United States v. Kirilyuk

Full title: United States of America, Plaintiff-Appellee, v. Ruslan Kirilyuk…

Court: United States Court of Appeals, Ninth Circuit

Date published: Apr 1, 2022

29 F.4th 1128 (9th Cir. 2022)

Opinion

No. 19-10447

04-01-2022

UNITED STATES of America, Plaintiff-Appellee, v. Ruslan KIRILYUK, Defendant-Appellant.

Gene D.Vorobyov (argued), San Francisco, California, for Defendant-Appellant. Matthew G. Morris (argued), Assistant United States Attorney; Camil A. Skipper, Appellate Chief; Phillip A. Talbert, Acting United States Attorney; United States Attorney's Office, Sacramento, California; for Plaintiff-Appellee.

BUMATAY, Circuit Judge

Gene D.Vorobyov (argued), San Francisco, California, for Defendant-Appellant.

Matthew G. Morris (argued), Assistant United States Attorney; Camil A. Skipper, Appellate Chief; Phillip A. Talbert, Acting United States Attorney; United States Attorney's Office, Sacramento, California; for Plaintiff-Appellee.

Before: Daniel A. Bress and Patrick J. Bumatay, Circuit Judges, and Douglas L. Rayes,^* District Judge.

* The Honorable Douglas L. Rayes, United States District Judge for the District of Arizona, sitting by designation.

BUMATAY, Circuit Judge: Ruslan Kirilyuk was sentenced to 27 years' imprisonment for his part in a complex fraud conspiracy spanning multiple countries and involving over 120,000 stolen American Express cards. Kirilyuk was charged and convicted of 28 felonies, the bulk of which were wire and mail fraud counts. To reach the 27-year sentence, the district court relied on multiple sentencing enhancements under the U.S. Sentencing Guidelines ("U.S.S.G.").

In this opinion, we turn our attention to two of these enhancements: (1) the calculation of "loss" as $500 per stolen credit card under U.S.S.G. § 2B1.1(b)(1)(L) and Application Note 3(F)(i); and (2) the two-level enhancement for use of an "authentication feature" under U.S.S.G. § 2B1.1(b)(11)(A)(ii). We conclude that the district court erred in applying the enhancements. We also hold that the district court erred when it imposed a prison term of 264 months for each of the fraud counts—two years above their 240-month statutory maximum. See 18 U.S.C. §§ 1341, 1343. We therefore vacate Kirilyuk's sentence and remand for resentencing.¹

1 In a concurrently filed memorandum disposition, we reject Kirilyuk's remaining objections to his sentence.

I.

For three years, Kirilyuk and his associates engaged in a massive, international fraud scheme. The operation involved layers of sophistication.

First, the conspirators stole account information from nearly 120,000 American Express credit and debit cards.

Second, the group created dozens of fake online businesses using stolen identities and opened merchant accounts for the sham businesses. In setting up these businesses, the enterprise used identities pilfered from three Russian nationals who traveled to the United States on student visas and 220 California high school students whose transcripts had been stolen.

Third, one of Kirilyuk's Russian partners used the false merchant accounts to make fraudulent charges on the stolen AMEX cards. Typically, the charges were in small amounts, between $15 and $30, to avoid detection by the accountholders. The schemers even set up phone lines for the fake businesses to field complaints from AMEX customers seeking refunds for the illicit charges. By refunding the fraudulent charges, they would deter the customers from notifying AMEX.

Next, after being credited for fraudulent charges, the conspirators would transfer the funds from the merchant accounts to nominee bank accounts. They would then withdraw the money from ATMs, wire the funds overseas, or make purchases from other companies.

In total, Kirilyuk's fraud scheme involved more than five conspirators, over 70 shell companies, over 220 stolen identities, almost 120,000 fraud victims, and over 190,000 fraudulent transactions, including 84,000 transfers of funds to fake merchant accounts. Altogether, according to the Probation Office, Kirilyuk and his associates stole over $1.4 million and the intended loss was found to be more than $3.4 million.

By Fall 2015, the FBI unraveled the fraud scheme and arrested Kirilyuk and his co-conspirators. Kirilyuk was indicted on 24 counts of wire fraud, two counts of mail fraud, and one count of aggravated identity theft. See 18 U.S.C. §§ 1028(a)(1), 1341, 1343. Following his arrest and release on bond, Kirilyuk failed to appear for his February 2017 trial and the government added a failure-to-appear charge after he was apprehended in Mexico City, Mexico. See 18 U.S.C. § 3146(a)(1), (b)(1)(A)(i). A jury convicted Kirilyuk on all 28 counts. The presentence investigation report ("PSR") determined that Kirilyuk's offense level reached the maximum of 43. As part of its calculations, the Probation Office recommended several sentencing enhancements and adjustments, including:

• ÷22 levels for the 119,913 AMEX credit card numbers used, multiplied by $500 each, resulting in a loss of $59,956,500, U.S.S.G. § 2B1.1(b)(1)(L) & cmt. n.3(F)(i).

• ÷2 levels for ten or more victims, U.S.S.G. § 2B1.1(b)(2)(A)(i) ;

• ÷2 levels for sophisticated means, U.S.S.G. § 2B1.1(b)(10)(C) ;

• ÷2 levels for the use of an authentication feature, U.S.S.G. § 2B1.1(b)(11)(A)(ii) ;

• ÷4 levels for aggravated role, U.S.S.G. § 3B1.1(a) ;

• ÷2 levels for obstruction of justice, U.S.S.G. § 3C1.1 ; and

• ÷3 levels for commission of an offense while on pretrial release, U.S.S.G. § 3C1.3.

The resulting Guideline range for Kirilyuk's offense level was life, limited by the charges' maximum terms of imprisonment. The maximum term of imprisonment for both wire and mail fraud is 20 years for each count. The identify theft and failure-to-appear counts have maximum terms of two years and ten years, respectively. The district court accepted the PSR's offense level recommendations and sentenced Kirilyuk to 324 months' total imprisonment. The district court's sentence was based on a 264-month, concurrent sentence on each of the wire and mail fraud counts, and consecutive terms of 24 months on the aggravated identity theft count and 36 months on the failure-to-appear count.

On appeal, Kirilyuk only challenges his sentence. We review the district court's interpretation of the Sentencing Guidelines de novo, its application of the Guidelines to the facts of the case for abuse of discretion, and its factual findings for clear error. United States v. Gasca-Ruiz , 852 F.3d 1167, 1170 (9th Cir. 2017) (en banc). We review whether a sentence exceeds the maximum term of imprisonment de novo. United States v. Gementera , 379 F.3d 596, 612 n.5 (9th Cir. 2004).

II.

Kirilyuk appeals his sentence on several grounds. In this opinion, we tackle three issues: (1) whether the district court erred in calculating loss based on a $500-per-card multiplier under Application Note 3(F)(i) to § 2B1.1 ; (2) whether the district court properly applied the "authentication feature" enhancement under § 2B1.1(b)(11)(A)(ii) ; and (3) whether the district court imposed an illegal sentence.

We find merit in all three claims, vacate the sentence, and remand for resentencing.

A. Section 2B1.1's $500-Per-Card Multiplier

We first address Kirilyuk's challenge to the district court's application of Application Note 3(F)(i) of § 2B1.1. The Application Note mandates that "loss" for use of "[c]ounterfeit [c]redit [c]ards" must be calculated at "not less than $500" per credit card used. See U.S.S.G. § 2B1.1 cmt. n.3(F)(i). The effect of the Application Note was enormous. Although Kirilyuk's offense only caused an actual loss of $1.4 million and had an intended loss of only $3.4 million, the Application Note's multiplier skyrocketed the "loss" to nearly $60 million and led to a 22-level enhancement. See U.S.S.G. § 2B1.1(b)(1)(L).

Kirilyuk contends that Application Note 3(F)(i)'s mandatory $500-per-card minimum conflicts with the plain meaning of "loss" under § 2B1.1, and he asks us to find it non-binding under Stinson v. United States , 508 U.S. 36, 38, 113 S.Ct. 1913, 123 L.Ed.2d 598 (1993). We agree and do not consider the Application Note authoritative.

1.

Before turning to the merits of Kirilyuk's claim, we address two important housekeeping issues. First, we look to see whether our prior precedent forecloses Kirilyuk's challenge to Application Note 3(F)(i). Second, we determine whether Kirilyuk properly raised this argument on appeal.

On the question of precedent, we conclude that no Ninth Circuit case has considered whether Application Note 3(F)(i)'s $500-per-card multiplier conflicts with the meaning of "loss" in § 2B1.1. So it remains an open question in our circuit. To be sure, in two published cases, we interpreted and applied Application Note 3(F)(i) or its predecessor. See United States v. Yellowe , 24 F.3d 1110 (9th Cir. 1994) ; United States v. Gainza , 982 F.3d 762 (9th Cir. 2020). But neither case analyzed the Note's validity under Stinson , so neither case binds us on this question.

Prior precedent that does not "squarely address" a particular issue does not bind later panels on the question. Brecht v. Abrahamson , 507 U.S. 619, 631, 113 S.Ct. 1710, 123 L.Ed.2d 353 (1993). As we have repeatedly stated, "[q]uestions which merely lurk in the record, neither brought to the attention of the court nor ruled upon, are not to be considered as having been so decided as to constitute precedents." United States v. Ped , 943 F.3d 427, 434 (9th Cir. 2019) (simplified). Thus, cases are "not precedential for propositions not considered," United States v. Pepe , 895 F.3d 679, 688 (9th Cir. 2018), or for matters that are "simply assumed," Sonner v. Premier Nutrition Corp. , 971 F.3d 834, 842 n.5 (9th Cir. 2020). Indeed, if a prior case does not "raise or consider the implications" of a legal argument, it does "not constrain our analysis." United States v. Cassel , 408 F.3d 622, 633 n.9 (9th Cir. 2005).

In Yellowe , we considered the applicability of Application Note 3(F)(i)'s $500-per-card multiplier's predecessor, former Application Note 4's $100-per-card multiplier, in a particular context.² In his briefing, Yellowe argued that the Application Note's multiplier only applied to offenses involving stolen credit cards , not to cases merely using credit card numbers. When only credit card numbers are used, Yellowe asserted that another Application Note—former Application Note 7—required the court to determine "loss" using "intended loss." See U.S.S.G. § 2F1.1 cmt. n.7 (1993). Yellowe contended that the multiplier's use of a "presumed loss" figure conflicted with the plain meaning of Note 7's "intended loss" requirement. Yellowe also argued that, even if the multiplier applied to credit card numbers, the focus should be on his "knowledge and intentions" since "intended loss" was the goal of the Guidelines and the district court should have applied his estimate of the likely return on the fraud. We rejected each contention in Yellowe.

2 Former Application Note 4 provided that "loss includes any unauthorized charges made with stolen credit cards, but in no event less than $100 per card." U.S.S.G. § 2B1.1 cmt. n.4 (1993). See United States v. King , 861 F.3d 692, 694 n.1 (7th Cir. 2017) (noting that in 2000, the Sentencing Commission moved Application Note 4 to Application Note 3(F)(i) and changed the minimum loss amount from $100 to $500 per card).

Right off the bat, the first sentence of Yellowe acknowledged that the central question in that case was narrow: "This appeal requires us to decide whether Application Note 4 to U.S.S.G. § 2B1.1 ... applies to unauthorized use of credit card numbers as well as the card itself." 24 F.3d at 1111. We held that the $100-per-card multiplier applied equally to credit card numbers and to "unauthorized charges made with the plastic itself." Id. at 1113. Thus, "loss under Application Note 4 to § 2B1.1 includes any unauthorized charges made with stolen credit card numbers or with physical credit cards." Id. Second, we observed that the multiplier set a "presumed loss" for credit card fraud, "setting a floor beneath which neither ‘actual’ nor ‘intended’ loss may fall." Id. Given this floor, we concluded that Yellowe's "subjective intent" in the offense was "immaterial." Id. Finally, we noted that former Application Note 4 said "nothing about probabilities" of success of the fraud and so it wasn't clearly erroneous for the district court to disregard Yellowe's "estimate of the likely return" of his fraud. Id. Those were the precise holdings of Yellowe . Yellowe never argued that—and we never analyzed whether—the $100-per-card multiplier was consistent with the plain meaning of "loss" under Stinson . That is the question we face today.³

3 Our fine dissenting colleague misunderstands Kirilyuk's argument on appeal to conclude that Yellowe governs the matter. Contrary to the dissent's characterization, Kirilyuk is not arguing that "the focus" of the loss inquiry "should be on his knowledge and intentions" since "intended loss" is the "goal of the guideline." Dissent 1143–44. Rather, Kirilyuk argues that the Application Note's use of a $500-per-card multiplier to determine his offense level conflicts with the plain meaning of "loss" in § 2B1.1. Under Stinson , if an Application Note "is inconsistent with" the Guidelines, we must follow the Guidelines. 508 U.S. at 38, 113 S.Ct. 1913.

Gainza is even further afield from the issue in this case. There, we reviewed "[t]he pivotal question" of how to determine "how many account numbers [were] obtained" for Application Note 3(F)(i) purposes. 982 F.3d at 765. We found the district court's finding on the quantity of account numbers clearly erroneous. Id. In doing so, we considered the various types of proof permissible to make the numerical determination and observed such a calculation didn't require "mathematical precision." Id. Nowhere in that decision did we address whether the $500-per-card multiplier aligned with the text of § 2B1.1.⁴

4 The government cites no other published Ninth Circuit decision that purports to bind us on Kirilyuk's argument. We note that several unpublished decisions of this court, including that of Kirilyuk's co-conspirator, Mihran Melkonyan, have also applied the $500 multiplier. But none of these cases analyzed whether the multiplier conflicts with the plain meaning of "loss." See, e.g., United States v. Melkonyan , 831 F. App'x 319, 319–20 & n.1 (9th Cir. 2020) (unpublished); United States v. Chew , 804 F. App'x 492, 494–95 (9th Cir. 2020) (unpublished); United States v. Gaussiran , 787 F. App'x 458, 460 (9th Cir. 2019) (unpublished); United States v. Nguyen , 543 F. App'x 715, 716 (9th Cir. 2013) (unpublished); United States v. Levine , 87 F. App'x 44, 45 (9th Cir. 2004) (unpublished). Contrary to the dissent's suggestion, Melkonyan did not frame his Stinson challenge as Kirilyuk did. Dissent 1147. The only mention of Stinson in Melkonyan's briefing was to note that the $500-per-card multiplier conflicted with Application Notes 3(A) and (C)—not that the multiplier is inconsistent with the meaning of "loss."

So the Stinson issue was neither "brought to the attention of the court nor ruled upon" in Yellowe, Gainza , or any other published Ninth Circuit opinion. Ped , 943 F.3d at 434 (simplified). As no prior panel has "squarely addressed" the issue, we may address its merits. Brecht , 507 U.S. at 631, 113 S.Ct. 1710.

Even if precedent doesn't foreclose reaching the merits, the government still contends that Kirilyuk forfeited his Stinson challenge by not raising it in the district court. This is inaccurate. "[I]t is claims that are deemed waived or forfeited, not arguments." United States v. Lloyd , 807 F.3d 1128, 1174–75 (9th Cir. 2015) (simplified). Before the district court, Kirilyuk specifically objected to the applicability of Application Note 3(F)(1) as "arbitrary," "artificially high," and "contrary to relevant case law and concepts of justice." "Once a federal claim is properly presented, a party can make any argument in support of that claim; parties are not limited to the precise arguments they made below." Id. at 1175 (simplified). Thus, Kirilyuk's sentencing objection was enough to preserve our de novo review of his Stinson challenge.

Finally, the government argues we should not reach the Stinson issue because Kirilyuk didn't raise it until his reply brief. It is true that an appellant generally waives any argument not raised in the opening brief. See Friends of Yosemite Valley v. Kempthorne , 520 F.3d 1024, 1033 (9th Cir. 2008). But we've recognized two exceptions to that rule: (1) when failure to consider the argument would lead to "manifest injustice," and (2) when the "opposing party will not suffer prejudice." Hall v. City of Los Angeles , 697 F.3d 1059, 1071 (9th Cir. 2012). Both exceptions apply here. First, Application Note 3(F)(1) boosted Kirilyuk's base offense level—from ÷ 16 to ÷ 22, see U.S.S.G. § 2B1.1(b)(1)(I), (L), significantly increasing his sentencing range and raising a concern for a manifest injustice. Second, both parties had ample opportunity to address this question in supplemental briefing and so we see no prejudice to the government. We thus exercise our discretion to consider Kirilyuk's Stinson argument and turn to the merits.

2.

As we've recently observed, "Application Notes are not formally part of the Guidelines, but serve to ‘interpret[ ]’ and ‘explain[ ]’ the Guidelines for district courts." United States v. Prien-Pinto , 917 F.3d 1155, 1157 (9th Cir. 2019) (quoting Stinson , 508 U.S. at 38, 113 S.Ct. 1913 ). The U.S. Sentencing Commission drafts both the Guidelines and Application Notes. Id. As Application Notes are based on the Commission's "particular area of concern and expertise" and "represent the most accurate indications of how the Commission deems that the guidelines should be applied," they are generally binding on federal courts. Stinson , 508 U.S. at 45, 113 S.Ct. 1913.

But that is not the end of the story. There's an important distinction between the Guidelines and Application Notes. That's because Congress too has a role over the Guidelines. Id. at 44, 113 S.Ct. 1913. Congress charged the Commission with promulgating the Guidelines and retains the right to review the Guidelines. Id. at 44–45, 113 S.Ct. 1913. So any amendment to the Guidelines must be submitted to Congress for a six-month period of review, during which time Congress can "modify or disapprove them." Id. at 41, 113 S.Ct. 1913. By contrast, "Congress lacks the power to modify or disapprove of Application Notes," which the Commission has unbridled discretion to issue. Prien-Pinto , 917 F.3d at 1157.

Given this difference, the Supreme Court has told us that there is a limit to the binding nature of the Application Notes. Stinson says that an Application Note "that interprets or explains a guideline is authoritative unless it ... is inconsistent with, or a plainly erroneous reading of, that guideline." Id. (quoting Stinson , 508 U.S. at 40, 113 S.Ct. 1913 ). So, "we ascribe somewhat less legal weight to the Application Notes than to the Guidelines proper: if the Guideline and Application Note are inconsistent, the Guideline prevails." Id.

We've been "troubled" by the Commission's prior attempts to use its interpretive authority to improperly change the scope of a Guideline provision. See, e.g., United States v. Crum , 934 F.3d 963, 966 (9th Cir. 2019). And multiple times, we've found Application Notes non-binding for conflicting with the Guidelines. For example, we've applied Stinson to reject an Application Note that altered the "temporal restriction" imposed by the "language of the Guideline." United States v. Rising Sun , 522 F.3d 989, 996 (9th Cir. 2008) (holding that an Application Note can't expand an enhancement for obstruction of justice "during" an investigation to include conduct taking place before an investigation). We've also ruled that the Guidelines commentary need not be followed when it establishes a "narrowing" construction not "found in the Guideline text." United States v. Lambert , 498 F.3d 963, 971 (9th Cir. 2007). Lastly, we've determined that an Application Note can't render a part of the Guidelines "meaningless." United States v. Powell , 6 F.3d 611, 614 (9th Cir. 1993).

With this legal background, we turn to Application Note 3(F)(i)'s $500-per-card multiplier. Section 2B1.1 generally applies to crimes involving theft, stolen property, fraud, and counterfeiting. U.S.S.G. § 2B1.1. The offense level of § 2B1.1 is determined in large part by the crime's "loss" amount. As we've explained, the Guideline "provide[s] for graduated increases to the base offense level depending on the amount of loss caused by the crime." Gainza , 982 F.3d at 764. Subsection (b)(1) of § 2B1.1 establishes a table that increases the base offense by commensurate levels of loss "[i]f the loss exceed[s] $6,500[.]" U.S.S.G. § 2B1.1(b)(1).

Application Note 3(F) provides for "special rules" to "be used to assist in determining loss." Id. cmt. n.3(F). In particular, it provides that in a case involving stolen or counterfeit credit cards, "loss includes any unauthorized charges made with the [credit cards] and shall be not less than $500 per [credit card]." Id. cmt. n.3(F)(i). In other words, Note 3(F)(i) creates a rigid, fictional $500 minimum loss amount per credit card—no matter the facts of the particular case.

The question here is simple: Is Note 3(F)(i)'s "special rule" for calculating loss by using a minimum $500-per-card multiplier consistent with the plain meaning of "loss"? We hold that it is not.

To begin, § 2B1.1 does not define "loss."⁵ In interpreting the Guidelines, we apply the ordinary tools of statutory interpretation and look to the plain meaning of its terms. United States v. Turnipseed , 159 F.3d 383, 387 (9th Cir. 1998). Such tools include "consult[ing] dictionary definitions, which we trust to capture the common contemporary understandings of the word." United States v. Flores , 729 F.3d 910, 914 (9th Cir. 2013).

5 Commentary to § 2B1.1, however, does define "loss." It states that "loss is the greater of actual loss or intended loss," U.S.S.G. § 2B1.1 cmt. n.3(A), with "actual loss" being "the reasonably foreseeable pecuniary harm that resulted from the offense," id. cmt. n.3(A)(i), and "intended loss" consisting of "the pecuniary harm that the defendant purposely sought to inflict," id. cmt. n.3(A)(ii). Given our ruling under Stinson , we need not reach Kirilyuk's alternative argument that Application Note 3(F)(i) conflicts with the Commentary's definitions.

As the Sixth Circuit recently showed, a review of dictionaries reveals that "loss" can have a range of meanings:

One dictionary defines the word to mean, among other things, the "amount of something lost" or the "harm or suffering caused by losing or being lost." American Heritage Dictionary of the English Language 1063 (3d ed. 1992). Another says it can mean "the damage, trouble, disadvantage, [or] deprivation

... caused by losing something" or "the person, thing, or amount lost." Webster's New World College Dictionary 799 (3d ed. 1996). A third defines it as "the being deprived of, or the failure to keep (a possession, appurtenance, right, quality, faculty, or the like)," the "[d]imunition of one's possessions or advantages," or the "detriment or disadvantage involved in being deprived of something[.]" 9 Oxford English Dictionary 37 (2d ed. 1989).

United States v. Riccardi , 989 F.3d 476, 486 (6th Cir. 2021).

Though dictionary definitions for "loss" may vary, they make one thing clear: "No reasonable person would define the ‘loss’ from a stolen [credit] card as an automatic $500" rather than a fact-specific amount. Id. Instead, § 2B1.1 is driven by "the amount of loss caused by the crime. " Gainza , 982 F.3d at 764 (emphasis added). So "loss" cannot mean a pre-determined, contrived amount with no connection to the crime committed, even if it is based on the Commission's "research and data." See U.S.S.G. amend. 596 (Nov. 2000). Application Note (3)(F)(i) thus doesn't illuminate the meaning of "loss," but modifies it. Yet " Stinson requires that commentary interpret the guidelines, not contradict or add to them." Riccardi , 989 F.3d at 493 (Nalbandian, J., concurring).⁶

6 Our dissenting colleague claims that "loss" can also mean "a presumptive dollar amount" so long as it is "intended reasonably to estimate the pecuniary harm resulting from a particular offense." Dissent 1150. But as a matter of plain meaning, that is incorrect, and the dissent provides no support for this contention. When deciding such important matters, we should rely on more than ipse dixit.

This case illustrates the egregious problem with the Application Note's expansion of the meaning of "loss." As determined by the Probation Office, Kirilyuk's conspiracy involved $1.4 million in actual losses or $3.4 million in intended losses. Applying the $500-per-card multiplier balloons the "loss" to $60 million—17 times greater than the intended loss. While the conspiracy was designed to charge only $15 to $30 per credit card, the Application Note asks us to deem each loss to be $500. Application Note 3(F)(i) thus operates as an enhanced punishment , rather than an assessment of "loss" tied to the facts of the case. But Stinson makes clear that the role of the Application Notes is to explain the Guidelines, not enact policy changes to them. We thus hold that Application Note 3(F)(i)'s expansion of the meaning of "loss" is "clearly inconsistent with the language of the Guideline" and is not binding under Stinson . Rising Sun , 522 F.3d at 996.⁷

7 The government argues that we should ignore these concerns because the Commission was only responding to a Congressional directive in amending the Application Note. See U.S.S.G. amend. 596 (Nov. 2000) (noting that the Commission increased the multiplier to $500 after Congress enacted § 2 of the Wireless Telephone Protection Act, Pub. L. 105-172 ). But we decline to create an exception to Stinson based on the Commission's response to a Congressional directive. There may be good reason for the $500-per-card multiplier, but the Commission could put it on firmer ground by adding it to the text of § 2B1.1 itself (as it has with the presumed loss rules of § 2T1.1(c)(1) and (2)). Indeed, as it stands, nothing prevents the Commission from amending the multiplier to a low of $1 or a high of $1 million in the next edition of the Guidelines—all with no say by Congress. Lastly, Application Note 21(C)'s "Downward Departure" safety valve doesn't save the improper expansion of "loss" under Application Note 3(F)(i), as the government argues. Just because a district court has the discretion to lessen the sting of Note 3(F)(i)'s enhanced punishment, that does not make it consistent with the Guidelines under Stinson .

With this holding, we align ourselves with the Sixth Circuit—the only other court to consider this issue. In Riccardi , 989 F.3d 476, the majority of the court held that Application Note 3(F)(i) was not binding, though by applying the narrower deference set out in Kisor v. Wilkie , ––– U.S. ––––, 139 S. Ct. 2400, 204 L.Ed.2d 841 (2019). We do not express a view on that analysis. Instead, our reasoning tracks with Judge Nalbandian's Riccardi concurrence, which relied on Stinson to conclude that "[a]scribing a certain number to ‘loss’ is not a definition." Id. at 493 (Nalbandian, J., concurring). We thus follow our general path of "not creat[ing] a direct conflict with other circuits" in resolving this issue. United States v. Cuevas-Lopez , 934 F.3d 1056, 1067 (9th Cir. 2019) (simplified).

Because Application Note 3(F)(i) contorts the meaning of "loss" to equal "$500" in credit card cases, we hold that it is not binding and that Kirilyuk's 22-level enhancement cannot stand.⁸

8 The dissent warns that our decision will be "far-reaching and destabilizing." Dissent 1157. First, we disagree that our decision is so earthshattering. As we've said, the Commission need only put the $500-per-card multiplier in the Guidelines, rather than in the commentary, to protect it from Stinson scrutiny. Second, even if the dissent is correct, our duty as judges is to apply the law regardless of any disfavored consequences.

B. Section 2B1.1's Authentication Feature Enhancement

We next turn to the enhancement for use of an "authentication feature" under § 2B1.1(b)(11)(A)(ii). We hold that the district court erred in imposing that enhancement because the purported authentication features used here were issued by American Express or a bank, not an "issuing authority" as defined by the Guidelines.

Section 2B1.1(b)(11)(A)(ii) provides for a two-level increase "[i]f the offense involved ... the possession or use of any ... authentication feature." The relevant application note defines "authentication feature" as the term is used in 18 U.S.C. § 1028(d)(1). U.S.S.G. § 2B1.1 cmt. n.10(A). That provision defines "authentication feature" as:

[A]ny hologram, watermark, certification, symbol, code, image, sequence of numbers or letters, or other feature that either individually or in combination with another feature is used by the issuing authority on an identification document, document-making implement, or means of identification to determine if the document is counterfeit, altered, or otherwise falsified.

18 U.S.C. § 1028(d)(1) (emphasis added). "Issuing authority," in turn, is defined as "any governmental entity or agency that is authorized to issue identification documents, means of identification, or authentication features." Id. § 1028(d)(6)(A).

The district court applied the enhancement based on the PSR's reasoning that "credit card numbers, passwords, and bank account[ ] numbers" involved in the scheme constituted authentication features. But the issuers of these authentication features—either American Express or another private financial institution—do not fit within the definition of "issuing authority." Cf. United States v. Sardariani , 754 F.3d 1118, 1121 (9th Cir. 2014) ("issuing authority" includes a notary public, who takes "actions ... based upon the authority of the state"). Here, the government does not allege that American Express issued the credit card numbers, passwords, and bank account numbers "based upon the authority of the state" or any other government entity. Id. So they do not constitute a "governmental entity or agency." 18 U.S.C. § 1028(d)(6). It was thus improper for the district court to apply the enhancement based on the rationale presented in the PSR. The government points out that Kirilyuk's scheme also involved stolen social security numbers, drivers' licenses, and material from public school transcripts. Though true, the government's argument does not carry the day because it is not apparent that the district court relied on these facts to impose the enhancement. On remand, the government is free to re-argue for the authentication feature enhancement on these or any other grounds supported by the record.⁹

9 Because we hold the district court erroneously imposed the authentication feature enhancement, we do not address Kirilyuk's alternative argument that the enhancement also constituted impermissible double-counting under Application Note 2 of § 2B1.6. Kirilyuk may raise that argument, as appropriate, at resentencing.

C. Illegal Sentence

Although not raised by Kirilyuk, the government commendably concedes that the district court imposed an illegal sentence by imposing a 264-month sentence on each of Kirilyuk's wire and mail fraud counts. Both wire and mail fraud carry a maximum penalty of 240 months' imprisonment for each count. See 18 U.S.C. § 1343 (wire fraud); 18 U.S.C. § 1341 (mail fraud). So the government is right that the district court's sentence of 264 months per count was illegal. See United States v. Grimaldo , 993 F.3d 1077, 1083 (9th Cir. 2021) ("An illegal sentence is one in excess of the permissible statutory penalty for [a] crime." (simplified)).

Yet since Kirilyuk failed to raise this issue himself in the district court, we must review whether the illegal sentence constitutes plain error. United States v. Hayat , 710 F.3d 875, 895 (9th Cir. 2013). Under the plain-error standard, relief may be granted only when there was: (1) an error; (2) that was plain; (3) that affected the defendant's substantial rights; and (4) the error seriously affects the fairness, integrity, or public reputation of judicial proceedings. Id. (simplified).

The government admits that the illegal sentence was plain error, satisfying the first two prongs of plain-error review. Once again, we agree with the government. We have held that the "[i]mposition of a sentence exceeding a statutory maximum constitutes plain error." Grimaldo , 993 F.3d at 1083 (simplified).

The government instead counters that Kirilyuk cannot carry his burden under the last two prongs of plain-error review because he was convicted of multiple fraud counts that could be stacked consecutively to impose the district court's total sentence of 264 months. In other words, the district court could have sentenced Kirilyuk to the statutory maximum of 240 months on 25 of his 26 fraud convictions with a 24-month consecutive sentence on the 26th count—totaling 264 months imprisonment.

We disagree. Even if the district court could restructure Kirilyuk's sentence to reach the same result, we decline to decide that for the district court. Kirilyuk cogently argues that the district court may have set a lower sentence had it realized that the maximum sentence on the fraud counts was 240 months. This is especially true now that we are vacating Kirilyuk's sentence based on the error in imposing two other enhancements. As we have said, we should "try to avoid" ruling in a manner that leaves "everyone ... wonder[ing] about whether the sentencing court might have acted differently." United States v. Ameline , 409 F.3d 1073, 1081 (9th Cir. 2005).

The government relies on two cases with limited applicability here. See United States v. Buckland , 289 F.3d 558 (9th Cir. 2002) (en banc); United States v. Kentz , 251 F.3d 835 (9th Cir. 2001). In those cases, we affirmed illegal sentences because "the court could have imposed the sentences on certain counts to run consecutive to one another, there can't possibly be plain error that requires resentencing." Kentz , 251 F.3d at 843 ; see also Buckland , 289 F.3d at 569–70. But both cases were decided before United States v. Booker , 543 U.S. 220, 125 S.Ct. 738, 160 L.Ed.2d 621 (2005), when it was mandatory that district courts apply the Guidelines. Pre-Booker , the Guidelines would have required the district court to impose consecutive sentences to reach the total proper punishment under the Guidelines if it exceeded the statutory maximum on a single count. See Buckland , 289 F.3d at 572 (citing U.S.S.G. § 5G1.2(d) ).¹⁰

10 At the time of Buckland , § 5G1.2(d) provided, "[i]f the sentence imposed on the count carrying the highest statutory maximum is less than the total punishment, then the sentence imposed on one or more of the other counts shall run consecutively, but only to the extent necessary to produce a combined sentence equal to the total punishment." U.S.S.G. § 5G1.2(d) (2001).

Today, however, the district court would have been free to hand down a shorter sentence had it realized the error in the imposed sentence. Thus, it would be "a miscarriage of justice to give [Kirilyuk] an illegal sentence" in this case, United States v. Schopp , 938 F.3d 1053, 1069 (9th Cir. 2019), and we vacate and remand.

III.

The district court erred when it imposed a 22-level enhancement based on the $500-per-card multiplier, applied the two-level enhancement for use of an authentication feature, and handed down an illegal sentence. We vacate Kirilyuk's sentence and remand for resentencing on an open record.

VACATED and REMANDED.

BRESS, Circuit Judge, dissenting:

I respectfully dissent because the majority opinion vastly exceeds the powers of a three-judge panel in overturning circuit precedent that has been the established law of our western states for nearly three decades. In United States v. Yellowe , 24 F.3d 1110 (9th Cir. 1994), we rejected a challenge to Sentencing Guidelines commentary allowing district courts to impose a presumptive dollar-per-credit-card multiplier (then $100, now $500) for calculating "loss" enhancements under the Guidelines for certain fraud offenses. In doing so, we held that courts could impose this loss amount even if a defendant's "actual" or "intended" loss was lower. Since Yellowe , we have repeatedly upheld district courts' application of the per-card multiplier. We recently did so in the case of this defendant's own co-defendant , for his role in the very same credit-card fraud scheme.

Yet today the majority invalidates the per-card multiplier, even though there has been no intervening change in the law since Yellowe . Instead, the majority effectively concludes that our past cases were all wrongly decided and credits what is merely a re-stated version of the same argument we rejected in Yellowe . Under today's decision, Yellowe and many other cases should have come out the other way. Through this serious over-extension, the majority contravenes the fundamental principle that if we think circuit precedent should be revisited, we must engage the en banc process, not take matters into our own hands at the panel level.

Even if three decades of circuit precedent and practice were not enough, the majority's decision is still wrong on its own terms. The Sentencing Commission's per-card multiplier is not inconsistent with or a plainly erroneous reading of "loss" under the Guidelines, and the Commission's interpretation is therefore binding under Stinson v. United States , 508 U.S. 36, 113 S.Ct. 1913, 123 L.Ed.2d 598 (1993). That position is not novel: courts across the country have routinely applied the per-card multiplier, just as we did before today. In siding with an outlier concurring opinion in a Sixth Circuit case involving materially different facts, the majority puts our circuit at odds with the others, to say nothing of our own prior precedent. In the process, the majority disrupts long-established sentencing practices in our district courts—practices we had long approved.

While the majority professes that the defendant's sentence here is "egregious," it is easy to understand why the Sentencing Commission disagreed with the majority's theory of penology. Ruslan Kirilyuk and his co-conspirators stole more than 220 identities, many of them high school students whose confidential information the conspirators lifted from stolen school transcripts. They opened sham bank and merchant accounts in the names of these unsuspecting victims and then used stolen account information for 120,000 American Express cards to make more than 190,000 fraudulent transactions. The Sentencing Commission and district court (which declined to exercise its discretion to impose a lower loss enhancement) could easily conclude that the $1.4 million in completed charges resulting from Kirilyuk's scheme—halted only because authorities discovered it—nowhere near reflects the true scope of his criminality. That, of course, is the judgment Yellowe held the Sentencing Commission and district courts were permitted to reach.

While I join the remainder of the majority opinion, I dissent from that part of the decision invalidating the per-card multiplier.

I

A

I begin where this issue should have ended: the binding force of circuit precedent. Adherence to circuit precedent is not a mere "housekeeping" matter, as the majority would have it. Maj. Op. 1133–34. Three-judge panels must follow circuit precedent except "where the reasoning or theory of our prior circuit authority is clearly irreconcilable with the reasoning or theory of intervening higher authority." Miller v. Gammie , 335 F.3d 889, 893 (9th Cir. 2003) (en banc). "This venerable principle commands our utmost respect and is central to the rule of law in appellate decision-making." Lambert v. Saul , 980 F.3d 1266, 1274 (9th Cir. 2020). En banc review, not panel re-review, is the required mechanism for addressing prior decisions that we believe are wrongly decided. Miller , 335 F.3d at 900. The majority violates these fundamental precepts.

The Sentencing Guidelines provide that, for certain crimes, if "the loss exceeded $6,500," the defendant's offense level should be increased based on the amount of loss. U.S.S.G. § 2B1.1(b)(1). The Guidelines do not themselves define "loss." But in the Guidelines commentary, "loss" is generally defined as "the greater of actual loss or intended loss." Id. § 2B1.1 cmt. n.3(A). "Actual loss" is "the reasonably foreseeable pecuniary harm that resulted from the offense," while "intended loss" generally consists of "the pecuniary harm that the defendant purposely sought to inflict," including harm "that would have been impossible or unlikely to occur." Id. cmt. n.3(A)(ii). Regardless of the method of measurement, "[t]he court need only make a reasonable estimate of the loss." Id. cmt. n.3(C).

The Guidelines commentary provides some special rules "to assist in determining loss" for certain offenses. Id. cmt. n.3(F). In cases involving credit cards, among other "unauthorized access devices," "loss includes any unauthorized charges made with the unauthorized access device and shall not be less than $500 per access device." Id. cmt. n.3(F)(i). The Sentencing Commission originally determined that the loss would presumptively be at least $100 per credit card. Id. cmt. n.4 (1987). The Commission then increased the amount to $500 more than twenty years ago, after Congress directed it to review whether the amount was sufficiently punitive. U.S.S.G. app. C, vol. II, at 57–63. But the commentary makes clear that "[t]here may be cases in which the offense level determined under this guideline substantially overstates the seriousness of the offense," and that "[i]n such cases, a downward departure may be warranted." U.S.S.G. § 2B1.1, cmt. n.21(C).

In United States v. Yellowe , 24 F.3d 1110 (9th Cir. 1994), the defendant pleaded guilty to conspiring to possess and use unauthorized access devices. Id. at 1111. Through a scheme that bears eerie similarity to Kirilyuk's, Yellowe conspired to use thousands of credit card numbers to make fraudulent purchases, routed the charges into fake merchant accounts, and then transferred the funds to his company. Id. at 1111–12.

The district court found Yellowe had misused at least 7,000 credit card numbers and multiplied that number "by the presumed minimum loss of $100 per card" that was then set out in Application Note 4 (now Application Note 3(F)(i)) to U.S.S.G. § 2B1.1. Id. at 1112. The district court thus concluded that Yellowe was responsible for over $700,000 in loss, setting his offense level and Guidelines range accordingly. Id.

We affirmed the district court's application of the $100-per-card multiplier. In doing so, we specifically rejected Yellowe's argument that "the district court misapplied the Guidelines by using the $100 minimum loss mandated when a credit card is used rather than determining the intended loss based on what Yellowe believed the scheme would produce. " Id. at 1112 (emphasis added). The majority here nonetheless adopts the position we rejected in Yellowe , holding that Kirilyuk can be responsible only for the $1.4 million in losses his scheme actually caused (or possibly the $3.4 million in losses he intended to cause, although the majority's reasoning makes even that doubtful, as I explain below). Maj. Op. 1138. In so holding, the majority opinion directly contradicts Yellowe , which confirmed that the "loss" need not be actual or intended loss and may instead be the $100-per-card (now $500-per-card) presumptive loss that Kirilyuk challenges. See 24 F.3d at 1113.

Yellowe also explained why the $100-per-card multiplier was valid. Specifically, because "the value of the unauthorized use exceeds the intrinsic value of the device," "to determine loss based on the value of the card or the number alone would understate the severity of the offense." Id. at 1113. Yellowe 's holding was clear: "We now explicitly hold that loss under Application Note 4 to § 2B1.1 includes any unauthorized charges made with stolen credit card numbers, as well as cards. This loss is a presumed loss, setting a floor beneath which neither ‘actual’ nor ‘intended’ loss may fall." Id. (emphasis in original). Once the per-card multiplier applies and "there is no dispute about the number of stolen numbers," Yellowe held, the defendant's "subjective intent is immaterial. " Id. (emphasis added). We thus rejected Yellowe's argument—which is Kirilyuk's same argument—that "because determining ‘intended loss’ is the goal of the guideline, the focus should be on his knowledge and intentions." The majority claims that Kirilyuk is not making this same argument as Yellowe, Maj. Op. 1135 n.3, but in substance, this is Kirilyuk's entire point. Throughout his briefing, Kirilyuk maintains that the multiplier wrongly imposes a $500 per-card loss "regardless of the actual or intended pecuniary harm, even when (as here) all available evidence is that actual and intended losses are far lower." Kirilyuk Opening Br. 54; see also id. at 57 ("mandatory use of the application note 3, regardless of whether it is consistent with the evidence of actual or intended loss, creates an artificially high loss amount in cases like this one, which does not accurately reflect the seriousness of the crime"). As Kirilyuk informed us, the problem with the per-card multiplier is that it "artificially inflates the amount of the intended loss and, by logical implication, overstates the seriousness of the offense." Id. at 56. There are numerous similar statements throughout Kirilyuk's briefing,¹ confirming that at its core, the argument Kirilyuk makes here is the same one Yellowe made many years ago.

1 See, e.g., id. at 23 (arguing that "the loss amount is grossly disproportionate to the actual or intended loss or Kirilyuk's role in the crime"); id. at 52–53 (arguing that "the district court had to take a realistic, economic approach to determine what losses [Kirilyuk] truly caused or intended to cause" (alteration in original) (quotations omitted)); id. at 53 ("the district court should not have ascribed to Kirilyuk a larger loss tha[n] he inflicted or intended to inflict"); id. at 55 ("Presumed loss completely divorced from the evidence of actual or intended loss would greatly overstate the seriousness of the crime."); Kirilyuk Reply Br. 2 ("there is no evidence that actual or intended loss here can be reasonably estimated as $500 per card"); id. at 3 n.2 ("One of the arguments we made to the $500-per-card multiplier is that it conflicts with § 2B1.1 because it requires imposition of $500-per-card loss, regardless of the losses Kirilyuk did or intended to cause."); id. at 36 (challenging the use of the multiplier "despite lack of evidence of the actual or intended loss being anywhere close to that amount"); Kirilyuk 2d Suppl. Br. 1–2 ("at least when, as here, the actual or intended loss is known to be far lower than $500 per access device, reliance on that commentary violates Stinson "); id. at 19 (arguing that the Stinson argument was preserved because "Kirilyuk objected to the use of the $500-per-access-device rule because it is not based on pecuniary harm (whether actual or intended)").

Yellowe also held that the per-card multiplier applied to credit card numbers in addition to credit cards. The majority seizes on that point to claim that "the central question in [ Yellowe ] was narrow." Maj. Op. 1134. The majority's reading of Yellowe is unduly narrow. Yellowe did hold that the multiplier applies to credit card numbers in addition to cards. But that was not the full extent of Yellowe 's holding. Instead, Yellowe squarely and necessarily held that the per-card multiplier is a permissible application of "loss" under § 2B1.1. 24 F.3d at 1113. The defendant wanted a lower loss amount based on actual or intended loss, and we said "no." This holding had nothing to do with any distinction between credit card numbers and physical credit cards. And in any event, Kirilyuk himself misused credit card numbers and not the card themselves; he is squarely within Yellowe , factually and legally.

The majority is not free to ignore Yellowe simply because Yellowe did not cite Stinson v. United States , 508 U.S. 36, 113 S.Ct. 1913, 123 L.Ed.2d 598 (1993), or perform an analysis specifically tailored to that case. The Stinson argument that Kirilyuk makes now, and that the majority credits, is simply a refreshed version of the argument we rejected in Yellowe . Substantively, the two arguments are the same. Stinson requires us to ask whether the per-card multiplier is "inconsistent with, or a plainly erroneous reading of" the Guidelines. 508 U.S. at 38, 113 S.Ct. 1913. When interpreting the Guidelines, we "will most often begin and end with the structure of the Guidelines," and "may also look to the provision's history and purpose, such as by consulting the Commission's statements of reason for a particular amendment." United States v. Martinez , 870 F.3d 1163, 1166 (9th Cir. 2017) (quotations omitted).

Yellowe undertook that inquiry. Yellowe reviewed the relationship of the per-card multiplier to the Guidelines and to other commentary, as well as the multiplier's purpose—to more closely approximate "the severity of the offense." See 24 F.3d at 1112–13. Yellowe also acknowledged the Guidelines' references to "intended" loss and how " ‘loss need not be determined with precision, and may be inferred from any reasonably reliable information available, including the scope of the operation.’ " Id. at 1112 & n.1 (quoting U.S.S.G. § 2B1.1, cmt. n.3). Yellowe then held that the multiplier was a proper application of "loss," and not a "misappli[cation] [of] the Guidelines," as Yellowe had argued. Id. at 1113. We thus allowed the multiplier notwithstanding Yellowe's argument that "because determining ‘intended loss’ is the goal of the guideline, the focus should be on his knowledge and intentions." Id.

The majority's suggestion that the validity of the per-card multiplier was merely "lurk[ing] in the record" in Yellowe , or "not considered" or merely "assumed" there, is simply inaccurate. Maj. Op. 1134. At issue in Yellowe was whether the multiplier was a permissible application of the Guidelines in a case like this one, where the multiplier resulted in a higher loss amount than the defendant's actual or intended loss. Yellowe upheld the use of the multiplier against that challenge. And contrary to the majority, Yellowe very much "consider[ed] the implications" of this determination, which was to result in a longer sentence for Yellowe and those like him. Maj. Op. 1134.²

2 Perhaps ironically, the majority concludes that Kirilyuk preserved his Stinson argument in the district court, even though he did not couch it in terms of Stinson . Maj. Op. 1137–38. But for the majority, that same reasoning is not sufficient when it comes to interpreting our own circuit precedent. If Stinson is truly a different argument than the one we addressed in Yellowe , the majority should find Kirilyuk's argument forfeited.

Circuit precedent must be read for its holdings and its reasoning, in tandem. "In determining whether we are bound by an earlier decision, we consider not only the rule announced, but also the facts giving rise to the dispute, [and] other rules considered and rejected." In re NCAA Athletic Grant-in-Aid Cap Antitrust Litig. , 958 F.3d 1239, 1253 (9th Cir. 2020) (quotations and alterations omitted). Judicial decisions, conceptual in nature, are not statutes; the omission of a particular word (or here, a case citation) in a judicial opinion does not establish that the case did not hold what it clearly did. Today's decision is directly opposed to Yellowe in its result and reasoning. Under the majority opinion, Yellowe was wrongly decided.

The majority has exceeded its authority. Three-judge panels are "not free to disregard the decision of another panel of our court simply because we think the arguments have been characterized differently or more persuasively by a new litigant." United States v. Ramos-Medina , 706 F.3d 932, 939 (9th Cir. 2013) ; see also Cnty. of San Mateo v. Chevron Corp. , 960 F.3d 586, 597 (9th Cir. 2020), vacated on other grounds , ––– U.S. ––––, 141 S. Ct. 2666, 210 L.Ed.2d 830 (2021) ("Precedents ... do not cease to be authoritative merely because counsel in a later case advances new arguments."). When we have "found a similar argument insufficient" in a prior case, "we are bound by that holding to reach the same conclusion here." Pensinger v. Chappell , 787 F.3d 1014, 1028 (9th Cir. 2015). As one of our colleagues has written, "[t]he fact that a prior panel may not have considered a particular argument, or line of thought, in reaching its bottom line does not provide a valid basis for distinguishing otherwise controlling precedent." United States v. Davis , 428 F.3d 802, 809 (9th Cir. 2005) (Callahan, J., dissenting in part).

Because Yellowe is not "clearly irreconcilable with the reasoning or theory of intervening higher authority," Miller , 335 F.3d at 893, we must follow it. The majority seriously errs in concluding otherwise.

B

Further confirming that Yellowe supplies our rule of decision, numerous cases from this court have upheld applications of the per-card multiplier or have otherwise acknowledged the multiplier as the governing rule. Although the majority gestures in a footnote to a handful of these cases, there are far more than the majority acknowledges. This lengthy set of decisions—which includes that of Kirilyuk's own co-defendant —shows that we certainly have not regarded the validity of the multiplier as the "open question" the majority posits:

• United States v. Chew , 804 F. App'x 492, 494–95 (9th Cir. 2020), relying on Yellowe , affirmed the district court's use of the multiplier. We cited Yellowe as "holding that it was not clearly erroneous for a district court to calculate loss by multiplying the minimum loss calculation by the amount of useable credit cards in the defendant's possession." Id.

• United States v. Gaussiran , 787 F. App'x 458, 460 (9th Cir. 2019), affirmed that the defendant had possessed sufficient useable unauthorized access devices to support the district court's loss calculation, which employed the per-card multiplier. We cited the multiplier rule and concluded that "the district court did not abuse its discretion in including $500 for each card in its calculation." Id.

• United States v. Jackson , 721 F. App'x 631, 633 (9th Cir. 2018), explained that the "plain language of the guidelines indicates there is a floor on each device: the greater of the loss resulting from the unauthorized charges or $500," and affirmed the district court's application of the multiplier.

• United States v. Dobadzhyan , 677 F. App'x 454, 455 (9th Cir. 2017), relied on Yellowe to conclude that courts "may impose a charge of $500 per counterfeit access device number," and held "that the district court did not err by adding $643,500 to the total amount of loss based on the 1,287 access device numbers."

• United States v. Wilburn , 627 F. App'x 659, 659–60 (9th Cir. 2015), affirmed the defendant's sentence where he received a 12-level increase because "calculating loss at $500 per access device," he "possessed at least 471 unique stolen account numbers resulting in an intended loss of $235,000."

• United States v. Masters , 613 F. App'x 618, 621 (9th Cir. 2015), affirmed application of the multiplier and resulting 14-level loss enhancement.

• United States v. Nguyen , 543 F. App'x 715, 716 (9th Cir. 2013), explained that "[f]or crimes involving stolen or counterfeit credits cards and access devices, loss may be calculated at $500 per access device." Citing Yellowe , we noted that the district court "was not required to take into account Nguyen's anticipated likelihood of success using access devices he obtained." Id.

We thus affirmed the defendant's 20-level loss enhancement.

• United States v. Karapetian , 473 F. App'x 603 (9th Cir. 2012), affirmed the defendant's sentence, which was based on the district court's $500-per-card loss calculation.

• United States v. Truong , 587 F.3d 1049, 1051–52 (9th Cir. 2009) (per curiam), rejected the defendant's argument that gift cards did not qualify as access devices, and thus affirmed the district court's sentence, which was based on the per-card multiplier.

• United States v. Camper , 337 F. App'x 631, 632–33 (9th Cir. 2009), relied on Yellowe to conclude that the "district court correctly calculated the total loss by applying the Sentencing Guidelines' $500 presumed loss to each of the 1,531 stolen credit cards." We explained that Yellowe held "that the district court did not clearly err when it calculated loss by multiplying the Sentencing Guidelines' minimum loss figure by the number of workable credit card numbers in Yellowe's possession, even though none of the numbers had been used to purchase items fraudulently." Id. at 633.

• United States v. Levine , 87 F. App'x 44, 45 (9th Cir. 2004), held that the district court did not err "in calculating the total loss by multiplying 2,071 by the $500 minimum loss calculation." We emphasized, citing Yellowe , that the "minimum loss calculation applies regardless of whether the unauthorized credit card was actually used to make fraudulent purchases or not." Id.

• United States v. Nguyen , 81 F.3d 912, 913–15 (9th Cir. 1996), held that blank credit cards qualify as access devices, so the district court had not erred in applying the $100-per-card multiplier. "Because there was no actual loss in this case, each access device was assigned a loss of $100." Id. at 914 (citing U.S.S.G. § 2B1.1, cmt. n.4).

These cases confirm that we have consistently recognized and applied Yellowe 's rule for decades. Under today's decision, however, these cases should have come out differently. And that is to say nothing of the many instances in which district courts applied the multiplier but where the defendant did not appeal that issue, or the case did not reach us at all, because a challenge to the multiplier was so clearly foreclosed by precedent.

But there is more. Remarkably, and though it buries the point in a footnote, the majority invalidates the per-card multiplier even though we recently upheld its application to Kirilyuk's own co-defendant, Mihran Melkonyan. See United States v. Melkonyan , 831 F. App'x 319 (9th Cir. 2020). And we did so even though Melkonyan relied on Stinson 's test.

Melkonyan, like Kirilyuk, received the same 22-level enhancement based on the $500-per-card multiplier for his role in Kirilyuk's same fraudulent scheme. Id. at 319. In his opening brief on appeal, Melkonyan argued that the per-card multiplier "conflicts with the court's role of determining a reasonable estimate of the loss," and that "the court should not be forced to follow the $500 per credit card rule." Melkonyan Opening Brief at 30–31, No. 19-10026, Dkt. No. 12. Just like today's majority opinion, Melkonyan relied on United States v. Rising Sun , 522 F.3d 989, 996 (9th Cir. 2008), a Stinson case, to advance Stinson 's rule: "Application notes like this one are treated as authoritative interpretations of the Sentencing Guidelines unless they violate the Constitution or a federal statute or are inconsistent with, or [are a] plainly erroneous reading of, the Guideline they are meant to interpret." Melkonyan Opening Brief at 31–32 (quoting Rising Sun , 522 F.3d at 996, and noting that Rising Sun was "citing Stinson "). Just like Kirilyuk, Melkonyan argued that he caused approximately $1.5 million in actual losses, so that the multiplier's application resulted in an "inflated figure." Id. at 32.³

3 The majority claims that Melkonyan only invoked Stinson to argue that the per-card multiplier conflicted with other portions of the Guidelines commentary. Maj. Op. 1135 n.4. But Stinson concerns conflicts with the Guidelines themselves. Melkonyan thus argued that his sentence should be based on his actual or intended loss, which is what he claimed should have governed the assessment of how much "loss" he was responsible for under the Guidelines.

In its answering brief, the government responded that Yellowe foreclosed Melkonyan's argument. Answering Brief at 21–27, Dkt. No. 26. The government also relied on Stinson to argue that Application Note 3(F)(i) did not conflict with "the Guidelines section it interprets" or any other Guidelines commentary. Id. at 21–22. And it pointed out that "this Court has repeatedly upheld sentencing courts' use of the $500 valuation to calculate loss." Id. at 24.

We rejected Melkonyan's argument. We quoted Yellowe and reiterated that the per-card multiplier "establishes ‘a presumed loss, setting a floor beneath which neither "actual" nor "intended" loss may fall.’ " Melkonyan , 831 F. App'x at 319 (quoting Yellowe , 24 F.3d at 1113 ). Citing Yellowe , we continued: "Here, because the number of unauthorized access devices is not in dispute, multiplying that number by $500 is the correct application of the Sentencing Guidelines , and the defendant's subjective intent as to actual loss is immaterial. " Id. at 319–20 (citing Yellowe , 24 F.3d at 1113 ) (emphasis added). The majority here not only contradicts Yellowe but treats differently two co-defendants who received identical loss enhancements for the same fraudulent scheme. It is hard to see what justice there is in that, especially when Kirilyuk was the ringleader.

Melkonyan was an unpublished decision. But in these circumstances, that makes the majority's opinion here even more troubling. That the disposition was non-precedential confirms that Melkonyan's Stinson argument was readily resolved based on established law. The same can be said of the other unpublished cases I cited above. In our unpublished dispositions, there should be "no new legal holdings, just applications of established law to facts." Grimm v. City of Portland , 971 F.3d 1060, 1067 (9th Cir. 2020). Indeed, the Melkonyan panel evidently viewed the case as so straightforward that it submitted the matter on the briefs without oral argument. Melkonyan , 831 F. App'x at 319 n.*.

Melkonyan is just further indication of what decades of circuit precedent confirms: that before today, it was settled law that the per-card multiplier is a permissible application of the Guidelines. That the law was so settled likely explains why we do not have even more cases involving such challenges: they would clearly fail, just as every one of them has until now.

II

Even setting aside Yellowe , the majority's holding is incorrect on its own terms. In invalidating the multiplier, the majority badly misapplies Stinson , creates a lopsided split with our sister circuits, and embraces an anomalous concurring opinion from a Sixth Circuit case that involved critically different facts. In the process, the majority converts one possible approach to sentencing policy into a hard legal rule, preventing district courts in appropriate cases from effectuating the Sentencing Commission's judgment that other measures of loss do not adequately capture the seriousness of offenses like the one before us.

A

In 1987, the Sentencing Commission promulgated the first version of the Guidelines Manual, which includes the Guidelines and the Commission's commentary. The Guidelines state that the commentary "may interpret the guideline or explain how it is to be applied." U.S.S.G. § 1B1.7. The Guidelines further make clear that "[f]ailure to follow such commentary could constitute an incorrect application of the guidelines, subjecting the sentence to possible reversal on appeal." Id. Citing the Guidelines' treatment of the commentary and Congress's grant of authority to the Commission, Stinson held that the Guidelines commentary is "authoritative," "controls," and is "binding" on courts. 508 U.S. at 40, 42–43, 113 S.Ct. 1913.

Stinson also carved out narrow situations in which Guidelines commentary is not binding. Specifically, commentary "that interprets or explains a guideline is authoritative unless it violates the Constitution or a federal statute, or is inconsistent with, or a plainly erroneous reading of, that guideline." Stinson , 508 U.S. at 40, 113 S.Ct. 1913. But Stinson clarified that "inconsistent with" means diametrically opposed, where following either the Guidelines or the commentary "will result in violating the dictates of the other. " Id. at 43, 113 S.Ct. 1913 (emphasis added); see also Rising Sun , 522 F.3d at 996 (same). Stinson thus disapproved of courts "refus[ing] to follow commentary in situations falling short of such flat inconsistency." Id. Stinson also recognized that "commentary explains the guidelines and provides concrete guidance as to how even unambiguous guidelines are to be applied in practice." Id. at 44, 113 S.Ct. 1913.

Two circuits, including the Sixth Circuit in United States v. Riccardi , 989 F.3d 476, 483 (6th Cir. 2021), have recently held that courts should now evaluate the validity of Guidelines commentary under the less deferential test set forth in Kisor v. Wilkie , ––– U.S. ––––, 139 S. Ct. 2400, 204 L.Ed.2d 841 (2019). See also United States v. Nasir , 17 F.4th 459, 471–72 (3d Cir. 2021) (en banc). Kisor held that a court can defer to an agency's interpretation of a regulation only after determining that the regulation is "genuinely ambiguous," and only if the agency's interpretation falls "within the zone of ambiguity the court has identified after employing all its interpretive tools." Id. at 2414, 2416.

Other courts have disagreed and have held that Stinson continues to apply to Guidelines commentary. See United States v. Moses , 23 F.4th 347 348-49 (4th Cir. 2022) ; United States v. Cruz-Flores , 799 F. App'x 245, 246 (5th Cir. 2021). The issue is a weighty one because if Kisor were to apply, it "would negate much of the Commission's efforts in providing commentary to fulfill its congressionally designated mission," while leading to "substantial litigation and divisions of authority regarding the extent to which each Guideline is ‘genuinely ambiguous,’ even after ‘all the traditional tools of construction’ have been ‘exhaust[ed].’ " Moses , 23 F.4th at 357 (quoting Kisor , 139 S. Ct. at 2415 ).

The debate over Stinson versus Kisor should be irrelevant in this case because our circuit has continued to apply Stinson to Guidelines commentary after Kisor . See, e.g., United States v. Herrera , 974 F.3d 1040, 1047 (9th Cir. 2020) ; United States v. George , 949 F.3d 1181, 1185 (9th Cir. 2020) ; United States v. Wang , 944 F.3d 1081, 1086 (9th Cir. 2019) ; United States v. Cuevas-Lopez , 934 F.3d 1056, 1061 (9th Cir. 2019) ; United States v. Crum , 934 F.3d 963, 966 (9th Cir. 2019). In this circuit, Stinson is still the governing law for evaluating Guidelines commentary. The majority here thus purports to apply Stinson while disclaiming any position on whether Kisor should be the right test. Maj. Op. 1138–39. As we will see, however, the majority's application of Stinson is nothing of the sort. And its refusal to acknowledge legitimate and long-applied commentary is just Kisor in disguise.

The $500-per-card multiplier easily satisfies Stinson . No one suggests that the multiplier violates the Constitution or a federal statute. See Stinson , 508 U.S. at 40, 113 S.Ct. 1913. Nor is the multiplier "inconsistent with, or a plainly erroneous reading of" the Guidelines. See id. The Guidelines create graduated offense level increases based on the amount of "the loss." U.S.S.G. § 2B1.1(b)(1). "Loss" is the only operative word here. But the Guidelines do not define what "loss" means, how to calculate it, or the precision by which a "loss" amount should be assessed.

"Loss" could mean "actual" out-of-pocket loss, but it could also mean "intended" loss, even if the intended loss "would have been impossible or unlikely to occur." Id. § 2B1.1 cmt. n.3(A)(ii) (listing as an example a government sting operation); see also United States v. Popov , 742 F.3d 911, 915 (9th Cir. 2014). "Loss" could even mean "the gain that resulted from the offense," provided "there is a loss but it reasonably cannot be determined." Id. § 2B1.1 cmt. n.3(b); see also United States v. Martin , 796 F.3d 1101, 1111 (9th Cir. 2015) ("[D]istrict courts may use the defendant's gain as another way to measure the loss."). None of these ways of viewing "loss" is inconsistent with or a plainly erroneous reading of the Guidelines. See Stinson , 508 U.S. at 38, 113 S.Ct. 1913. The Commentary also states that "[t]he court need only make a reasonable estimate of the loss." Id. cmt. n.3(C). We have long recognized that this too is appropriate. See, e.g., United States v. Tadios , 822 F.3d 501, 503 (9th Cir. 2016) ; United States v. Armstead , 552 F.3d 769, 778 (9th Cir. 2008) ; United States v. Lopez , 64 F.3d 1425, 1427 (9th Cir. 1995).

The per-card multiplier is just an offense-specific application of "loss" under § 2B1.1, consistent with the "reasonable estimate" requirements. See Yellowe , 24 F.3d at 1112–13. Just as "loss" can mean actual or intended loss—two concepts that are not specifically stated in the Guidelines either—it can mean a presumptive dollar amount that is intended reasonably to estimate the pecuniary harm resulting from a particular offense. Nothing in the word "loss" prohibits the Sentencing Commission from assigning a presumptive monetary value to a given act of misconduct based on the Commission's experience with that crime. See id. at 1113 (explaining that loss can be "a presumed loss, setting a floor beneath which neither ‘actual’ nor ‘intended’ loss may fall"). Thus, following the per-card multiplier in the commentary does not lead the sentencer to "violate the dictates" of the Guidelines. Stinson , 508 U.S. at 43, 113 S.Ct. 1913.

The history of the multiplier supports this. The original $100-per-card multiplier was included in the first iteration of the Guidelines commentary. Riccardi , 989 F.3d at 482. Prior to taking effect, the first Guidelines Manual was made available for public comment and congressional review. See Moses , 23 F.4th at 353-54. The original $100-per-card multiplier remained unchanged until 2000, when, in response to a directive from Congress, see Identity Theft and Assumption Deterrence Act of 1998, Pub. L. 105-318, § 4(a), 112 Stat. 3007, 3009 (1998); Wireless Telephone Protection Act, Pub. L. 105-172, § 2(e)(1), 112 Stat. 53, 55 (1998), the Sentencing Commission adopted Amendment 596, which increased the multiplier to $500. U.S.S.G. app. C, vol. II, at 57–63. In Amendment 596, the Commission explained that Congress had directed it to "review the extent to which the value of the loss caused by the offenses is an adequate measure of establishing penalties." Id. at 62 (quotations and alterations omitted). The Commission concluded that its "research and data supported increasing the minimum loss amount ... from $100 to $500 per access device." Id.

This history shows that the Commission aimed to approximate the loss associated with an offense like Kirilyuk's. Under Stinson , nothing in the word "loss" precluded it from doing so. And using a presumptive dollar value per card has the benefit of promoting uniformity in sentencing, making it more likely that defendants who commit credit-card fraud offenses will receive similar loss enhancements.

B

The majority nevertheless claims the multiplier is "clearly inconsistent" with the Guidelines. Maj. Op. 1138–39. This holding is deeply flawed. It would of course be surprising to learn that in the decades since Yellowe , numerous judges applying the per-card multiplier at the trial and appellate levels have unwittingly enforced a facially improper application of "loss." Fortunately, that is not the case. Guidelines commentary is impermissibly inconsistent with the Guidelines only when "following one will violate the dictates of the other." Stinson , 508 U.S. at 43, 113 S.Ct. 1913. And there is no such "flat inconsistency" here. Id.

On this point, I recite the majority's core reasoning in full. The majority quotes the discussion of dictionary definitions of "loss" in Riccardi and then says:

Though dictionary definitions for "loss" may vary, they make one thing clear: "No reasonable person would define the ‘loss’ from a stolen [credit] card as an automatic $500" rather than a fact-specific amount. Id. Instead, § 2B1.1 is driven by "the amount of loss caused by the crime. " Gainza , 982 F.3d at 764 (emphasis added). So "loss" cannot mean a pre-determined, contrived amount with no connection to the crime committed, even if it is based on the Commission's "research and data." Application Note 3(F)(i) thus doesn't illuminate the meaning of ‘loss,’ but modifies it.

Maj. Op. 1138 (brackets and alterations in original; the "Id. " citation is of Riccardi ). Unpacking each piece of this is critical for appreciating the majority's error.

First , the majority's reliance on Riccardi raises obvious red flags because the Sixth Circuit in that case invalidated the $500-per-card multiplier under Kisor , not Stinson . See Riccardi , 989 F.3d at 486. Invoking Riccardi 's Kisor -based discussion of dictionary definitions and what a "reasonable person" would think, the majority asserts that "loss" must be "the amount of loss caused by the crime ," Maj. Op. 1137(quoting Gainza , 982 F.3d at 764 ) (emphasis in majority opinion), by which the majority apparently means the actual loss associated with a crime.

But this is not a Stinson analysis: the question here is not whether the word "loss" is ambiguous in some sense but whether the multiplier is flatly inconsistent with the Guidelines. Stinson , 508 U.S. at 43, 113 S.Ct. 1913 ; Rising Sun , 522 F.3d at 996. The majority says that " ‘loss’ cannot mean a pre-determined" amount. Maj. Op. 1138. But under Stinson , where is the flat inconsistency? Nothing in the word "loss" or the Guidelines more generally prevents the Commissioner from ascribing a presumptive loss value to a particular offense, much less creates a situation in which applying the commentary "violates the dictates of" the Guidelines. Stinson , 508 U.S. at 43, 113 S.Ct. 1913 ; see also Yellowe , 24 F.3d at 1112–13.

Not only does the majority fail to apply a true Stinson analysis, the majority's assertion that we are required to invalidate the per-card multiplier based on the "plain meaning" of "loss" is simply unfounded. Maj. Op. 1137–38 & n.6. The unadorned and undefined word "loss" does not remotely demand the majority's wooden reading, especially in the context of the Guidelines as a whole, which are designed to ensure punishments that reflect a defendant's criminality. U.S.S.G. Ch. 1, Pt. A, intro., 3.

The majority's supposed "plain meaning" approach would also call into question many other established measures of "loss." As I have explained above, the Commentary has long treated "loss" as including "intended" loss, U.S.S.G. § 2B1.1 cmt. n.3(A)(ii), among many other offense-specific interpretations of "loss," id. § 2B1.1 cmt. n.3(A), (F). "Intended" loss also does not reflect "loss caused by the crime. " Maj. Op. 1138. But we have repeatedly treated this and other measures of "loss" as valid. See, e.g., Popov , 742 F.3d at 915 ; Martin , 796 F.3d at 1111. These other longstanding measures of "loss" would also now appear vulnerable under the majority's improperly narrow reading of "loss."

The majority's repeated assertions that the Commission is "modifying" the Guidelines or making "policy" judgments prove nothing. Maj. Op. 1138. Many aspects of the Guidelines and commentary reflect policy judgments; that does not make them unlawful under Stinson . And here, the commentary expressly notes that if the offense level "substantially overstates the seriousness of the offense," "a downward departure may be warranted." U.S.S.G. § 2B1.1, cmt. n.21(C). I fail to see how the commentary squarely violates the dictates of the Guidelines when it allows this flexibility.

Of course, for all its reliance on Riccardi , the majority alters its key quotation of that case by adding the word "credit" in brackets in place of "gift" card. Quoting Riccardi , the majority says: " ‘No reasonable person would define the "loss" from a stolen [credit] card as an automatic $500’ rather than a fact-specific amount." Maj. Op. 1138 (quoting Riccardi , 989 F.3d at 486 ) (emphasis added). Riccardi involved gift cards, not credit cards. As I will explain further below, gift cards present a very different situation than credit cards because they have a finite face value and their theft produces fewer collateral costs. But suffice to say, Riccardi had no occasion to extend its Kisor analysis to credit cards. And the Sixth Circuit has since implied that Riccardi may not apply to credit cards. See United States v. Nicolescu , 17 F.4th 706, 720 (6th Cir. 2021) ("even if this court's recent decision in United States v. Riccardi renders invalid any loss calculation based on a $500-per-stolen-credit-card multiplier ...."). By editing the quotation, the majority implies that Riccardi sweeps broader than the Sixth Circuit thus far has recognized.

Second , the majority purports to derive its "loss caused by the crime" test from United States v. Gainza , 982 F.3d 762 (9th Cir. 2020). But the majority quotes a stray phrase in that case out of context. The language the majority quotes comes from this sentence: "For economic crimes, the Sentencing Guidelines provide for graduated increases to the base offense level depending on the amount of loss caused by the crime." Gainza , 982 F.3d at 764. Nothing in this generic language in Gainza or the case as a whole purported to foreclose application of the per-card multiplier or any other established measure of "loss," much less adopt the majority's unnecessarily cramped "plain meaning" interpretation of that term.

In fact, the next sentence of Gainza states: "loss includes any unauthorized charges made with the unauthorized access device and shall be not less than $500 per access device." Id. (quoting U.S.S.G. § 2B1.1 cmt. n.3(F)(i) (alterations omitted)). Gainza thus recognized the very rule the majority invalidates here. We even relied on Gainza in affirming Kirilyuk's co-defendant's sentence. There, we described Gainza as a case "referring with approval to the court's use of the $500 minimum per access device found in Application Note 3(F)(i) to determine amount of loss." Melkonyan , 831 F. App'x at 319 n.1 (emphasis added).

Third , the majority is improperly dismissive of the Commission's guidance in claiming that the commentary's approach to "loss" is a "contrived" one "with no connection to the crime committed." Maj. Op. 1138. That is a severe mischaracterization. After Congress directed the Commission to reevaluate its penalty provisions, the Commission's Economic Crimes Policy Team produced two reports, which are publicly available, discussing possible changes to the loss amount. See Econ. Crimes Pol'y Team, U.S. Sent'g Comm'n, Cellular Phone Cloning Final Report, at 27 (Jan. 25, 2000); Econ. Crimes Pol'y Team, U.S. Sent'g Comm'n, Identity Theft Final Report, at 23 (Dec. 15, 1999); see also Riccardi , 989 F.3d at 482–83.

In one report, the Commission explained that the Department of Treasury had recommended increasing the presumptive loss per card to $1,000. Cellular Phone Cloning Final Report at 27 & n.47. Treasury had "cited credit card industry data that showed the average fraud loss in 1998 to be $1,040.59 per credit card." Id. at 27. The Commission further recounted how "Treasury also cited 1999 Secret Service statistics indicating an average fraud loss per credit card of $2,218." Id. These amounts were lower than the Commission's own estimate of $3,775, which was based on "a sample of 109 federally sentenced credit card fraud cases for which the exact number of credit cards and exact amount of charges were known." Id. at 27 & n.48. These loss figures considered only the fraudulent charges made on credit cards, not other pecuniary harms suffered by the cardholder or financial institutions.

In another report, the Commission acknowledged that the Guidelines arguably did "not provide adequate punishment" because they did "not provide for consideration of indirect monetary harms to the individual victims, such as the costs incurred in attempting to repair damaged credit ratings." Identity Theft Final Report at 23. This report cited the egregious example of an identity theft victim who spent nearly a year, "500 hours of her time," and "incurred out-of-pocket costs of approximately $10,000" as a result of the offense. Id. at 23 n.37. The report also contemplated that "loss," as it stood then, might not reflect "harm to an individual's financial reputation, as well as the ensuing inconvenience" of identity theft. Id. at 24. In raising the per-card loss multiplier from $100, the Commission thus considered not only average fraudulent charges per card, but also other associated losses.

The Commission then solicited comments identifying three potential alternatives: $500, $750, and $1,000. Sentencing Guidelines for United States Courts, 65 Fed. Reg. 2663, 2665–66 (Jan. 18, 2000). It received comments from, among others, the Department of Justice, a federal defender organization, and Treasury. See U.S. Sent'g Comm'n, Public Comment from March 2000, Part I, Amendment 6. The Commission ultimately went with the lowest amount, explaining in Amendment 596 that its "research and data supported increasing the minimum loss amount ... from $100 to $500 per access device." U.S.S.G. app. C, vol. II, at 62. Again, this was substantially lower than the average loss per card from the Commission's own data-set, as well as the information provided by Treasury and the Secret Service.

The Commission provided notice of the amended per-card multiplier to Congress for its review. Sentencing Guidelines for United States Courts, 65 Fed. Reg. 26,880, 26,895 (May 9, 2000). Congress did not take further action and the amendment went into effect later that year. All these events belie the majority's castigation of the multiplier as "contrived." And they confirm what we said in Yellowe : that other measures of loss "would understate the severity of the offense." 24 F.3d at 1113.⁴

4 The majority's suggestion that Congress "lacks the power to modify or disapprove of Application Notes," Maj. Op. 1136, is incorrect. The majority relies on United States v. Prien-Pinto , 917 F.3d 1155, 1157 (9th Cir. 2019), which in turn cited Stinson 's statement that "Congress does not review amendments to the Commentary," 508 U.S. at 40, 113 S.Ct. 1913. That the Sentencing Commission is not required to submit amendments to the commentary to Congress does not deprive Congress of the power to modify or disapprove of them. And here, the Commission did submit the amendment to Congress. 65 Fed. Reg. at 26,895.

In a footnote, the majority suggests that "nothing prevents the Commission from amending the multiplier to a low of $1 or a high of $1 million." Maj. Op. 1138 n.7. This is pure hyperbole. In the decades since the per-card multiplier was adopted, the Sentencing Commission has used only two amounts: $100, and then, when prompted to revisit that amount by Congress, $500. The Sentencing Commission has not entertained an increase in the multiplier on anything like the scale the majority imagines.

And there would of course be constraints on the Commission's ability to set the multiplier at $1 million per card. That preposterous number—which would far exceed any credit card spending limits of which I am aware—would not reflect a "reasonable estimate of the loss." U.S.S.G. § 2B1.1 cmt. n.3(c). It would also presumably not be supported by any data; would not be consistent with the 18 U.S.C. § 3553(a) factors, see U.S.S.G. § 1B1.1 (directing courts to consider § 3553(a) ); and could implicate other due process and statutory limits. The majority's undeveloped attempt to claim the Sentencing Commission could run roughshod in this area has no basis in law or reality.

The upshot is the following: applying a presumptive $500-per-card multiplier does not require us to "violate the dictates of" the Guidelines. Stinson , U.S. at 43, 113 S.Ct. 1913. This case thus bears no resemblance to the few cases the majority cites invalidating commentary under Stinson , because each involved a direct conflict between a Guideline and application note. In Rising Sun , for instance, the relevant guideline provided a sentencing adjustment if the defendant obstructed justice during an investigation or prosecution, and we held that commentary suggesting obstruction could occur before the start of an investigation was "clearly inconsistent" with the guideline. Rising Sun , 522 F.3d at 995–96. Our decision in United States v. Lambert , 498 F.3d 963, 971 (9th Cir. 2007), involved a similarly square conflict. And the statement the majority relies on from United States v. Powell , 6 F.3d 611, 614 (9th Cir. 1993) —that we cannot render part of the Guidelines "meaningless"—is inapposite. The per-card multiplier does not render § 2B1.1 "meaningless."

The majority identifies no case supporting its Kisor -esque application of Stinson , which in any event directly conflicts with the holding and reasoning of Yellowe .

C

In disregarding our own law, the majority claims it is avoiding a circuit split. Maj. Op. 1138–39. Untrue. The state of the law is this: until now, no court has rejected the per-card multiplier under Stinson ; our sister circuits have routinely applied it, just as we did before today; and the only circuit to have rejected it (the Sixth) did so under Kisor and in a case involving gift cards, not credit cards.

The Seventh Circuit, for instance, has held that the $500-per-card multiplier applies to all access devices a defendant possesses. See United States v. Moore , 788 F.3d 693, 695 (7th Cir. 2015). Moore explained that the "commentary following the guideline is an authoritative interpretive aid for how the guideline should be applied," and held that the plain text of Application Note 3(F)(i) establishes "that the $500 per unauthorized access device amount of loss ... applies to all unauthorized access devices in a case." 788 F.3d at 695. The Eighth Circuit reached the same conclusion, holding that a district court did not err in applying the $500-per-card multiplier to counterfeit credit cards, regardless of whether the defendant used the card. United States v. Thomas , 841 F.3d 760, 763–65 (8th Cir. 2016).

The First Circuit similarly rejected a challenge to the $500-per-card multiplier, explaining that "Application Note 3(F)(i) provides that loss both (1) shall include any unauthorized charges made with the counterfeit access device or unauthorized access device and (2) shall be not less than $500 per access device regardless of whether each access device was actually charged." United States v. Rueda , 933 F.3d 6, 9–11 (1st Cir. 2019) (quotations and alterations omitted). The First Circuit thus rejected the defendant's argument "that the loss attributable to her offense should not be the $1,290,000 calculated by the District Court," but "the $24,673.60 that was reflected in the victim impact statement," which related to actual wrongful charges. Id. at 10. Under the majority's reasoning, the First, Seventh, and Eighth Circuit decisions would have to come out the other way.

Indeed, these and many other circuits—in total, the First, Second, Third, Fourth, Fifth, Seventh, Eighth, Tenth, and Eleventh—have repeatedly applied the per-card multiplier post- Stinson , recognizing it as the governing rule. See, e.g., United States v. Acevedo , 860 F. App'x 604, 612–13 (11th Cir. 2021) ; United States v. Graveran-Palacios , 835 F. App'x 436, 445 (11th Cir. 2020) ; United States v. Carver , 916 F.3d 398, 402 (4th Cir. 2019) ; United States v. Sosa , 773 F. App'x 140, 140–41 (4th Cir. 2019) ; United States v. Fleitas , 766 F. App'x 805, 807–08 (11th Cir. 2019) ; United States v. Maitre , 898 F.3d 1151, 1159–61 (11th Cir. 2018) ; United States v. Delima , 886 F.3d 64, 72 (1st Cir. 2018) ; United States v. Garcia , 727 F. App'x 599, 602 (11th Cir. 2018) ; United States v. Nelson , 724 F. App'x 814, 818–19 (11th Cir. 2018) ; United States v. Popovski , 872 F.3d 552, 553 (7th Cir. 2017) ; United States v. Wright , 862 F.3d 1265, 1274–75 (11th Cir. 2017) ; United States v. Bendelladj , 710 F. App'x 384, 386–87 (11th Cir. 2017) ; United States v. Haywood , 681 F. App'x 290, 292 (4th Cir. 2017) ; United States v. Garcia , 634 F. App'x 242, 244–45 (11th Cir. 2015) ; United States v. Balde , 616 F. App'x 578, 583–84 (4th Cir. 2015) ; United States v. Cardenas , 598 F. App'x 264, 266–67 (5th Cir. 2015) ; United States v. Nelson , 597 F. App'x 17, 18 (2d Cir. 2015) ; United States v. Mendez , 589 F. App'x 642, 645 (4th Cir. 2014) ; United States v. Torres-Bonilla , 556 F. App'x 875, 882 (11th Cir. 2014) ; United States v. Bermudez , 536 F. App'x 869, 871 (11th Cir. 2013) ; United States v. Miralles , 521 F. App'x 837, 839–40 (11th Cir. 2013) ; United States v. Volynskiy , 431 F. App'x 8, 9–10 (2d Cir. 2011) ; United States v. Heath , 424 F. App'x 730, 735–37 (10th Cir. 2011) ; United States v. Harris , 597 F.3d 242, 249 (5th Cir. 2010) ; United States v. Warthen , 390 F. App'x 977, 978 n.1, 981 n.3 (11th Cir. 2010) ; United States v. Dodson , 357 F. App'x 324, 325–26 (2d Cir. 2009) ; United States v. Jones , 332 F. App'x 801, 804 n.1, 807 (3d Cir. 2009) ; United States v. Kowal , 527 F.3d 741, 748 (8th Cir. 2008) ; United States v. Lewis , 312 F. App'x 515, 518 (4th Cir. 2008) ; United States v. Carralero , 195 F. App'x 874, 877–78 (11th Cir. 2006) ; United States v. Muhammed , 108 F. App'x 775, 777 (4th Cir. 2004) ; United States v. Scott , 250 F.3d 550, 551–53 (7th Cir. 2001) ; United States v. Sowels , 998 F.2d 249, 252 (5th Cir. 1993).

The only circuit that has taken a different approach is the Sixth Circuit in Riccardi . But the majority here cannot purport to "align [itself] with the Sixth Circuit," Maj. Op. 1138, when that court applied Kisor , not Stinson . See Riccardi , 989 F.3d at 484–85. And as I flagged above, Riccardi involved a critically different set of facts: the defendant was charged with stealing 1,505 gift cards from the mail. Id. at 479. This included, for example, a $25 Starbucks gift card, with the gift cards having an average face value of $35 each. Id. at 479–80.

Although the Sixth Circuit concluded that the gift cards were "access devices" subject to the $500-per-card multiplier, id. at 479, 482–83, there is a world of difference between gift cards and credit cards. Charges on a gift card max out at the value of the card, whereas much more can be charged to a credit card than $35. The collateral damage also differs substantially. When a gift card is stolen, there may be some costs associated with replacing it. But those amounts surely pale in comparison to the costs associated with credit card fraud, which are borne by cardholders and financial institutions alike. See United States v. Pham , 545 F.3d 712, 721 (9th Cir. 2008) (identifying some of these costs). In purporting to associate our court with the Sixth Circuit, the majority thus relies on a case that involved materially different facts.

Of course, when the majority says it is aligning with the Sixth Circuit, what it means is that it is siding with a separate opinion in Riccardi that did not command a majority there. See Riccardi , 989 F.3d at 490–93 (Nalbandian, J., concurring). Judge Nalbandian believed Stinson still applied to Guidelines commentary but would invalidate the per-card multiplier under Stinson . Id. at 492, 113 S.Ct. 1913. Until today, no court had adopted that position. In aligning our circuit with a novel separate writing in this area of law, the majority puts us at odds with both our own precedent and that of every other circuit.

D

Finally, the majority cannot justify its departure from settled law by claiming that Kirilyuk's sentence is an "egregious problem" considering that the amounts he actually charged to credit cards were much lower than the presumptive loss amount under the per-card multiplier. Maj. Op. 1138. Here too the majority misperceives our role, enshrining into law one possible perspective on sentencing policy rather than respecting the judgment of the Sentencing Commission and the experienced district court judge who conducted Kirilyuk's sentencing.

As the majority itself explains, "Kirilyuk and his associates engaged in a massive, international fraud scheme" that "involved layers of sophistication." Maj. Op. 1132. Working with Russian partners and with Kirilyuk in the lead, the co-conspirators created sham merchant accounts and linked bank accounts using stolen identities, including those of 220 Sacramento high school students through their stolen school transcripts. The merchant accounts had names like "Best Box," "Chevran," and "CVS Store," designed to look close enough to real companies to avoid detection.

Then, using nearly 120,000 stolen credit card numbers obtained by Russian hackers, the co-conspirators made over 190,000 fraudulent charges to the fake merchant accounts, routed the money to the sham bank accounts, and then withdrew it from the banks or wired it overseas. They even set up phone numbers for the fake businesses, so that when an American Express cardholder saw a suspicious charge and called the number, the co-conspirators could remove the charge and prolong their scheme. It took the combined efforts of American Express, Wells Fargo, and the FBI to uncover this multi-year criminal operation. Even after his arrest, Kirilyuk skipped bail and had to be apprehended in Mexico City. At trial, he was convicted on all charges.

The majority finds Kirilyuk's 22-level loss enhancement "egregious" because his fraudulent scheme "involved $1.4 million in actual losses or $3.4 million in intended losses." Maj. Op. 1138. But the Sentencing Commission and district court could recognize that using those lower numbers would create an "egregious" problem of its own: those much lesser amounts do not nearly reflect the full magnitude of Kirilyuk's criminality. Indeed, this is precisely why Yellowe held that the multiplier was appropriate: because other metrics "would understate the severity of the offense." 24 F.3d at 1113.

It is also important to unpack the majority's numbers here. The $1.4 million figure, which the majority calls "actual losses," is the amount of false merchant account charges that American Express had already approved. The $3.4 million figure, which the majority calls the "intended loss," includes the additional charges that Kirilyuk submitted but which American Express declined to process.

The majority reasons that "the conspiracy was designed to charge only $15 to $30 per credit card." Maj. Op. 1138. But all we know is that Kirilyuk's conspiracy was stopped at this point, before further fraudulent charges were incurred. See Delima , 886 F.3d at 75 ("Actual losses were lower than intended losses because federal agents seized the conspirators' equipment and inventory, preventing the conspirators from profiting from the remaining numbers."). The Sentencing Commission and the district court were not required to accept Kirilyuk's self-serving attempt to put a ceiling on the ambitions of his massive international fraud operation. Nor were they required to blind themselves to the many other costs that this kind of scheme necessarily imposes beyond the bare amounts charged to the stolen card numbers.

Of course, if the district court judge who conducted Kirilyuk's trial and sentencing thought that a 22-level loss enhancement was too high, the judge had the discretion to grant a downward departure. See U.S.S.G. § 2B1.1, cmt. n.21(C). Kirilyuk in fact asked the district court for such a downward departure, but the court declined. The Guidelines themselves are also advisory in nature only. But rather than allowing district courts the option of a downward departure or a lower sentence, the majority is now effectively forcing this upon them. In my respectful view, that choice is not the majority's to make.

III

The full implications of the majority's opinion will likely be far-reaching and destabilizing. The per-card multiplier has been applied in our circuit, and in district courts across our circuit, for decades. Now it can no longer be applied to any defendant. In its place, district courts are left with uncertainty as to the proper measure of loss for unauthorized access device crimes. And it remains unclear whether other measures of Guidelines "loss" may also now be invalid as well. Indeed, many aspects of the Guidelines commentary may be open to challenge under the majority's Kisor -charged approach to Stinson —even commentary provisions backed by thirty years of circuit precedent. On these and many other questions, we should likely expect a wave of post-conviction motions and requests for re-sentencing.

We should not have gone down this road. Yellowe foreclosed it. Stinson foreclosed it. The law of nearly every other circuit counseled against it. I respectfully dissent from this unprecedented departure from governing law.