdraw his guilty plea based on newly discovered evidence. Specifically, Mr. Williams submitted affidavits in support of his claim that his guilty plea was involuntarily entered because the law enforcement officers investigating his case planted evidence, gave false testimony, and used threats and intimidation to suborn perjury from other witnesses. After conducting an evidentiary hearing, the district court vacated Mr. Williams's convictions. It concluded that the officers had perpetrated a fraud on the court and that vacating Mr. Williams's convictions was necessary to correct the fraud and to prevent a miscarriage of justice. The government appeals, asserting the district court lacked jurisdiction over Mr. Williams's motion due to his failure to first obtain certification from this court, as required by 28 U.S.C. § 2255, permitting him to file a second or successive petition for habeas corpus relief.

Exercising jurisdiction pursuant to 28 U.S.C. §§ 1291 and 2244, we hold that the Antiterrorism and Effective Death Penalty Act of 1996 (AEDPA) limits the courts' power to correct fraud on the court and to prevent a miscarriage of justice when the court vacates a conviction in response to a second or successive habeas petition. Although a district court may invoke these powers sua sponte, the district court here acted on Mr. Williams's motion to withdraw his guilty plea, which is a second or successive petition. We therefore reverse the district court's order vacating Mr. Williams's conviction for lack of jurisdiction. But we exercise our discretion to treat Mr. Williams's appellate brief as a request to file a second or successive motion under 28 U.S.C. § 2255(h), and we grant that request in part.

II. BACKGROUND

A. Factual History1

According to the government, from at least 1995 to 1997, Mr. Williams operated and directed an illegal drug distribution operation, primarily in Tulsa, Creek, and Osage Counties, Oklahoma. He was an experienced methamphetamine cook and charged large fees to teach others to manufacture methamphetamine. Mr. Williams also employed associates to help him manufacture methamphetamine, to purchase necessary precursor chemicals, and to find locations to manufacture and store his product.

Local authorities arrested individuals connected with the conspiracy beginning in 1994 but they did not realize the full scope of the conspiracy until 1997. By that point, the authorities had made controlled buys, seized methamphetamine and paraphernalia at traffic stops, and interviewed various witnesses about their involvement with Mr. Williams and a coconspirator, James Edmondson. At that point, law enforcement officers turned the focus of their investigation to Mr. Williams. Various searches of Mr. Williams's vehicle, residence, and the residence of his then-girlfriend uncovered drugs, drug paraphernalia, drug manufacturing equipment, and a firearm. As a result, Mr. Williams was arrested and ultimately charged under federal law.

B. Procedural History

1. The Federal Indictment, Guilty Plea, and Sentencing

Mr. Williams pled guilty to four counts alleged in a federal indictment: conspiring to manufacture and possess methamphetamine (the conspiracy count); two counts of possessing with intent to distribute methamphetamine on March 12, 1997, and July 22, 1997 (the drug counts); and possessing a firearm during and in relation to a drug trafficking crime (the firearm count). At sentencing, Mr. Williams disputed the amount of methamphetamine he was responsible for manufacturing and/or distributing, but the district court rejected Mr. Williams's argument and instead accepted the testimony of a DEA agent. The district court therefore sentenced Mr. Williams to a total term of 420 months imprisonment, including 360 months for each of the conspiracy and drug counts, running concurrently, and 60 months for the gun count, running consecutively.

2. The Direct Appeal and Prior Post-Conviction Relief Proceedings

Mr. Williams appealed his conviction, and we affirmed. See United States v. Williams, 201 F.3d 449 (10th Cir.1999) (unpublished table decision). He then filed a series of collateral attacks. His first four § 2255 motions were denied by the district court and the Tenth Circuit. See, e.g., In re Williams, No. 08-5079 (10th Cir.2008); United States v. Williams, 167 Fed.Appx. 25 (10th Cir. 2006) (per curiam) (unpublished); United States v. Williams, 44 Fed.Appx. 443 (10th Cir.2002) (unpublished).

On November 12, 2010, Mr. Williams filed a fifth request for postconviction relief, in which he first raised the allegations relevant to this appeal. Mr. Williams claimed at least five Tulsa Police Department officers who were involved in the investigation of his case were also the subject of an investigation into corruption at the Tulsa Police Department (the corruption investigation), which uncovered instances of planted evidence and perjury in other criminal cases. According to Mr. Williams, these officers fabricated and manipulated evidence, intimidated witnesses, and used false informants in his case.

Although we denied Mr. Williams authorization to file a successive habeas petition because he had not provided evidence to support his allegations, we noted our denial was "without prejudice to Mr. Williams's refiling, in this court, a motion for authorization containing a complete description of all relevant facts and circumstances, with supporting evidence."

In an attempt to comply with our instructions, Mr. Williams refiled his request for authorization to file a second or successive § 2255 petition with this court and attached notarized statements from various witnesses. We again denied the motion because the new evidence did not establish a connection between any officer implicated in the corruption investigation and the witnesses and federal agents involved in Mr. Williams's case.

3. The Current Postconviction Relief Proceeding

a. Mr. Williams's Motion

Mr. Williams next filed a pro se "Motion to Withdraw and Nullify Guilty Plea" (the Motion) with the district court on January 17, 2012, which is the subject of this appeal. In the Motion, he argued that "the same officers [involved in the corruption investigation had] engaged in the same type of illegal conduct during the investigation, searches and seizures [in Mr. Williams's case,]" and that this conduct resulted in Mr. Williams's guilty plea. Mr. Williams attached evidence undermining the testimony of the DEA agent used to support the drug quantity findings at Mr. Williams's sentencing.2 In addition, Mr. Williams attached his own affidavit in which he claimed Tulsa officers conducted searches without probable cause, coerced testimony regarding the gun charge, and seized less methamphetamine than they reported. Mr. Williams further stated that he was pressured to plead guilty because his counsel advised him the government was planning to file a Fourth Superseding Indictment which would add a continuing criminal enterprise (CCE) charge against Mr. Williams and would name Mr. Williams's younger sister as a codefendant.

b. The District Court's Response

The district court construed the Motion as a request under Federal Rule of Civil Procedure 60(d)(3), which acknowledges a district court's inherent authority to set aside a judgment for fraud on the court. The district court also appointed counsel for Mr. Williams, ordered the parties to conduct discovery, and held an evidentiary hearing, which was conducted intermittently between May 2012 and June 2012.3

At the hearing, the district court received testimony from multiple witnesses who supported Mr. Williams's claims that the Tulsa police officers acted fraudulently during the investigation of his case.

The district court issued an opinion and order vacating Mr. Williams's judgment and sentence and dismissing the third superseding indictment. In so doing, the district court found that Tulsa Police Department officers committed a fraud on the court that required Mr. Williams's convictions to be set aside. In the alternative, the court granted Mr. Williams's motion to withdraw his guilty plea based on the court's common law authority to prevent a miscarriage of justice, specifically finding that Mr. Williams had demonstrated actual innocence by a preponderance of the evidence. To distinguish an intervening decision of this court, United States v. Baker, 718 F.3d 1204 (10th Cir.2013),4 which held that motions alleging fraud on the court should be treated as second or successive petitions, the district court explained that "it is the Court, not the defendant, that has invoked its inherent authority and has construed Mr. Williams'[s] motion to Withdraw and Nullify Guilty Plea as a fraud on the court motion." United States v. Williams, 16 F.Supp.3d 1301, 1313 n. 5 (N.D.Okla.2014).

The district court entered its order on April 18, 2014, and the government filed an appeal of that decision on June 16, 2014. In response, Mr. Williams filed a motion in this court seeking to dismiss the government's appeal as untimely.5

III. DISCUSSION

On appeal, the government argues the district court lacked jurisdiction to vacate Mr. Williams's conviction because Mr. Williams failed to first obtain certification from this court permitting him to file a second or successive petition, as required under AEDPA. In response, Mr. Williams claims the district court had independent jurisdiction because (1) the Motion was not a second or successive petition or (2) irrespective of AEDPA's limitations on second or successive petitions, the court had the inherent authority to correct fraud on the court or, alternatively, to prevent a miscarriage of justice.

For the reasons explained below, we conclude that although AEDPA has not eliminated the courts' power to correct fraud on the court or to prevent a miscarriage of justice entirely, AEDPA has restricted those powers in a way that divested the district court of jurisdiction to act on the Motion. In so ruling, we first analyze whether the Motion is a second or successive petition for habeas relief and conclude that it is. Next, we explain that although courts have the inherent authority to correct fraud on the court or to prevent a miscarriage of justice, AEDPA has limited a court's ability to exercise those inherent powers when the court is acting on a second or successive petition for habeas relief. Ultimately, we conclude that the district court did not properly invoke its inherent powers in Mr. Williams's case and therefore it lacked subject matter jurisdiction to vacate his conviction. But we exercise our discretion to construe Mr. Williams's arguments on appeal as a request for authorization to file a successive habeas petition and grant that request in part.

A. The Motion is a Second or Successive Petition for Habeas Corpus Relief.

We begin by determining whether Mr. Williams's Motion is a second or successive petition and is therefore governed by AEDPA. After a federal prisoner has filed one postjudgment habeas petition, which is permitted under 28 U.S.C. § 2255(a), another postjudgment motion is treated as a second or successive § 2255 motion if it asserts or reasserts claims of error in the prisoner's conviction. United States v. Nelson, 465 F.3d 1145, 1147 (10th Cir.2006); see Gonzalez v. Crosby, 545 U.S. 524, 530-31, 125 S.Ct. 2641, 162 L.Ed.2d 480 (2005) (involving a § 2254 motion). To discern whether a postjudgment motion styled otherwise is, in fact, a successive § 2255 motion, we "look at the relief sought, rather than a pleading's title or its form." United States v. Baker, 718 F.3d 1204, 1206-08 (10th Cir.2013). If a petitioner challenges his underlying conviction, his filing is a successive § 2255 motion. See id. (concluding that a petition filed alleging fraud on the conviction court was a second or successive petition). The distinction is important because § 2255(h) deprives a district court of jurisdiction over uncertified second or successive habeas petitions. Nelson, 465 F.3d at 1148.6

Mr. Williams's motion to withdraw his guilty plea falls squarely within the definition of a second or successive petition. He filed an initial § 2255 motion on January 4, 2001, which the district court reviewed and denied on the merits. His current Motion asked the district court to allow him to withdraw his guilty plea and thus challenged his underlying conviction. It is therefore a successive § 2255 motion, and the district court did "not even have jurisdiction to deny the relief sought in the pleading" in the absence of a certification from this court. Nelson, 465 F.3d at 1148; see also 28 U.S.C. § 2255(h) (explaining that a second or successive motion must receive certification from the appropriate court of appeals before it may be considered by a district court).

Despite this, Mr. Williams asks us to consider the Motion as an initial § 2255 petition. He relies on our decision in In re Weathersby, where we held that a petition technically filed after an initial petition need not be treated as a second or successive petition if "the purported defect did not arise, or the claim did not ripen, until after the conclusion of the previous petition." 717 F.3d 1108, 1111 (10th Cir.2013) (internal quotation marks omitted). Attempting to qualify for the Weathersby exception, Mr. Williams alleges the defect in his conviction did not exist when he filed his first habeas petition, and thus his habeas claim was not ripe, because the Tulsa police corruption had not been discovered at that time. We are not persuaded.

Weathersby outlined a narrow exception to the bar on successive § 2255 motions for circumstances where a particular claim cannot be raised in a defendant's initial § 2255 motion. Id. at 1111. This occurs where the factual basis for a claim does not yet exist — not where it has simply not yet been discovered — at the time of a defendant's first motion. For instance, in Weathersby, the defendant's federal sentence was enhanced based on prior state convictions. Id. at 1109. After sentencing and after his first § 2255 motion, the defendant successfully challenged his state convictions and they were expunged. Id. He then raised a separate challenge to his federal sentence. Although the defendant in Weathersby filed a new motion for postconviction relief, this court held it was not a successive petition because the challenge to his enhanced sentence was not ripe when he filed his first § 2255 motion. Id. at 1111. We explained that the factual basis for his claim — that his state convictions were expunged — did not exist when he filed the first habeas corpus motion. Id.

In contrast, the factual basis of Mr. Williams's claim — that Tulsa police officers and various witnesses lied, planted evidence, fabricated witness statements, misled federal agents, and otherwise coerced Mr. Williams to plead guilty — existed at the time of his initial habeas petition. Because these acts occurred well before Mr. Williams filed the Motion, he cannot rely on the Weathersby exception. See United States v. McCalister, 545 F. App'x 718, 720-21 (10th Cir.2013) (unpublished) (treating an allegation of police corruption as a successive § 2255 motion because the police corruption, and hence the defendant's claim, existed at the time of trial).7 Although the proof of those allegations may not have been available to Mr. Williams when he filed his first petition, newly available proof implicates the newly discovered evidence exception in § 2255(h)(1), not whether Mr. Williams's claim should be treated as an initial habeas petition.8

For these reasons, we conclude Mr. Williams's petition must be construed as a second or successive petition governed by § 2255(h) of AEDPA. Under § 2255(h), Mr. Williams was prohibited from filing a second or successive petition without first obtaining certification from this court, which he failed to do. We therefore proceed to Mr. Williams's alternative argument that, even though his Motion is a second or successive petition, the district court had jurisdiction under its inherent authority to correct fraud on the court or to prevent a miscarriage of justice.

B. The Courts' Inherent Powers and AEDPA's Limitations on Them

Federal courts enjoy numerous inherent equitable powers. See Chambers v. NASCO, Inc., 501 U.S. 32, 43-45, 111 S.Ct. 2123, 115 L.Ed.2d 27 (1991) (discussing various inherent powers, including the power to "control admission to its bar and to discipline attorneys who appear before it," to "punish for contempt," to correct fraud on the court, to control its courtroom, to dismiss for forum non conveniens, and to sua sponte dismiss for failure to prosecute); Eash v. Riggins Trucking Inc., 757 F.2d 557, 561-64 (3d. Cir.1985) (outlining three general categories of inherent powers); see also McQuiggin v. Perkins, ___ U.S. ___, 133 S.Ct. 1924, 1934, 185 L.Ed.2d 1019 (2013) (evaluating the courts' equitable authority to prevent a miscarriage of justice). These powers are necessary to "control and direct the conduct of litigation" and are exercised "without any express authorization in a constitution, statute, or written rule of court." Joseph J. Anclien, Broader Is Better: The Inherent Powers of Federal Courts, 64 N.Y.U. Ann. Surv. Am. L. 37, 44 (2008) (alteration and internal quotation marks omitted). Two of these powers, the power to correct a fraud on the court and to prevent a miscarriage of justice, are at issue here.

Although the courts' equitable powers have "always been characterized by flexibility" so that they may "meet new situations which demand equitable intervention, and to accord all the relief necessary to correct the particular injustices involved in these situations," Hazel-Atlas Glass Co. v. Hartford-Empire Co., 322 U.S. 238, 248, 64 S.Ct. 997, 88 L.Ed. 1250 (1944), the judicial power is not boundless. Instead, the Supreme Court has recognized that because Congress created the federal district and circuit courts, it may statutorily limit their inherent powers, so long as such limitations are clearly reflected by statute. Chambers, 501 U.S. at 47, 111 S.Ct. 2123. But see Eash, 757 F.2d at 562 (opining that the judiciary's "irreducible inherent authority" authorizes the exercise of inherent powers which are "so fundamental to the essence of a court as a constitutional tribunal" that Congress cannot limit them). Therefore, where a statute includes "clear[] command" limiting an inherent judicial power, courts cannot exercise that power in contravention of the statute. McQuiggin, 133 S.Ct. at 1934 (quoting Holland v. Florida, 560 U.S. 631, 646, 130 S.Ct. 2549, 177 L.Ed.2d 130 (2010)); see, e.g., Carlisle v. United States, 517 U.S. 416, 428, 116 S.Ct. 1460, 134 L.Ed.2d 613 (1996) (rejecting the argument that a court has the inherent power to act in contradiction to an applicable rule and holding that a district court could not sua sponte grant an untimely motion for judgment of acquittal); Bank of Nova Scotia v. United States, 487 U.S. 250, 254, 108 S.Ct. 2369, 101 L.Ed.2d 228 (1988) (holding that Federal Rule of Criminal Procedure 52 sets the boundaries of a federal court's supervisory power to dismiss an indictment procured by prosecutorial misconduct).

We are therefore charged with determining whether the relevant AEDPA provision, 28 U.S.C. § 2255, reflects the type of clear statutory command that limits the courts' power to correct fraud on the court and prevent a miscarriage of justice in this case.

1. Fraud on the Court

The government argues AEDPA has limited the courts' inherent power to correct fraud on the court where a petitioner has raised allegations of fraud in a second or successive petition but has failed to obtain the required certification under § 2255(h).9 The district court rejected this argument, reasoning that any procedural bars contained in AEDPA were inapplicable because the court did not act on the Motion, but instead acted sua sponte.

For the reasons explained below, we agree with the government that AEDPA limits a court's inherent authority and prohibits district courts from acting on an unauthorized second or successive petition, even if the petition alleges fraud on the court. We also hold that although the district court was correct that AEDPA (and any procedural bars it may contain) does not apply where a court acts sua sponte, rather than on a second or successive petition, the district court did not act sua sponte here. Therefore AEDPA's certification requirement applies and the district court lacked jurisdiction to vacate Mr. Williams's convictions based on fraud on the court.

a. The inherent power to correct fraud on the court

Courts have historically enjoyed the inherent authority to correct judgments obtained by the commission of fraud on the court, regardless of ordinary procedural bars like statutes of limitations or the time limits imposed on motions to set aside judgments for fraud. See Hazel-Atlas, 322 U.S. at 244-46, 64 S.Ct. 997 (acknowledging an exception to the general prohibition against altering a judgment after the expiration of the court term if the judgment was the result of fraud on the court); see also Fed.R.Civ.P. 60(d)(3) ("This rule does not limit a court's power to ... set aside a judgment for fraud on the court."); Zurich N. Am. v. Matrix Serv., Inc., 426 F.3d 1281, 1290-91 (10th Cir.2005) (explaining that procedural limitations on Rule 60(b) motions do not apply to fraud on the court allegations). This is because "a decision produced by fraud on the court is not in essence a decision at all, and never becomes final." Kenner v. Comm'r, 387 F.2d 689, 691 (7th Cir.1968).

The courts' interest in correcting a fraud on the court stems from "far more than an injury to a single litigant." Hazel-Atlas, 322 U.S. at 246, 64 S.Ct. 997. Instead, our primary objective when correcting a fraud on the court is to redress harm to "the integrity of the judicial process." United States v. Estate of Stonehill, 660 F.3d 415, 444 (9th Cir.2011) (internal quotation marks omitted). Although a court may "vacate its own judgment upon proof that a fraud has been perpetrated upon the court," it may also "fashion an appropriate sanction" short of disturbing an otherwise valid judgment. Chambers, 501 U.S. at 44, 111 S.Ct. 2123. Such sanctions include assessing attorney fees against the culpable party, id. at 46, 111 S.Ct. 2123, or suspension, disbarment, or other reprimand against the attorney who abuses the judicial process, Eash, 757 F.2d at 561. Any relief a party may obtain when we correct a fraud on the court is subordinate to our primary interest in restoring the court's integrity.

b. AEDPA restricts the court's jurisdiction to consider a second or successive petition, even where the petition alleges fraud on the court.

Having discussed the nature of and remedies available under the courts' inherent power to correct fraud on the court, we now consider whether AEDPA has limited that power in the context of a second or successive habeas petition. As explained above, supra Part III.A, AEDPA plainly requires that before filing a second or successive § 2255 motion, a petitioner must obtain certification from the appropriate court of appeals. 28 U.S.C. § 2255(h). In the years immediately after Congress passed AEDPA, precedent from this court and other circuits acknowledged the possibility that a court might have jurisdiction to consider a second or successive petition, despite § 2255's procedural bar, where the petition alleges fraud on the court. See, e.g., United States v. McVeigh, 9 Fed.Appx. 980, 983 (10th Cir. 2001) (unpublished) (considering a petitioner's argument that the prosecution committed fraud on the court and assuming "the existence of a fraud on the court exception to the gatekeeping requirements and affirmative limitations in § 2255 applicable to second or successive motions"); Fierro v. Johnson, 197 F.3d 147, 153 (5th Cir.1999) (recognizing the potential that the inherent power to correct fraud on the court may give life to a challenge that AEDPA's procedural bars would otherwise forbid).

As the federal case law interpreting AEDPA progressed, however, it called into question the practice of relabeling what would otherwise constitute a second or successive petition as a motion alleging fraud in an attempt to avoid the reach of AEDPA's procedural bars. See Gonzalez v. Crosby, 545 U.S. 524, 531-32, 125 S.Ct. 2641, 162 L.Ed.2d 480 (2005) (holding that a pleading labeled as a Rule 60(b)(6) motion will be considered a second or successive habeas motion subject to AEDPA if it seeks to add new grounds for relief or attacks a previous habeas determination on the merits); Spitznas v. Boone, 464 F.3d 1213, 1216 n. 4 (10th Cir.2006) (cautioning that "spurious attempts to re-cast substantive habeas arguments in the guise of `fraud on the court' ... will properly be treated as an attempt to allege or re-allege substantive grounds for habeas relief, thus presenting a second or successive petition").

Most recently, in United States v. Baker, we squarely rejected the possibility that alleging fraud on the court of conviction creates an exception to the procedural bars under § 2255(h). 718 F.3d 1204 (10th Cir.2013). In Baker, a federal prisoner filed a second or successive motion asking the court to correct fraud on the court that allegedly took place during his criminal trial, citing Federal Rule of Civil Procedure 60(d)(3)'s savings clause. Id. at 1205. The district court ruled it lacked jurisdiction to hear the motion because the petitioner had not received authorization from this court to file the motion. Id. at 1205-06. On appeal, we agreed and held that a motion asking the district court to invoke its inherent power to correct fraud on the criminal trial court is not exempt from § 2255(h)'s preauthorization requirements. Id. at 1208. We reasoned that despite a court's inherent authority to correct a fraud on the court, a petitioner cannot circumvent AEDPA's statutory certification requirements applicable to second-or-successive applications by filing pleadings that are labeled as motions under Rule 60(b) but are actually habeas corpus petitions in substance. We therefore concluded that a claim seeking the same relief as a second or successive collateral attack, even if it invokes the district court's inherent power, is not exempt from the certification requirements in § 2255(h). Id.10

From Baker, we learn that in this circuit, AEDPA has indeed limited the courts' power to correct fraud on the conviction court where the petitioner has already received fair habeas review.11 Therefore, without authorization from the appropriate court of appeals, a district court may not act on a petitioner's second or successive motion to vacate a petitioner's conviction, even where that motion alleges fraud on the court.12 Because we have determined that the Motion Mr. Williams filed in the district court is a second or successive petition, the district court lacked the jurisdiction under § 2255(h) to act on it in the absence of certification from this court.

c. AEDPA's procedural bars do not apply where a court is not acting pursuant to a motion.

Notwithstanding AEDPA's limitations, Mr. Williams argues the district court here did not act on his second or successive petition, but instead, invoked its inherent authority sua sponte. We thus next address whether the district court could, and, in fact, did, properly act sua sponte.

Mr. Williams is correct that AEDPA will not constrain a court's authority to employ its inherent equitable powers if the court is acting on its own initiative, rather than upon the application filed by the petitioner. Calderon v. Thompson, 523 U.S. 538, 554, 118 S.Ct. 1489, 140 L.Ed.2d 728 (1998). This is so because AEDPA — and any limitations it may contain — governs courts' actions in response to habeas petitions and is therefore applicable only in instances in which "the court acts pursuant to a prisoner's [habeas] `application.'" Id.

In Calderon, a state prisoner in a capital case filed a motion asking the Ninth Circuit to exercise its inherent power to recall its mandate denying his original request for habeas corpus relief based on arguments challenging his underlying conviction. Id. at 546, 118 S.Ct. 1489. The Ninth Circuit thereafter recalled its mandate and granted relief. Id. at 548-51, 118 S.Ct. 1489. The Supreme Court granted the state's petition for certiorari. Id. at 549, 118 S.Ct. 1489. The Court first explained that the defendant's motion asking the circuit court to exercise its inherent power to recall the mandate was effectively a second or successive application for habeas relief governed by § 2244(b) because it challenged the underlying conviction. Id. at 553, 118 S.Ct. 1489. The Supreme Court then reviewed the plain text of § 2244(b), and instructed that if the court acts on the defendant's application for a successive petition, its jurisdiction is constrained by AEDPA, but where the court is truly acting on its own initiative, AEDPA, by its plain language, does not govern the matter. Id.

Next, the Supreme Court undertook the task of determining whether the Ninth Circuit had acted on the defendant's successive petition, beginning with an examination of the circuit's stated basis of its action. Id. The Supreme Court cautioned, "The court's characterization of its action as sua sponte does not, of course, prove this point; had the court considered claims or evidence presented in [the defendant's] later filings, its action would have been based on a successive application, and so would be subject to § 2244(b)." Id. In other words, if, in taking action, "the court considers new claims or evidence presented in a successive application for habeas relief, it is proper to regard the court's action as based on that application" and the court is therefore constrained by AEDPA. Id.13

In Calderon, the Supreme Court concluded the Ninth Circuit recalled its mandate because of voting irregularities related to the first petition and that it did not consider the claims or evidence provided in the petitioner's successive motion. Id. The Supreme Court explained that the Ninth Circuit reconsidered and "acted on the first application [for habeas relief] rather than a successive one." Id. As a result, the Supreme Court held that the Ninth Circuit indeed acted sua sponte, and therefore AEDPA's procedural bars on second or successive petitions were not directly applicable.

Nonetheless, the Court reversed the Ninth Circuit's sua sponte decision, explaining that even when a court acts sua sponte, its must still "exercise its discretion in a manner consistent with the objects of" AEDPA — specifically, promoting the state's interest in the finality of judgment. Id. at 554, 118 S.Ct. 1489. Because the court of appeals' justification for recalling the mandate in Calderon did not comport with the objects of AEDPA, the Court ruled that recalling the mandate constituted an abuse its discretion. Id. at 554-566, 118 S.Ct. 1489. In so ruling, the Court took care to note that Calderon was not a case involving fraud on the court, which would presumably provide a proper basis for reversing a mandate sua sponte. Id. at 557, 118 S.Ct. 1489. Notably, the Court made this fraud-on-the-court distinction in the second part of its analysis, where it explained the circumstances under which a court is within its discretion in reversing a mandate sua sponte. It did not indicate that the presence of fraud on the court would alter the first part of its analysis, wherein it established the test for determining whether a court acts sua sponte.

Calderon thus instructs that in the context of a second or successive petition, AEDPA does not bar the exercise of inherent authority where a court truly acts sua sponte — that is, where it acted without considering claims or evidence presented in the successive habeas petition. Although the dissent would limit the application of Calderon's sua sponte test to cases involving an appellate court's recall of a habeas mandate, we are not convinced the Court intended such a narrow reading. Rather, we read Calderon to establish a general test for determining whether a (district or appellate) habeas court acts sua sponte. This reading of Calderon is more consistent with AEDPA's general ban on consideration of second or successive petitions, which applies at both the district court and appellate court level, as well as with this court's prior interpretation of the term sua sponte. See Bylin v. Billings, 568 F.3d 1224, 1228 n. 6 (10th Cir.2009) (explaining that "[a] court raises an issue sua sponte when it does so `[w]ithout prompting or suggestion; on its own motion'" and ruling that a district court purporting to act sua sponte did not, in fact, do so because the defendant had raised the relevant issue in a motion (second alteration in original)) (quoting Black's Law Dictionary 1437 (7th ed.1999)). We therefore proceed by applying the Calderon test to the facts of Mr. Williams's case.

d. The district court did not act sua sponte in Mr. Williams's case.

Guided by Calderon's principles, we now decide whether the district court acted pursuant to the Motion or whether it acted sua sponte. To do so, we first ascertain the "underlying basis of the court's action." Id. at 554, 118 S.Ct. 1489. Here, the district court claimed it acted sua sponte to correct a fraud on the court. United States v. Williams, 16 F.Supp.3d 1301, 1313 n. 5 (N.D.Okla.2014). But the district court's characterization is not the end of our inquiry; Calderon instructs us to confirm that claim by determining whether the district court "consider[ed] new claims or evidence presented in a successive application for habeas relief." 523 U.S. at 554, 118 S.Ct. 1489.

Applying this test to the facts here, we easily conclude that the district court acted on Mr. Williams's successive application for habeas relief because the district court considered the new claims and evidence raised in Mr. Williams's Motion. In the Motion, Mr. Williams alleged Tulsa police officers lied at the state preliminary hearing and a witness made false statements to a DEA agent, which resulted in an artificial increase in the quantity of methamphetamine involved. Based on these and other allegations, Mr. Williams moved the court to withdraw and nullify his guilty plea or, alternatively, to remand the case for an evidentiary hearing and to appoint him counsel. In response, the district court first granted the alternative relief requested: it appointed counsel, opened discovery, and conducted an evidentiary hearing to further develop the evidence Mr. Williams presented in his Motion. Based on the evidence presented at the hearing, the district court vacated Mr. Williams's judgment. This is essentially the same relief Mr. Williams requested because, had the district court granted Mr. Williams's motion to withdraw his guilty plea, it necessarily would have also vacated the judgment. Cf. Hicks v. Franklin, 546 F.3d 1279, 1287 (10th Cir.2008) (granting a motion to withdraw a guilty plea and vacating the judgment). Under the test announced in Calderon, the district court was acting on Mr. Williams's successive application, something it lacked jurisdiction to do in the absence of certification from this court. Accordingly, we must reverse the district court's decision to vacate Mr. Williams's convictions on the basis of fraud on the conviction court.

2. Miscarriage of Justice

Because we conclude the district court lacked jurisdiction to vacate Mr. Williams's convictions on the basis of fraud on the court, we turn to whether the court had inherent authority to vacate the convictions to prevent a miscarriage of justice. As with Mr. Williams's fraud on the court claim, we first explain the nature of this inherent power and then look to the relevant AEDPA provision, § 2255(h), to determine whether and to what extent Congress has limited the courts' power to prevent a miscarriage of justice.

a. The inherent authority to prevent a miscarriage of justice

Historically, the common law miscarriage of justice exception (also referred to as the actual innocence exception) allowed a petitioner to overcome certain procedural bars to postconviction relief if a petitioner could show "it is more likely than not that no reasonable juror would have convicted him in light of the new evidence." Schlup v. Delo, 513 U.S. 298, 327, 115 S.Ct. 851, 130 L.Ed.2d 808 (1995). This authority "is grounded in the equitable discretion of habeas courts to see that federal constitutional errors do not result in the incarceration of innocent persons." McQuiggin v. Perkins, ___ U.S. ___, 133 S.Ct. 1924, 1931, 185 L.Ed.2d 1019 (2013) (internal quotation marks omitted). Thus, pre-AEDPA "a court could grant relief on a second-or-successive petition, then known as an `abusive' petition, if the petitioner could show that a fundamental miscarriage of justice would [more likely than not] result." Id. at 1933 (internal quotation marks omitted).

But, as with the courts' inherent power to correct a fraud on the court, Congress may statutorily limit the courts' authority to prevent a miscarriage of justice if it does so by "the clearest command." McQuiggin, 133 S.Ct. at 1934 (quoting Holland v. Florida, 560 U.S. 631, 648, 130 S.Ct. 2549, 177 L.Ed.2d 130 (2010)). In McQuiggin, the Supreme Court analyzed whether Congress had limited the courts' inherent power to prevent a miscarriage of justice by enacting AEDPA's one-year statute of limitations on § 2254 petitions. Id. The Court acknowledged that the provisions in AEDPA that allow petitioners to evade procedural bars by presenting clear and convincing evidence of actual innocence contain a miscarriage of justice exception, but an exception that has been modified from the preponderance standard employed at common law. Id. For those provisions, Congress clearly intended that courts may no longer invoke their common law miscarriage of justice authority to allow petitioners to bypass the relevant procedural bar. Id. Instead, courts must apply the exception as modified by Congress. Id. In contrast, "In a case not governed by those provisions... the [common law] miscarriage of justice exception survived AEDPA's passage intact and unrestricted." Id. Because AEDPA's statute of limitations governing first habeas petitions was silent as to a miscarriage of justice exception, the Supreme Court held the common law exception had survived. Id.

b. Section 2255(h) contains a modified miscarriage of justice exception.

Applying McQuiggin's analysis to Mr. Williams's case, we must determine whether the relevant AEDPA provision — in this case, § 2255(h) — contains a modified miscarriage of justice exception. If so, the district court was not free to invoke its common law authority to correct a miscarriage of justice but was instead constrained by the limitations on that power established by Congress.

Section 2255(h)(1) allows a petitioner to bring a second or successive petition if the appropriate court of appeals first certifies that the petition contains "newly discovered evidence that, if proven and viewed in light of the evidence as a whole, would be sufficient to establish by clear and convincing evidence that no reasonable factfinder would have found the movant guilty of the offense." The statute thus alters the common law miscarriage of justice exception in two ways: (1) by changing the standard from "more likely than not" to "clear and convincing evidence," and (2) by requiring preauthorization from the appropriate court of appeals.

Because § 2255(h) contains a modified miscarriage of justice exception, the district court here could not invoke the common law miscarriage of justice exception. Instead, before bringing his Motion in the district court, Mr. Williams was required by § 2255(h)'s procedures to first obtain certification from this court. Mr. Williams did not obtain certification, and as a result the district court lacked jurisdiction to vacate his convictions to prevent a miscarriage of justice.

In sum, a district court's ability to exercise its inherent powers has been constrained by AEDPA where the court acts on a petitioner's successive motion. Because the district court acted on Mr. Williams's Motion in invoking its powers, it did so in contravention of AEDPA. We therefore reverse the district court's decision for lack of jurisdiction.14

C. We Construe Mr. Williams's Appellate Briefing as a Request to File a Second or Successive Petition, which We Grant in Part.

Having concluded that Mr. Williams's Motion was a successive § 2255 petition and that the district court lacked jurisdiction to grant the requested relief, we now exercise our discretion to treat his appellate brief as a request for authorization to file a successive § 2255 motion. See United States v. Nelson, 465 F.3d 1145, 1149 (10th Cir.2006) (recognizing that we have the discretion to treat a petitioner's appeal and appellate briefing as an implied application for leave to file a successive motion); Spitznas v. Boone, 464 F.3d 1213, 1219 n. 8 (10th Cir.2006) (same). To qualify for permission to file a successive petition, Mr. Williams must present "newly discovered evidence that, if proven and viewed in light of the evidence as a whole, would be sufficient to establish by clear and convincing evidence that no reasonable factfinder would have found the movant guilty of the offense." 28 U.S.C. § 2255(h)(1); see also id. § 2244(b)(3)(C) ("The court of appeals may authorize the filing of a second or successive application only if it determines that the application makes a prima facie showing that the application satisfies the requirements of this subsection.").

In deciding whether the evidence as a whole supports Mr. Williams's factual innocence claim, we "must consider all the evidence, old and new, incriminating and exculpatory, without regard to whether it would necessarily be admitted under rules of admissibility that would govern at trial." House v. Bell, 547 U.S. 518, 538, 126 S.Ct. 2064, 165 L.Ed.2d 1 (2006) (internal quotation marks omitted); United States v. MacDonald, 641 F.3d 596, 612-13 (4th Cir.2011) (adopting the House rule for the factual innocence determination § 2255(h)(1) requires). In engaging in this inquiry, we may also consider the evidence presented to the district court at the evidentiary hearing.15 Each of the parties was represented by counsel and the testimony was taken under oath before the district court.

Considering all of the evidence in the total record, we "must make a probabilistic determination about what reasonable, properly instructed jurors would do. The court's function is not to make an independent factual determination about what likely occurred, but rather to assess the likely impact of the evidence on reasonable jurors." House, 547 U.S. at 538, 126 S.Ct. 2064 (citations and internal quotation marks omitted). Our review "does not require absolute certainty about the petitioner's guilt or innocence." Id.

Because of the fact-intensive nature of this inquiry, we pause to detail the evidence referenced in Mr. Williams's presentence investigation report (PSR), his guilty plea, and the evidence he has introduced attacking his conviction before analyzing whether this evidence entitles Mr. Williams to a certification under § 2255.

1. The Evidence as a Whole

We begin with the evidence contained in the PSR from Mr. Williams's 1998 criminal prosecution. The PSR discusses five searches and seizures that resulted in the evidence against Mr. Williams. Mr. Williams has not challenged two of those searches, which occurred at his residence on September 9, 1996 and January 2, 1997, and revealed methamphetamine, laboratory equipment, and precursor chemicals associated with manufacturing methamphetamine. Mr. Williams and an associate, Eileen Farmer, were arrested during the execution of the search warrant on January 2, 1997.

Mr. Williams has, however, challenged some of the evidence related to searches that took place on March 12 and July 22, 1997. On March 12, police officers conducted a controlled buy and then searched the apartment of Katherine Lanning,16 Mr. Williams's then-girlfriend. According to the PSR, the search revealed 2.5 grams of methamphetamine, a small amount of marijuana, a revolver, and a pager. That same day, police officers discovered Mr. Williams at a nearby parking lot with a disabled vehicle. Following a search of Mr. Williams's person and vehicle, the officers seized 25 grams of methamphetamine, $622, and a pager. Tulsa Police Officer John Gray testified at the state preliminary hearing that Mr. Williams consented to the search. Mr. Williams was arrested at the scene.

The other search Mr. Williams challenges occurred on July 22, 1997, after Mr. Williams was released on bond. On that day, Tulsa Police officers, including Officer Gray and Officer Jeff Henderson, stopped a vehicle driven by Mr. Williams. Ms. Lanning and a third individual, Doyle "Skip" Owens, were in the car with Mr. Williams. Officers discovered methamphetamine, a loaded handgun, and other contraband in the vehicle. Ms. Lanning and Mr. Owens claimed the firearm and the methamphetamine belonged to Mr. Williams. Law enforcement officers again arrested Mr. Williams.

The evidence developed at the 2012 postconviction evidentiary hearing called into question some of the information contained in the 1998 PSR. The district court received testimony from multiple witnesses, four of whom provided particularly relevant information.

First, Ms. Lanning testified at the 2012 hearing that on March 12, 1997, she was home alone and no controlled buy took place at the apartment she shared with Mr. Williams. Ms. Lanning testified that when police officers searched her apartment that day, they did not find methamphetamine. But she claimed that Officer Gray produced a bag of methamphetamine and that she believed he planned to use it to frame Mr. Williams.

Ms. Lanning also described the July 22, 1997, traffic stop by Officers Gray and Henderson. She testified that after the officers stopped the vehicle, Mr. Owens became agitated because he had a firearm tucked into the front of his pants. Mr. Owens pulled the firearm out of his clothing, and officers later recovered it from the vehicle. After the officers recovered the firearm, they asked Ms. Lanning who owned it. When she truthfully reported the firearm belonged to Mr. Owens, Officers Gray and Henderson detained her and continued to question her about the ownership of the firearm. They told her she was in trouble and asked if she wanted to go to prison. The officers indicated she could avoid prison if she said the firearm belonged to Mr. Williams. Ms. Lanning eventually complied; she stated the firearm belonged to Mr. Williams, even though she knew it did not.

Second, Edward Farner, who had served as Officer Gray's informant, testified at the 2012 postconviction evidentiary hearing. He stated Officers Gray, Harold Wells, and Henderson coerced him into working as a confidential informant from 1994 to 1999. At one point, the officers ordered Mr. Farner to target Mr. Williams, first as a potential informant, and later as a potential defendant. Mr. Farner testified that he did not make a controlled buy at the apartment shared by Ms. Lanning and Mr. Williams on March 12, 1997, as represented in the search warrant affidavit.

The next witness, Greg Fillmore, was another informant who worked for Officers Gray and Henderson in Mr. Williams's case. Like Mr. Farner, he claimed the officers coerced him into working for them. Mr. Fillmore testified he was forced by one of the officers to call DEA agents and to lie about having helped Mr. Williams manufacture methamphetamine. Mr. Fillmore also testified that he sensed the DEA agents were expecting his call implicating Mr. Williams. Mr. Fillmore recanted the original statements he had made to the DEA agents, and instead testified that he had never seen Mr. Williams manufacture methamphetamine and had never bought drugs from him. Mr. Fillmore also denied providing evidence that could support the drug quantity attributed to Mr. Williams in the PSR. When asked about written reports DEA Agent Leon Francis made of his interviews with Mr. Fillmore, Mr. Fillmore denied making many of the statements Agent Francis had included in the reports.

Finally, Agent Francis testified at the 2012 hearing. He claimed that his investigation included interviews with approximately seventy witnesses and that these interviews established that Mr. Williams had manufactured methamphetamine and had taught others how to manufacture it. But Agent Francis admitted to a number of irregularities regarding his reliance on Mr. Fillmore as an informant. For example, Agent Francis did not recall how Mr. Fillmore knew to contact the DEA. And he recalled that none of the other seventy witnesses mentioned knowing Mr. Fillmore or reported that Mr. Fillmore was a member of the conspiracy. Rather, Mr. Fillmore's relationship with the DEA began when Mr. Fillmore "walked in off the street" and implicated himself in the drug conspiracy. Notably, Mr. Fillmore did not ask for anything in exchange for his cooperation. In addition, Agent Francis admitted he did not question Mr. Fillmore about his knowledge of the manufacturing process or otherwise make any independent effort to corroborate Mr. Fillmore's story.

2. Application of the § 2255(h) Standard to Mr. Williams's Evidence

Considering the evidence detailed above, we agree with the district court that Mr. Williams has made a prima facie showing of factual innocence with respect to his firearm conviction. But we disagree that Mr. Williams has established actual innocence with respect to his conspiracy and drug possession convictions. We part ways with the district court on the charges other than the firearm conviction, in large part because we must hold Mr. Williams to a higher evidentiary standard than that applied by the district court. As previously explained, the district court considered Mr. Williams's claims through the lens of the common law miscarriage of justice exception, which required the petitioner to establish actual innocence only by a preponderance of the evidence. See McQuiggin v. Perkins, ___ U.S. ___, 133 S.Ct. 1924, 1935, 185 L.Ed.2d 1019 (2013). But to obtain authorization to file a successive petition under AEDPA's framework, Mr. Williams must establish actual innocence by clear and convincing evidence. 28 U.S.C. § 2255(h)(1).

a. Conspiracy Conviction

We turn first to Mr. Williams's conspiracy conviction. Even accepting the district court's conclusion that any evidence obtained from Officers Gray, Henderson, and Wells, including the statements of Mr. Fillmore and Mr. Farner implicating Mr. Williams in the conspiracy, is necessarily unreliable, Mr. Williams has not met his burden. The conspiracy conviction is not supported solely with evidence from these tainted sources. Law enforcement officers also recovered evidence connecting Mr. Williams with a conspiracy to manufacture and distribute methamphetamine when executing search warrants at his residence on September 9, 1996, and January 2, 1997. This included methamphetamine, laboratory equipment, precursor chemicals, scales, and objects containing methamphetamine residue. Cf. United States v. Tolman, 826 F.2d 971, 972-73 (10th Cir.1987) (holding that the possession of chemicals and laboratory equipment necessary to manufacture methamphetamine provides circumstantial evidence that a defendant manufactured methamphetamine). And he was arrested with an associate who manufactured and sold methamphetamine, which supports an inference that he was involved in a conspiracy.

Mr. Williams's prima facie case of factual innocence is also undermined by Agent Francis's testimony and Mr. Williams's plea allocution. Agent Francis testified that he received information from as many as seventy other witnesses, all of whom provided information related to Mr. Williams's involvement in the conspiracy and supported his conclusion that Mr. Williams conspired to manufacture at least 22 kilograms of methamphetamine. These witnesses provided detailed information about the conspiracy's methamphetamine distribution network, and many observed Mr. Williams personally distributing methamphetamine. Mr. Williams has only produced evidence directly contradicting two of these individuals' statements — Mr. Fillmore and Mr. Farner. That leaves dozens of witnesses Agent Francis relied upon in reaching his conclusions. Even if Mr. Williams had shown that all of Agent Francis's reports were unreliable under the preponderance standard the district court applied, he has not demonstrated pervasive problems with the entirety of Agent Francis's investigation that could show, by clear and convincing evidence, that no reasonable juror would credit any of these statements.

Although Mr. Williams has raised serious questions about whether Officers Gray, Wells, and Henderson manufactured false evidence, there is still ample untainted evidence remaining in the record that supports Mr. Williams's conspiracy conviction. See United States v. McCalister, 545 Fed. Appx. 718, 722 (10th Cir.2013) (unpublished) (holding the involvement of Officers Gray and Wells in a defendant's conspiracy case does not per se meet the clear and convincing standard for a successive motion where a defendant has not shown their involvement tainted all the witnesses against him). We hold that Mr. Williams has not met the heavy burden of producing new evidence that, if proven, would show by clear and convincing evidence that no reasonable juror would convict him of conspiring to manufacture and distribute methamphetamine. Therefore, he is not authorized to file a successive § 2255 motion attacking his conspiracy conviction.

b. Drug Possession Convictions

Likewise, Mr. Williams has not asserted actual innocence with respect to his convictions on the drug possession charges. Although he has presented testimony that Officer Gray appeared willing to plant methamphetamine on him and testimony that Officers Gray and Henderson were willing to coerce witnesses into lying about who possessed contraband, Mr. Williams has not claimed that Officers Gray, Henderson, Wells, or others planted methamphetamine on him on March 12 or July 22, 1997. Nor has he claimed that he did not possess methamphetamine on those dates.17 If anything, Mr. Williams's most recent affidavit actually implies his guilt with respect to both counts because he asserts that his sentence should have been calculated using only the methamphetamine seized on those dates as opposed to the methamphetamine attributable to the entire conspiracy. As a result, Mr. Williams has not met his burden of making a prima facie case of actual innocence for drug possession, and we deny authorization to file a successive § 2255 motion attacking those convictions.

c. Firearm Conviction

Finally, we address Mr. Williams's firearm conviction. We are persuaded that Mr. Williams has met his burden of establishing a prima facie case of actual innocence with respect to this conviction. There are four facts that support Mr. Williams's firearm conviction: (1) Officers Gray and Henderson discovered a firearm in Mr. Williams's vehicle; (2) Mr. Owens stated the gun belonged to Mr. Williams; (3) Ms. Lanning gave a statement that the firearm was Mr. Williams's; and (4) Mr. Williams pled guilty to possessing the firearm.

Ms. Lanning's testimony at the 2012 postconviction evidentiary hearing provides exculpatory evidence that, if proven, undermines much of this incriminating evidence. And the district court found Ms. Lanning credible. If her new version of the facts is true, it also undermines Mr. Owens's statement made to Officers Gray and Henderson that the gun belonged to Mr. Williams. According to Ms. Lanning, Mr. Owens became agitated and nervous when the police stopped the vehicle and then removed the gun from his waistband and placed it in the car. A reasonable inference is that he was concerned about facing criminal charges if caught in possession of a firearm. Thus, Mr. Owens had a strong motive to fabricate his statement that the firearm belonged to Mr. Williams.

That the firearm was discovered in Mr. Williams's vehicle and that he pled guilty to possession of it both support an inference that Mr. Williams possessed the firearm. Cf. United States v. Gorman, 312 F.3d 1159, 1164 (10th Cir.2002) (the presence of a firearm in a defendant's vehicle can establish constructive possession); Vanwinkle v. United States, 645 F.3d 365, 370 (6th Cir.2011) (same). But we agree with the district court that Ms. Lanning's recent account of events provides compelling evidence that Mr. Owens owned the firearm. And the district court concluded that Mr. Williams's explanation for entering into the plea agreement — that he was coerced — is credible evidence in light of Officers Gray's and Henderson's pattern of fraudulent activity. Recall that Mr. Farner and Mr. Fillmore testified to being coerced by these officers into making false statements on numerous occasions.

In short, Ms. Lanning's testimony regarding the firearm charge provides the link that is missing in Mr. Williams's attack on his conspiracy and drug convictions. Specifically, Mr. Williams has presented newly discovered evidence that, if proven and viewed in light of the evidence as a whole, would be sufficient to establish by clear and convincing evidence that no reasonable factfinder would have found him guilty of possessing a firearm. Therefore, we authorize Mr. Williams to file of a successive § 2255 motion attacking his firearm conviction.

IV. CONCLUSION

For the foregoing reasons, we REVERSE the district court's judgment and REMAND with instructions to dismiss the case for lack of jurisdiction. We also GRANT Mr. Williams authorization to file with the district court a successive § 2255 motion attacking his conviction for possession of a firearm.

BACHARACH, J., dissenting.

I agree with the majority that (1) Mr. Williams' motion, which alleged fraud on the court, constituted an unauthorized second-or-successive motion, (2) the district court would have lacked jurisdiction to grant this motion, and (3) the district court had the power to sua sponte raise fraud on the court. But I respectfully disagree with the majority in two respects:

Because the district court invoked fraud on the court sua sponte and had the power to do so, I would affirm. Accordingly, I respectfully dissent.

I. Power to Invoke Fraud on the Court Sua Sponte

The judiciary can vacate a judgment sua sponte when it was obtained through a fraud on the court. Hazel-Atlas Glass Co. v. Hartford-Empire Co., 322 U.S. 238, 244-45, 64 S.Ct. 997, 88 L.Ed. 1250 (1944), overruled on other grounds by Standard Oil Co. of Cal. v. United States, 429 U.S. 17, 97 S.Ct. 31, 50 L.Ed.2d 21 (1976). This fraud goes beyond an injury to a party because it injures the "integrity of the judicial process." Id. at 246, 64 S.Ct. 997. Thus, the fraud-on-the-court power does not depend on the filing of a motion and "the court may assert this power sua sponte." United States v. Buck, 281 F.3d 1336, 1341-42 (10th Cir.2002); see Chambers v. NASCO, Inc., 501 U.S. 32, 44, 111 S.Ct. 2123, 115 L.Ed.2d 27 (1991) ("[A] court has the power to conduct an independent investigation in order to determine whether it has been the victim of fraud."); see also Government's Reply Br. at 6 (acknowledging that a court can invoke fraud on the court sua sponte "to protect itself").

In certain contexts, the courts' ability to act sua sponte may be affected by the Antiterrorism and Effective Death Penalty Act's restrictions on second-or-successive motions to vacate a criminal judgment. 28 U.S.C. § 2255(h) (2012); United States v. Nelson, 465 F.3d 1145, 1148 (10th Cir. 2006). Like the majority, I believe the Act would prevent a district court from granting a second-or-successive motion, which relies on fraud on the court, in the absence of certification by our court. See Maj. Op. at 1072-73. As the majority acknowledges, the Act does not apply when the district court acts sua sponte. Id. at 1070-71. In my view, this acknowledgment requires us to affirm: The district court stated that it was acting sua sponte, and we should take the court at its word.

A. Calderon's Interpretive Rule

The majority discounts the district court's characterization based on Calderon v. Thompson, 523 U.S. 538, 553-56, 118 S.Ct. 1489, 140 L.Ed.2d 728 (1998). Maj. Op. at 1073-75. But Calderon does not address a judge's sua sponte invocation of fraud on the court.

Calderon involves a circuit court's ability to recall a mandate sua sponte based on judges' docketing errors. Calderon v. Thompson, 523 U.S. 538, 547-48, 118 S.Ct. 1489, 140 L.Ed.2d 728 (1998). In Part III of the opinion, the Supreme Court addresses whether the recall was proper, discussing the issue in three sections:

Id. at 553-66, 118 S.Ct. 1489.

The first section contains the rule applied by today's majority regarding when a court has acted sua sponte. But the second section disclaims any impact on fraud-on-the-court cases.

Calderon's procedural history helps us understand the first section of the opinion. Mr. Thompson, a state prisoner, filed a habeas petition alleging ineffective assistance of counsel. Id. at 545, 118 S.Ct. 1489. The petition was granted in part. Id. On appeal, a Ninth Circuit panel reversed in part and affirmed in part. Id. at 545-46, 118 S.Ct. 1489.

Mr. Thompson filed a petition for rehearing and a suggestion for rehearing en banc, which was circulated to all active judges. Id. at 546, 118 S.Ct. 1489. The panel issued an order in March 1997, denying the petitions. Id.

In July 1997, Mr. Thompson filed a second motion to recall the mandate, which was based on new evidence. Id. This motion constituted a second-or-successive petition, triggering the Antiterrorism and Effective Death Penalty Act, and the Ninth Circuit denied relief. Id. at 547, 118 S.Ct. 1489.

Two days later, the full court voted to recall its original mandate (issued in March 1997), claiming that it was acting sua sponte rather than in response to the request in July 1997 to recall the mandate. Id. at 547-48, 118 S.Ct. 1489. In deciding to recall the mandate, the court explained that two of the judges failed to request en banc review because of docketing mistakes. Id. at 548, 118 S.Ct. 1489. The court later considered the merits and decided to grant relief. Id. at 548-49, 118 S.Ct. 1489.

The Supreme Court granted certiorari. The Court's first step was to determine whether the Ninth Circuit had jurisdiction to recall the mandate. Id. at 553-54, 118 S.Ct. 1489. The Ninth Circuit's recall of the mandate would have implicated the Antiterrorism and Effective Death Penalty Act if the court's action had been based on Mr. Thompson's filing in July 1997. See id. Thus, to determine whether the Act applied, the Supreme Court had to determine whether the Ninth Circuit was truly acting sua sponte:

Id. at 554, 118 S.Ct. 1489.

This passage announces an interpretive rule for determining when a court is truly acting sua sponte in recalling a mandate. In our case, the majority relies on this passage to conclude that the district court did not act sua sponte because it considered claims and evidence presented in Mr. Williams' second-or-successive motion. See Maj. Op. at 1073-75.

But the reach of the interpretive rule is qualified by the second section of Part III, when the Supreme Court cautioned: "We should be clear about the circumstances we address in this case.... This ... is not a case of fraud upon the court, calling into question the very legitimacy of the judgment." Calderon, 523 U.S. at 557, 118 S.Ct. 1489.

B. Calderon's Impact on Fraud on the Court

The disclaimer in the second section creates a question about the scope of the rule relied on by the majority. Given the context of the disclaimer, two interpretations are possible.

One interpretation is that the disclaimer applies to the entire case, including the first section's analysis of whether the Court of Appeals acted sua sponte. If this interpretation is correct, nothing in Calderon should be applied to a fraud-on-the-court case and the Supreme Court's discussion about when a judge acts sua sponte does not affect our decision.

The second interpretation is that the disclaimer applies only to the rule regarding when an appellate court can recall a mandate sua sponte (the analysis in the second section). If this interpretation is correct

Reasonable minds can differ in interpreting Calderon's disclaimer. For two reasons, however, I believe the Calderon disclaimer applies to the entire case:

For both reasons, I do not believe Calderon's interpretive rule applies when a court states that it is acting to prevent fraud on the court. Thus, we must take the district court at its word when it said it was acting sua sponte.

1. Specificity of the Supreme Court's Language

The Supreme Court stressed that it was dealing with recall of the mandate when explaining the test to determine whether the Ninth Circuit was acting sua sponte:

Id. at 554, 118 S.Ct. 1489 (emphasis added). Through this language, the Supreme Court took great pains to make clear that the sua sponte action it was testing was recall of a mandate, as opposed to something else. The specificity of this language in the first section, combined with the disclaimer in the second section, suggest that the Supreme Court was limiting its interpretive rule to the recall of a mandate.

2. Distinctions Between the Issue in Calderon and this Case

This limitation is imperative here, for the Supreme Court had powerful reasons to distinguish between (1) all courts' ability to sua sponte invoke fraud on the court (which is at issue here) and (2) an appellate court's desire to recall its mandate (which was at issue in Calderon).

The majority points out that when a court is defrauded, the judgment never becomes final. See Maj. Op. at 1071 (stating that courts have historically enjoyed the power to invoke fraud on the court because judgments procured through fraud had never become final); see also Kenner v. Comm'r, 387 F.2d 689, 691 (7th Cir.1968) ("We think ... that it can be reasoned that a decision produced by fraud on the court is not in essence a decision at all, and never becomes final."). Because the judgment never became final, the court can act through its jurisdiction over the original proceeding. Otherwise, the case would continue in perpetuity.

In contrast, an appellate court transfers jurisdiction to the district court by issuing a mandate. See In re Sunset Sales, Inc., 195 F.3d 568, 571 (10th Cir.1999). Accordingly, when a court recalls its mandate sua sponte, it impinges on the finality of the judgment and its relationship to another court in a way that is absent when a court merely invokes its fraud-on-the-court power sua sponte to continue proceedings that had never terminated.

This difference accounts for the Supreme Court's effort to determine whether the Ninth Circuit was acting on the first habeas petition or the second habeas petition. The determination took on importance under the Antiterrorism and Effective Death Penalty Act because the first round of habeas proceedings had already terminated.

Here, we are attempting to determine whether the district court was acting in the original criminal case or the sixth round of federal post-conviction proceedings. As the majority acknowledges, if the original criminal proceedings had been tainted by a fraud on the court, those proceedings were never terminated. See Maj. Op. at 1071 (stating that when there is a fraud on the court, the judgment never became final). Surely the district court had the power to determine whether the original criminal case had ever ended.

Because a fraud on the court prevents termination of the tainted proceedings, no court has ever applied the statutory provisions governing second-or-successive petitions to a judge's power to remedy a fraud on the court. See Gonzalez v. Sec'y for Dep't of Corrs., 366 F.3d 1253, 1275 (11th Cir.2004) (stating that in Calderon, the Supreme Court recognized that a mandate in a habeas case can be recalled when there was a fraud on the court, creating a question about the legitimacy of the judgment); see also Douglas v. Workman, 560 F.3d 1156, 1193-94 (10th Cir.2009) (noting that the Antiterrorism and Effective Death Penalty Act is designed to protect against judicial abuse by the petitioner, not the State's perpetration of a fraud on the court).

In my view, Congress has given little indication of an intent to restrict the established power of a court to act sua sponte to address fraud on the court. In the absence of congressional repeal of this power, a district court continues to enjoy the necessary power to address fraud on itself — regardless of whether the petitioner files a second-or-successive motion to vacate the judgment.

II. The District Court's Characterization of its Action

Having concluded that a court may invoke fraud on the court sua sponte even after the filing of a motion, the resulting issue is whether the district court exercised that power. The majority acknowledges that the district court said it was acting sua sponte. Maj. Op. at 1075 ("Here, the district court claimed it acted sua sponte to correct a fraud on the court."). I would abide by the district court's characterization of its action.

Mr. Williams pleaded guilty to federal criminal charges and filed multiple post-conviction motions. His sixth motion, styled as a motion to withdraw and nullify his guilty plea, led to the district court's concern about a fraud on the court. Ultimately, however, the court invoked its fraud-on-the-court power sua sponte rather than in response to Mr. Williams' motion.

Initially, the district court did not believe Mr. Williams' motion would be considered second-or-successive based on his allegations of a fraud on the court. See Appellant's App., vol. II, at 359 ("[B]ecause Defendant seeks relief based on a fraud upon the court claim, it is distinct from a habeas petition and is not governed by [the Antiterrorism and Effective Death Penalty Act]."). As a result, the court allowed the case to proceed. But during the proceedings, the court changed its approach based on a newly issued opinion in our court: United States v. Baker, 718 F.3d 1204 (10th Cir.2013). There we held that "a motion alleging fraud on the court in a defendant's criminal proceeding must be considered a second-or-successive collateral attack because it asserts or reasserts a challenge to the defendant's underlying conviction." Baker, 718 F.3d at 1207. The district court correctly interpreted Baker to foreclose consideration of Mr. Williams' motion unless we were to authorize a second-or-successive motion under § 2255. Appellant's App., vol. II, at 383 n. 5.

Instead of continuing with the approach now forbidden by Baker, the district court initiated consideration of the issue under its authority to invoke fraud on the court sua sponte. In doing so, the district court distinguished its order from the one in Baker, stating in footnote five: "The instant case is distinguishable due to the fact it is the Court, not the defendant, that has invoked its inherent authority and has construed Williams' Motion to Withdraw and Nullify Guilty Plea as a fraud on the court motion." Id.1

Nothing in the district court's analysis is inconsistent with a sua sponte invocation of fraud on the court. Thus, I would conclude that the court invoked fraud on the court sua sponte, as it had the authority to do.

III. Merits

As the majority notes, the government waived arguments regarding the district court's application of fraud on the court by failing to raise these arguments until the reply brief. Maj. Op. at 1070 n. 9; see Wheeler v. Comm'r, 521 F.3d 1289, 1291 (10th Cir.2008) ("[I]ssues raised by an appellant for the first time on appeal in a reply brief are generally deemed waived.").2 Because I would reject the government's jurisdictional arguments, I would end the analysis there and affirm.

IV. Conclusion

We have long recognized the ability of a court to vacate a judgment procured through a fraud. The district court invoked that power, as it was empowered to do.

We need not decide whether the court correctly perceived the government's conduct as fraudulent because the government waived its arguments on the merits. Instead, the government appealed solely on a jurisdictional ground.

Though the district judge lacked jurisdiction to grant relief based on Mr. Williams' motion, that is not what happened. To the contrary, the district judge exercised his inherent power to sua sponte remedy a fraud on his own court. Accordingly, I would affirm. Because the majority has declined to do so, I respectfully dissent.