Dubin v. United States, 599 U.S. ___ (2023)

NOTICE: This opinion is subject to formal revision before publication in the United States Reports. Readers are requested to notify the Reporter of Decisions, Supreme Court of the United States, Washington, D. C. 20543, pio@supremecourt.gov, of any typographical or other formal errors.

SUPREME COURT OF THE UNITED STATES

_________________

No. 22–10

_________________

David Fox Dubin, PETITIONER v. UNITED STATES

on writ of certiorari to the united states court of appeals for the fifth circuit

[June 8, 2023]

Justice Sotomayor delivered the opinion of the Court.

There is no dispute that petitioner David Fox Dubin overbilled Medicaid for psychological testing. The question is whether, in defrauding Medicaid, he also committed “[a]ggravated identity theft,” 18 U. S. C. §1028A(a)(1), triggering a mandatory 2-year prison sentence. The Fifth Circuit found that he did, based on a reading of the statute that covers defendants who fraudulently inflate the price of a service or good they actually provided. On that sweeping reading, as long as a billing or payment method employs another person’s name or other identifying information, that is enough. A lawyer who rounds up her hours from 2.9 to 3 and bills her client electronically has committed aggravated identity theft. The same is true of a waiter who serves flank steak but charges for filet mignon using an electronic payment method.

The text and context of the statute do not support such a boundless interpretation. Instead, §1028A(a)(1) is violated when the defendant’s misuse of another person’s means of identification is at the crux of what makes the underlying offense criminal, rather than merely an ancillary feature of a billing method. Here, the crux of petitioner’s overbilling was inflating the value of services actually provided, while the patient’s means of identification was an ancillary part of the Medicaid billing process.

I

David Dubin helped his father manage a psychological services company. This company submitted a claim for reimbursement to Medicaid for psychological testing by a licensed psychologist. In fact, however, the claim overstated the qualifications of the employee who actually performed the testing and who was only a licensed psychological associate. This falsehood inflated the amount of reimbursement. Petitioner also changed the date on which the examination occurred.[1] Even with the inflation, the total reimbursement was only $338. App. 49. Petitioner was accordingly charged with healthcare fraud, a federal offense under 18 U. S. C. §1347. According to the Government, however, petitioner’s conduct also constituted “[a]ggravated identity theft” under §1028A(a)(1).

Section 1028A(a)(1) applies when a defendant, “during and in relation to any [predicate offense], knowingly transfers, possesses, or uses, without lawful authority, a means of identification of another person.” The predicate offenses include, among many others, healthcare fraud. §1028A(c)(4). Section 1028A(a)(1) carries a severe penalty: a mandatory minimum sentence of two years in prison “in addition to the punishment” for the predicate offense.

According to the Government, this is a clear aggravated identity theft case. The Government argued at trial that §1028A(a)(1) was automatically satisfied because petitioner’s fraudulent billing included the patient’s Medicaid reimbursement number (a “means of identification”). The District Court was less sure. “[T]his doesn’t seem to be an aggravated identity theft case” the court explained, as “the whole crux of this case is how [petitioner was] billing.” App. 37–38. This overbilling was “criminal,” but it “wasn’t aggravated identity theft.” Id., at 38. Nevertheless, the District Court denied petitioner’s post-trial challenge to his aggravated identity theft conviction, explaining that contrary Fifth Circuit precedent tied its hands. The court said that it “hope[d]” it would “get reversed.” Id., at 39.

On appeal, a Fifth Circuit panel affirmed. On rehearing en banc, a fractured court affirmed again. Five judges who agreed with the Government nonetheless acknowledged that under the Government’s reading of §1028A(a)(1), “the elements of [the] offense are not captured or even fairly described by the words ‘identity theft.’ ” 27 F. 4th 1021, 1024 (2022) (Richman, C. J., concurring). Eight dissenting judges agreed on this point.

This type of prosecution is not uncommon. The Government has, by its own admission, wielded §1028A(a)(1) well beyond ordinary understandings of identity theft. One prosecution targeted a defendant who “made a counterfeit handgun permit” for another person, using that person’s real name and at that person’s request. United States v. Spears, 729 F.3d 753, 754 (CA7 2013) (en banc). Another involved unlicensed doctors who violated the law by “issu[ing] prescriptions that their [actual] patients would then fill at . . . pharmacies.” United States v. Berroa, 856 F.3d 141, 148, 155–156 (CA1 2017). There was also a prosecution involving an ambulance service inflating its reimbursement rates by “mischaracteriz[ing] the nature of the transports, saying that the patients had required stretchers when they had not.” United States v. Michael, 882 F.3d 624, 628 (CA6 2018) (citing United States v. Medlock, 792 F.3d 700, 705 (CA6 2015)). Yet another prosecution involved a defendant who “provided massage services to patients to treat their pain,” but improperly billed this “as a Medicare-eligible physical therapy service.” United States v. Hong, 938 F.3d 1040, 1051 (CA9 2019).

Many lower courts have responded to such prosecutions with more restrained readings of the aggravated identity theft statute.[2] The Fifth Circuit did not. To resolve the conflict in the courts below, this Court granted certiorari, 598 U. S. ___ (2022), and now vacates the judgment of the Fifth Circuit and remands.[3]

II

A

This case turns on two of §1028A(a)(1)’s elements. Of the various possible ways to violate §1028A(a)(1), petitioner was convicted for “us[ing]” a patient’s means of identification “in relation to” healthcare fraud. The parties offer competing readings of these two elements.

The Government reads the terms broadly and in isolation. On the Government’s view, “[a] defendant uses a means of identification ‘in relation to’ a predicate offense if the use of that means of identification ‘facilitates or furthers’ the predicate offense in some way.” Brief for United States 10 (quoting Smith v. United States, 508 U.S. 223, 232 (1993)). As to “uses,” the Government seems just to mean “employ[s]” in any sense. Brief for United States 5, 7, 10–11. Section 1028A(a)(1) would thus apply automatically any time a name or other means of identification happens to be part of the payment or billing method used in the commission of a long list of predicate offenses. In other words, virtually all of the time.

Petitioner, in response, offers a more targeted reading. For petitioner, using a means of identification in relation to a predicate offense requires “a genuine nexus to the predicate offense.” Brief for Petitioner 15. On this reading, the means of identification is at the crux of what makes the predicate offense criminal, rather than merely an ancillary feature of a payment method. When the underlying crime involves fraud or deceit, as many of §1028A’s predicates do, this entails using a means of identification specifically in a fraudulent or deceitful manner.

To illustrate, petitioner borrows a heuristic from the Sixth Circuit. See Michael, 882 F. 3d, at 628. The relevant language in §1028A(a)(1) “covers misrepresenting who received a certain service,” but not “fraudulent claims regarding how or when a service was performed.” Brief for Petitioner 15. In other words, fraud going to identity, not misrepresentation about services actually provided. Take an ambulance service that actually transported patients but inflated the number of miles driven. The crux of this fraud was “how” services were rendered; the patients’ names were part of the billing process, but ancillary to what made the conduct fraudulent. See Michael, 882 F. 3d, at 628–629. In contrast, take the pharmacist who swipes information from the pharmacy’s files and uses it to open a bank account in a patient’s name. That “misuse of th[e] means of identification” would be “integral to” what made the conduct fraudulent, because misrepresentation about who was involved was at the crux of the fraud. Id., at 629.

In deciding between the parties’ readings, one limited and one near limitless, precedent and prudence require a careful examination of §1028A(a)(1)’s text and structure. While “uses” and “in relation to” are, in isolation, indeterminate, the statutory context, taken as a whole, points to a narrower reading.

B

In interpreting the scope of “uses” and “in relation to,” the Court begins with those terms themselves. Both terms have been singled out by this Court as particularly sensitive to context, and they do not, standing alone, conclusively resolve this case.

Start with “uses.” As the Court has observed more than once, “the word ‘use’ poses some interpretational difficulties because of the different meanings attributable to it.” Bailey v. United States, 516 U.S. 137, 143 (1995); see also Leocal v. Ashcroft, 543 U.S. 1, 9 (2004). The “ ‘ordinary or natural meaning’ ” of “ ‘use’ ” is “variously defined as ‘[t]o convert to one’s service,’ ‘to employ,’ ‘to avail oneself of,’ and ‘to carry out a purpose or action by means of.’ ” Bailey, 516 U. S., at 145. “These various definitions of ‘use’ imply action and implementation.” Ibid. Beyond that general concept, however, “ ‘use’ takes on different meanings depending on context,” and because it “draws meaning from its context, . . . we will look not only to the word itself, but also to the statute and the [surrounding] scheme, to determine the meaning Congress intended.” Id., at 143; see also Leocal, 543 U. S., at 9 (“Particularly when interpreting a statute that features as elastic a word as ‘use,’ we construe language in its context and in light of the terms surrounding it”).

For example, the federal arson statute only applies to buildings “ ‘used in’ commerce or commerce-affecting activity.” Jones v. United States, 529 U.S. 848, 850–851 (2000). In that statutory context, the Court distinguished between uses of a building as “the locus of any commercial undertaking,” and noncovered “passive,” “passing,” or ancillary uses of a building “as collateral to obtain and secure a mortgage” or to obtain an insurance policy. Id., at 855–856. It is statutory context, therefore, that determines what kind of active employment or conversion to one’s service triggers §1028A(a)(1)’s harsh penalty.

“In relation to” is similarly context sensitive. If “ ‘relate to’ were taken to extend to the furthest stretch of its indeterminacy, then for all practical purposes” there would be no limits, as “ ‘[r]eally, universally, relations stop nowhere.’ ” New York State Conference of Blue Cross & Blue Shield Plans v. Travelers Ins. Co., 514 U.S. 645, 655 (1995) (quoting H. James, Roderick Hudson xli (New York ed., World’s Classics 1980)). This language thus cannot be “considered in isolation,” Maracich v. Spears, 570 U.S. 48, 59 (2013), and the Court must “go beyond the unhelpful text and the frustrating difficulty of defining [this] key term” and look to statutory context. Travelers, 514 U. S., at 656. That the phrase refers to a relationship or nexus of some kind is clear. See Smith, 508 U. S., at 238 (“ ‘[I]n relation to’ ” requires “some purpose or effect” between two things). Yet the kind of relationship required, its nature and strength, will be informed by context.

The presence of two such context-dependent terms renders §1028A(a)(1) doubly attuned to its surroundings. The parties’ competing readings both fall within the range of meanings of “uses” and “in relation to,” taken alone. Resort to context is thus especially necessary here.[4]

C

Having found the key terms “use” and “in relation to” indeterminate, the next step is to look to their surrounding words. After all, “a statute’s meaning does not always turn solely on the broadest imaginable definitions of its component words.” Epic Systems Corp. v. Lewis, 584 U. S. ___, ___ (2018) (slip op., at 23) (internal quotation marks omitted). Instead, “[l]inguistic and statutory context also matter.” Ibid. Even in cases where “the literal language of the statute is neutral” in isolation, reading “the whole phrase” can point to a more targeted reading. Marinello v. United States, 584 U. S. ___, ___–___ (2018) (slip op., at 4–5).

Such is the case here. Section 1028A(a)(1)’s title and terms both point to a narrower reading, one centered around the ordinary understanding of identity theft. This cuts against the Government’s broad reading, which the Government admits bears little relationship to the common understanding of identity theft. In contrast, a more targeted reading accurately captures the ordinary understanding of identity theft, where misuse of a means of identification is at the crux of the criminality.

1

Start at the top, with the words Congress chose for §1028A’s title: “Aggravated identity theft.” 118Stat. 831. This Court has long considered that “ ‘the title of a statute and the heading of a section’ are ‘tools available for the resolution of a doubt’ about the meaning of a statute.” Almendarez-Torres v. United States, 523 U.S. 224, 234 (1998) (quoting Trainmen v. Baltimore & Ohio R. Co., 331 U.S. 519, 528–529 (1947)). A title will not, of course, “override the plain words” of a statute. Fulton v. Philadelphia, 593 U. S. ___, ___ (2021) (slip op., at 9). Yet here, the key terms are so “elastic” that they must be construed “in light of the terms surrounding [them],” Leocal, 543 U. S., at 9, and the title Congress chose is among those terms. Even the Government acknowledged that if the terms in §1028A(a)(1) are unclear, “the title is a useful clue.” Tr. of Oral Arg. 80.

Two additional points bolster this approach. First, the title here is not serving the unenviable role of pithily summarizing a list of “complicated and prolific” provisions. Trainmen, 331 U. S., at 528. Section 1028A is a focused, standalone provision. Second, a title is “especially valuable [where] it reinforces what the text’s nouns and verbs independently suggest.” Yates v. United States, 574 U.S. 528, 552 (2015) (Alito, J., concurring in judgment). As explained below, §1028A(a)(1)’s text independently suggests a focus on identity theft. See infra, at 12–15.

Indeed, this Court has already once used §1028A’s title and place in the statutory scheme to shed light on its text. In Flores-Figueroa v. United States, 556 U.S. 646 (2009), this Court pointed out that a neighboring provision, §1028, carries the broad title “ ‘Fraud and related activity in connection with identification documents, authentication features, and information.’ ” Id., at 655. Section 1028A, in contrast, is far more targeted, “us[ing] the words ‘identity theft.’ ” Ibid. That “Congress separated the [identity] fraud crime from the [identity] theft crime in” §1028A suggests that §1028A is focused on identity theft specifically, rather than all fraud involving means of identification. Ibid.[5]

Given that, it is abundantly clear why the Government urges the Court to ignore the title. The Government’s broad reading, covering any time another person’s means of identification is employed in a way that facilitates a crime, bears little resemblance to any ordinary meaning of “identity theft.” Consider again an unlicensed doctor who fills out a prescription actually requested by a patient; no one would call that identity theft. Even judges below who agreed with the Government’s reading of §1028A(a)(1), and ultimately the Government itself, acknowledged that its reading of §1028A(a)(1) does not fairly capture the ordinary meaning of identity theft. Nor is the difference just around the edges; the Government’s reading would, in practice, place garden-variety overbilling at the core of §1028A.

Instead, “identity theft” has a focused meaning. One dictionary defines identity theft as “the fraudulent appropriation and use of another person’s identifying data or documents, as a credit card.” Webster’s Unabridged Dictionary xi (2d ed. 2001) (Webster’s). Another similarly offers “[t]he unlawful taking and use of another person’s identifying information for fraudulent purposes; specif[ically] a crime in which someone steals personal information about and belonging to another, such as a bank-account number or driver’s-license number, and uses the information to deceive others.” Black’s Law Dictionary 894 (11th ed. 2019) (Black’s) (defining “identity theft”).[6]

This supports a reading of “in relation to” where use of the means of identification is at the crux of the underlying criminality. These definitions refer to offenses built around what the defendant does with the means of identification in particular. In other words, the means of identification specifically is a key mover in the criminality. This central role played by the means of identification, which serves to designate a specific person’s identity, explains why we say that the “identity” itself has been stolen. See, e.g., Spears, 729 F. 3d, at 756 (“identity theft” occurs when someone’s “identity has been stolen or misappropriated”). This helps explain why the examples resulting from the Government’s theory do not sound like identity theft. If a lawyer rounds up her hours from 2.9 to 3 and bills her client using his name, the name itself is not specifically a source of fraud; it only plays an ancillary role in the billing process. The same is true for the waiter who substitutes one cut of meat for another; we might say the filet mignon’s identity was stolen, perhaps, but not the diner’s.

This understanding of identity theft also supports a more targeted definition of “uses.” The word “use” appears in these definitions with a specific meaning: Identity theft encompasses when a defendant “uses the information to deceive others,” Black’s 894 (emphasis added), and “the fraudulent . . . use” of a means of identification, Webster’s xi (emphasis added). In other words, identity theft is committed when a defendant uses the means of identification itself to defraud or deceive. This tracks the Sixth Circuit’s heuristic. When a means of identification is used deceptively, this deception goes to “who” is involved, rather than just “how” or “when” services were provided. Use of the means of identification would therefore be at “the locus of [the criminal] undertaking,” rather than merely “passive,” “passing,” or ancillary employment in a crime. Jones, 529 U. S., at 855–856.

On top of that, §1028A’s title is not just “identity theft,” but “Aggravated identity theft.” Typically, “[a]n ‘aggravated’ offense is one ‘made worse or more serious by circumstances such as violence, the presence of a deadly weapon, or the intent to commit another crime.’ ” Carachuri-Rosendo v. Holder, 560 U.S. 563, 574 (2010) (quoting Black’s Law Dictionary 75 (9th ed. 2009)). This suggests that Congress had in mind a particularly serious form of identity theft. Yet the Government’s reading “would apply an ‘aggravated’ . . . label” to all manner of everyday overbilling offenses. Carachuri-Rosendo, 560 U. S., at 574. “Of course . . . Congress, like ‘Humpty Dumpty,’ has the power to give words unorthodox meanings.” Id., at 575. Yet where “the Government argues for a result that the English language tells us not to expect, . . . we must be very wary of the Government’s position.” Ibid. (internal quotation marks omitted).

The title suggests identity theft is at the core of §1028A(a)(1). On the Government’s reading, however, everyday overbilling would become the most common trigger for §1028A(a)(1)’s severe penalty. This would turn the core of “worse or more serious” identity theft into something the ordinary user of the English language would not consider identity theft at all.

2

The title is, by definition, just the beginning. A title does not supplant the actual text of the provision, as the Government observes. The problem for the Government is that §1028A(a)(1)’s language points in the same direction as its title. In particular, Congress used a trio of verbs that reflect an ordinary understanding of identity theft.

While “uses” is indeterminate in isolation, here it has company. Section 1028A(a)(1) applies when a defendant “knowingly transfers, possesses, or uses, without lawful authority, a means of identification of another person,” “during and in relation to” any predicate offense. (Emphasis added.) “Under the familiar interpretive canon noscitur a sociis, ‘a word is known by the company it keeps.’ ” McDonnell v. United States, 579 U.S. 550, 568–569 (2016) (quoting Jarecki v. G. D. Searle & Co., 367 U.S. 303, 307 (1961)). “[T]his canon is often wisely applied where a word is capable of many meanings in order to avoid the giving of unintended breadth to the Acts of Congress.” McDonnell, 579 U. S., at 569 (internal quotation marks omitted).

The two neighboring verbs here, “transfers” and “possesses,” are most naturally read in the context of §1028A(a)(1) to connote theft. While it is not necessary to determine the precise metes and bounds of these two verbs, their role in the provision points to this targeted reading. Section 1028A(a)(1) covers unlawful possession or transfer of a means of identification belonging to “another person.” Generally, to unlawfully “possess” something belonging to another person suggests it has been stolen. And to unlawfully “transfer” something belonging to another person similarly connotes misappropriating it and passing it along. In Flores-Figueroa, this Court drew a similarly intuitive link between a defendant taking a means of identification he knows belongs to another person and “ ‘theft.’ ” 556 U. S., at 655. The Government, at argument, agreed: these two verbs “refer to circumstances in which the information is stolen.” Tr. of Oral Arg. 90.[7]

“Transfer” and “possess” not only connote theft, but identity theft in particular. The verbs point to (1) theft of a (2) means of identification belonging to (3) another person. That tracks ordinary understandings of identity theft: “a crime in which someone [1] steals [2] personal information about and [3] belonging to another.” Black’s 894. Similarly, “the [1] fraudulent appropriation and use of [3] another person’s [2] identifying data or documents.” Webster’s xi. If this parallel were not enough, §1028A(a)(1)’s title indicates that the type of theft its verbs connote is identity theft specifically.

Because “transfer” and “possess” channel ordinary identity theft, noscitur a sociis indicates that “uses” should be read in a similar manner to its companions. See McDonnell, 579 U. S., at 568–569. “Uses” is quite amenable to such a reading, and not just because of its indeterminacy. As explained above, “using” another person’s means of identification to deceive or defraud is a common feature of identity theft. See Webster’s xi (“the fraudulent . . . use” of a means of identification (emphasis added)); Black’s 894 (when a defendant “uses the information to deceive others” (emphasis added)).

Congress thus employed a trio of verbs that capture various aspects of “classic identity theft.” Flores-Figueroa, 556 U. S., at 656. There is “the defendant [who] has gone through someone else’s trash to find discarded credit card and bank statements,” ibid., and thus has taken possession unlawfully. There is the bank employee who passes along customer information to an accomplice, and thus transfers it unlawfully. Then there is use involving fraud or deceit about identity: “a defendant [who] has used another person’s identification information to get access to that person’s bank account.” Ibid.

Another canon of construction offers a further point in favor of this narrow interpretation. The Court “assume[s] that Congress used [three] terms because it intended each term to a have a particular, nonsuperfluous meaning.” Bailey, 516 U. S., at 146. Reading §1028A(a)(1)’s operative verbs as tracking aspects of classic identity theft, each verb has an independent role to play. As the definitions reveal, identity theft covers both when “someone steals personal information about and belonging to another . . . and uses the information to deceive others,” Black’s 894 (emphasis added), and “fraudulent appropriation and use,” Webster’s xi (emphasis added). Identity theft thus intermingles aspects of theft and fraud, misappropriation and deceitful use. Section 1028A(a)(1)’s three verbs capture this complexity. While “transfer” and “possess” conjure up two steps of theft, “uses” supplies the deceitful use aspect.

In contrast, if §1028A(a)(1)’s verbs do not track identity theft and if the means of identification need only facilitate the predicate offense, the other two verbs threaten to leave “uses” without “virtually any function.” Bailey, 516 U. S., at 146. Return to a definition of “in relation to” that just means “ ‘facilitates or furthers’ the predicate offense in some way.” Brief for United States 10. In virtually all cases where a defendant employs a means of identification to facilitate a crime, the defendant will also possess or transfer the means of identification in a way that facilitates the crime. For example, petitioner’s possession of the patient’s means of identification facilitated the fraud, as did petitioner’s transfer of the patient’s means of identification to Medicaid. It is hard to imagine when “uses” would not similarly be covered by, at least, one of the two other verbs. This risk of superfluity suggests giving §1028A(a)(1) a more precise reading.

In sum, §1028A(a)(1)’s title and text are mutually reinforcing. Both point toward requiring the means of identification to be at the crux of the criminality.

D

Section 1028A’s list of predicate offenses points to yet another stumbling block for the Government’s broad reading. Section 1028A(a)(1) is an enhancement, and a severe one at that. It adds a 2-year mandatory prison sentence onto underlying offenses that do not impose a mandatory prison sentence of any kind. See, e.g., 18 U. S. C. §1035 (“[f]alse statements relating to health care matters,” setting no minimum sentence). This prevents sentencing judges from considering the severity of the offense, even if the amount of money involved was quite small or there are other mitigating factors. Interpretation of §1028A(a)(1) should thus reflect the “distinction between” the aggravated identity theft crimes that “Congress sought to distinguish for heightened punishment and other crimes.” Leocal, 543 U. S., at 11.

Far from distinguishing, the Government’s reading collapses the enhancement into the enhanced. Here, the Government claims that because petitioner’s overbilling was facilitated by the patient’s Medicaid reimbursement number, §1028A(a)(1) automatically applies. Patient names or other identifiers will, of course, be involved in the great majority of healthcare billing, whether Medicare for massages, Hong, 938 F. 3d, at 1051, or for ambulance stretcher services, Medlock, 792 F. 3d, at 706. Patient names will be on prescriptions, Berroa, 856 F. 3d, at 148, 155–156, and patients committing fraud on their own behalf will often have to include the names of others on their forms, such as doctors or employers. Under the Government’s own reading, such cases are “automatically identity theft,” Tr. of Oral Arg. 82, independent of whether the name itself had anything to do with the fraudulent aspect of the offense.

Nor are these implications confined to healthcare. Section 1028A(a)(1)’s predicates include a vast array of offenses, including wire fraud and mail fraud. §1028A(c)(5). The Government’s boundless reading of “uses” and “in relation to” would cover facilitating mail fraud by using another person’s name to address a letter to them.[8] Even beyond that, names or other means of identification are used routinely for billing and payment, whether payment apps, credit and debit cards, a bill sent by mail, or an invoice sent electronically. So long as the criteria for the broad predicate offenses are met, the Government’s reading creates an automatic 2-year sentence for generic overbilling that happens to use ubiquitous payment methods.

A far more sensible conclusion from the statutory structure is that §1028A(a)(1)’s enhancement is not indiscriminate, but targets situations where the means of identification itself plays a key role—one that warrants a 2-year mandatory minimum. This points once more to a targeted reading, where the means of identification is at the crux of the underlying criminality, not an ancillary feature of billing.

E

If more were needed, a final clue comes from the staggering breadth of the Government’s reading. This Court has “ ‘traditionally exercised restraint in assessing the reach of a federal criminal statute.’ ” Marinello, 584 U. S., at ___ (slip op., at 9) (quoting United States v. Aguilar, 515 U.S. 593, 600 (1995)); see also Arthur Andersen LLP v. United States, 544 U.S. 696, 703–704 (2005); McBoyle v. United States, 283 U.S. 25, 27 (1931). This restraint arises “both out of deference to the prerogatives of Congress and out of concern that a fair warning should be given to the world in language that the common world will understan[d] of what the law intends to do if a certain line is passed.” Marinello, 584 U. S., at ___ (slip op., at 4) (internal quotation marks omitted). After all, “[c]rimes are supposed to be defined by the legislature, not by clever prosecutors riffing on equivocal language.” Spears, 729 F. 3d, at 758.

Time and again, this Court has prudently avoided reading incongruous breadth into opaque language in criminal statutes. In Van Buren v. United States, 593 U. S. ___ (2021), the “far-reaching consequences” of the Government’s reading “underscore[d] the implausibility of the Government’s interpretation.” Id., at ___ (slip op., at 17). In Marinello, the Court rejected the Government’s reading of a statute about obstructing administration of the Tax Code that would have swept in the “person who pays a babysitter $41 per week in cash without withholding taxes,” as well as someone who “leaves a large cash tip in a restaurant, fails to keep donation receipts from every charity to which he or she contributes, or fails to provide every record to an accountant.” 584 U. S., at ___ (slip op., at 7). Nor was all such conduct innocent, as the statute required an individual to act “ ‘corruptly.’ ” Id., at ___ (slip op., at 8). Even still, “[h]ad Congress intended” to sweep so far, “it would have spoken with more clarity than it did.” Id., at ___ (slip op., at 7). In Yates, the Court held that the Government’s “unrestrained” reading would have turned a provision focused on “records” and “documents” into “an all-encompassing ban on the spoliation of evidence” that would “sweep within its reach physical objects of every kind,” including a fish. 574 U. S., at 536, 540 (plurality opinion). Had Congress set out to do so, “one would have expected a clearer indication of that intent.” Ibid.

So too here. The Government’s reading would sweep in the hour-inflating lawyer, the steak-switching waiter, the building contractor who tacks an extra $10 onto the price of the paint he purchased. So long as they used various common billing methods, they would all be subject to a mandatory two years in federal prison. To say that such a result is implausible would be an understatement.[9] Because everyday overbilling cases would account for the majority of violations in practice, the Government’s reading places at the core of the statute its most improbable applications.

Finally, the Government makes a familiar plea: There is no reason to mistrust its sweeping reading, because prosecutors will act responsibly. To this, the Court gives a just-as-familiar response: We “cannot construe a criminal statute on the assumption that the Government will ‘use it responsibly.’ ” McDonnell, 579 U. S., at 576 (quoting United States v. Stevens, 559 U.S. 460, 480 (2010)). “[T]o rely upon prosecutorial discretion to narrow the otherwise wide-ranging scope of a criminal statute’s highly abstract general statutory language places great power in the hands of the prosecutor.” Marinello, 584 U. S., at ___ (slip op., at 9). This concern is particularly salient here. If §1028A(a)(1) applies virtually automatically to a swath of predicate offenses, the prosecutor can hold the threat of charging an additional 2-year mandatory prison sentence over the head of any defendant who is considering going to trial.

III

All the points above are different wells drawing from the same source. The Court need not decide whether any of these points, standing alone, would be dispositive. Taken together, from text to context, from content to common sense, §1028A(a)(1) is not amenable to the Government’s attempt to push the statutory envelope. A defendant “uses” another person’s means of identification “in relation to” a predicate offense when this use is at the crux of what makes the conduct criminal. To be clear, being at the crux of the criminality requires more than a causal relationship, such as “ ‘facilitation’ ” of the offense or being a but-for cause of its “success.” Post, at 3, 5–6 (Gorsuch, J., concurring in judgment). Instead, with fraud or deceit crimes like the one in this case, the means of identification specifically must be used in a manner that is fraudulent or deceptive. Such fraud or deceit going to identity can often be succinctly summarized as going to “who” is involved.[10]

Here, petitioner’s use of the patient’s name was not at the crux of what made the underlying overbilling fraudulent. The crux of the healthcare fraud was a misrepresentation about the qualifications of petitioner’s employee. The patient’s name was an ancillary feature of the billing method employed. The Sixth Circuit’s more colloquial formulation is a helpful guide, though like any rule of thumb it will have its limits. Here, however, it neatly captures the thrust of the analysis, as petitioner’s fraud was in misrepresenting how and when services were provided to a patient, not who received the services.

*  *  *

Because petitioner did not use the patient’s means of identification in relation to a predicate offense within the meaning of §1028A(a)(1), the judgment of the Court of Appeals is vacated, and the case is remanded for further proceedings consistent with this opinion.

It is so ordered.

SUPREME COURT OF THE UNITED STATES

_________________

No. 22–10

_________________

David Fox Dubin, PETITIONER v. UNITED STATES

on writ of certiorari to the united states court of appeals for the fifth circuit

[June 8, 2023]

Justice Gorsuch, concurring in the judgment.

Whoever among you is not an “aggravated identity thief,” let him cast the first stone. The United States came to this Court with a view of 18 U. S. C. §1028A(a)(1) that would affix that unfortunate label on almost every adult American. Every bill splitter who has overcharged a friend using a mobile-payment service like Venmo. Every contractor who has rounded up his billed time by even a few minutes. Every college hopeful who has overstated his involvement in the high school glee club. All of those individuals, the United States says, engage in conduct that can invite a mandatory 2-year stint in federal prison. The Court today rightly rejects that unserious position. But in so holding, I worry the Court has stumbled upon a more fundamental problem with §1028A(a)(1). That provision is not much better than a Rorschach test. Depending on how you squint your eyes, you can stretch (or shrink) its meaning to convict (or exonerate) just about anyone. Doubtless, creative prosecutors and receptive judges can do the same. Truly, the statute fails to provide even rudimentary notice of what it does and does not criminalize. We have a term for laws like that. We call them vague. And “[i]n our constitutional order, a vague law is no law at all.” United States v. Davis, 588 U. S. ___, ___ (2019) (slip op., at 1).

The “[a]ggravated identity theft” statute stipulates that “[w]hoever, during and in relation to any felony violation” listed in a later subsection, “knowingly transfers, possesses, or uses, without lawful authority, a means of identification of another person shall, in addition to the punishment provided for such felony, be sentenced to a term of imprisonment of 2 years.” §1028A(a)(1). Today, the Court sets out to determine what conduct that law reaches. It is, as the Court acknowledges, no easy task. Both the term “us[e]” and the phrase “in relation to” can support a multitude of possible meanings. Ante, at 5–7. They of course “ ‘imply action and implementation.’ ” Ante, at 6 (quoting Bailey v. United States, 516 U.S. 137, 145 (1995)). Beyond that “general concept,” however, we must fend for ourselves based only on limited contextual clues. Ante, at 5–7.

The United States offers up a rapacious interpretation that would require only “the use of th[e] means of identification [to] ‘facilitat[e] or furthe[r]’ the predicate offense in some way.” Brief for United States 10 (emphasis added). Admittedly, this reading “fall[s] within the range” of plausible meanings the statute could support. Ante, at 7. But so too do other readings—ones that require a more demanding “nexus” between the “means of identification” and the underlying misconduct. Ante, at 7–8. For many of the reasons the Court gives (and more besides), I agree that we must adhere to those more restrained offerings. The United States’ maximalist approach has simplicity on its side, yes; an everybody-is-guilty standard is no challenge to administer. But the Constitution prohibits the Judiciary from resolving reasonable doubts about a criminal statute’s meaning by rounding up to the most punitive interpretation its text and context can tolerate. See Wooden v. United States, 595 U. S. ___, ___–___ (2022) (Gorsuch, J., concurring in judgment) (slip op., at 6–9). That insight alone means Mr. Dubin’s §1028A(a)(1) conviction cannot stand.

Unfortunately, our opinion cannot end there. Having told lower courts how not to read the statute, we owe them some guidance as to how they should read it. That is where the real challenge begins. Drawing on contextual clues and rules of statutory interpretation, the Court concludes that a violation of §1028A(a)(1) occurs whenever the “use of the means of identification is at the crux of the underlying criminality.” Ante, at 10 (emphasis added). “In other words, the means of identification specifically” must be in some way “a key mover in the criminality.” Ibid. (emphasis added). Put still another way, the “means of identification” must play the (or maybe a) “central role” in the commission of the offense. Ibid. (emphasis added).

Setting aside some definite-article inconsistency, those formulations all sound sensible enough. On closer review, however, they present intractable interpretive challenges of their own. When, exactly, is a “means of identification” “at the crux,” “a key mover,” or a “central role” player in an offense? No doubt, the answer “turns on causation, or at least causation often helps to answer the question.” United States v. Michael, 882 F.3d 624, 628 (CA6 2018). The Court agrees but stresses that “a causal relationship” of any kind will not suffice. Ante, at 20. At the same time, however, it studiously avoids indicating whether the appropriate standard is proximate cause or something else entirely novel. Ibid. All of which gives rise to further questions. In virtually every fraud, a “means of identification” plays some critical role in the fraud’s success—good luck committing a mail or wire fraud, for instance, without relying heavily on the name of the victim and likely the names of other third parties. Just how much “causation” must a prosecutor establish to sustain a §1028A(a)(1) conviction? For that matter, how does one even determine the extent to which a “means of identification” “caused” an offense, as compared to the many other necessary inputs?

The Court supplies no firm answer. Instead, it leans on various illustrations that only highlight the difficulties inherent in this exercise. Take, for instance, the Court’s assurance that a “waiter who serves flank steak but charges for filet mignon using an electronic payment method” has not committed aggravated identity theft. Ante, at 1, 11. Why not, exactly? In one sense, the “means of identification” (the credit card) lies “at the crux” of the fraud. The restaurant uses it to charge the customer for a product it never supplied. Maybe that feels less distasteful than a scenario in which an overseas hacker steals an individual’s credit card information and deploys it to order luxury goods on Amazon. But the Constitution’s promise of due process means that criminal statutes must provide rules “knowable in advance,” not intuitions discoverable only after a prosecutor has issued an indictment and a judge offers an opinion. Percoco v. United States, 598 U. S. ___, ___ (2023) (Gorsuch, J., concurring in judgment) (slip op., at 6).

Not yet convinced? Consider some tweaks to the Court’s hypothetical. Suppose that, instead of misrepresenting the cut of its steaks, a restaurant charged a customer for an appetizer he ordered that never arrived. What about an appetizer he never ordered? An additional entrée? Three? Three plus a $5,000 bottle of Moët? How about a Boeing 737? Now suppose the restaurant ran the customer’s credit card for the same steak twice. What if it waited an hour to do so? A day? A year? What if the waiter gave the credit card information to a different employee at the same restaurant to run the charge? A different employee at a different restaurant? What if the restaurant sold the customer’s credit card information on the dark web, and another restaurant ran the card for filet mignon? On the Court’s telling, the “crux” of the fraud in some of these examples lies merely in “how and when services were provided,” while in others the “crux” involves “who received the services.” Ante, at 20. But how to tell which is which?

The Court’s “crux” test seemingly offers no sure way through this “blizzard of . . . hypotheticals.” Ibid., n. 10. Nor is that because I have cherry-picked “hard cases.” Ibid. Scenarios like these—and variations of them—illustrate the sorts of problems that invariably arise in even simple §1028A(a)(1) cases involving bogus restaurant bills. Other contexts can present still greater complications and still deeper uncertainties. The problem we face, then, is not that §1028A(a)(1) presents some hard cases at its edges; the problem is this statute has no easy cases. Really, you could spend a whole day cooking up scenarios—ranging from the mundane to the fanciful—that collapse even your most basic intuitions about what §1028A(a)(1) does and does not criminalize. Try making up some of your own and running them by a friend or family member. You may be surprised at how sharply instincts diverge.

For the less adventurous, consider just the facts of the case now before us. On one framing, it seems outrageous to convict Mr. Dubin of aggravated identity theft. After all, the patient did (at one point) receive psychological testing. So you might say, as the Court does, that Mr. Dubin lied only about the qualifications of the individual who provided those services and the date on which they occurred. See ante, at 2, 20. But on another framing, the patient’s identity was “a key mover,” perhaps even “at the crux,” of the fraud. Mr. Dubin could not have successfully billed the insurance provider without accurately offering up some specific patient’s name and information. Nor, as the United States notes, could Mr. Dubin have simply drawn a random name from a hat. Rather, his fraud depended on purloining the specific identity of a “Texas Medicaid enrollee who had at least three hours of psychological-testing reimbursement left in his or her account.” Brief for United States 13. Along the way, Mr. Dubin’ compensable psychological services. From the patient’s perspective, Mr. Dubin’s use of his “means of identification” could hardly feel “ancillary.” Ante, at 1–2.

As an abstract exercise, debating fact patterns like these may seem good fun. But there is nothing entertaining about a 2-year mandatory federal prison sentence. Criminal statutes are not games to be played in the car on a cross-country road trip. To satisfy the constitutional minimum of due process, they must at least provide “ordinary people” with “fair notice of the conduct [they] punis[h].” Johnson v. United States, 576 U.S. 591, 595 (2015). And, respectfully, I do not see how §1028A(a)(1) can clear that threshold. Under the Court’s “crux” test, no boundary separates conduct that gives rise to liability from conduct that does not. And it appears I share this concern with the very lower court judges who will have to apply this standard prospectively. As even many of the Fifth Circuit dissenters below warned, the sort of “facilitation standard” the Court today adopts, “with its incidental/integral dividing line,” is unworkable because it “lacks clear lines and a limiting principle.” 27 F. 4th 1021, 1042 (2022) (en banc) (Costa, J., dissenting). In the end, it is hard not to worry that the Court’s “crux” test will simply become a fig leaf for judges’ and jurors’ own subjective moral judgments about whether (as the Court itself puts it) the defendant’s crime is “one that warrants a 2-year mandatory minimum.” Ante, at 17.

I do not question that the Court today has done the best it might to make sense of this statute. It’s just that it faces an impossible task. In the past when this Court has grappled with similar statutory language, it has done so in contexts where the relevant terms could carry only a few possible (and comparatively fixed) meanings. For example, when it comes to the “us[e]” of a firearm “in relation” to a crime of violence, 18 U. S. C. §924(c)(1)(A), the presence of a gun could be a but-for cause of (or a necessary ingredient of ) the offense—used, for example, as compensation in an exchange for illicit drugs. Smith v. United States, 508 U.S. 223, 237–238 (1993). Or the gun could be “ ‘used as a weapon’ ” by being discharged or brandished. Id., at 243 (Scalia, J., dissenting). Because both those interpretations are relatively bounded and understandable, this Court could use principles of statutory interpretation to choose between them. The same holds true for many of the other statutes the Court (mistakenly) frets I would call into doubt. See ante, at 20, n. 10.

The same cannot be said for §1028A(a)(1), though. There are an uncountable number of ways in which an individual could “us[e]” the “means of identification” of another to commit fraud. That list covers everything from including a victim’s name in the subject line of a fraudulent email; to misrepresenting information on a loan form involving a co-signer; to putting on a wig and walking into a bank with a fake ID. And no obvious neutral rule exists to separate those “uses” that violate §1028A(a)(1) from others that do not. In this way, §1028A(a)(1) is not just an “ambiguous” statute—“one that does define prohibited conduct with some precision, but [that] is subject to two or more different interpretations.” J. Decker, Addressing Vagueness, Ambiguity, and Other Uncertainty in American Criminal Laws, 80 Denver U. L. Rev. 241, 261 (2002) (emphasis added). Instead, it is a vague statute—one that “does not satisfactorily define the proscribed conduct” at all. Id., at 260–261.

I do not write this opinion as wishcasting. Perhaps, by applying the Court’s “crux” test, lower courts will achieve a consistency that has, to date, eluded them. Or perhaps they will, prompted by today’s decision, locate a previously unseen path through this statutory quagmire. But I would not hold my breath. Section 1028A(a)(1) simply does too little to specify which individuals deserve the inglorious title of “aggravated identity thief.” That is a problem Congress alone can fix. Until it does, I fear the issues that have long plagued lower courts will persist. And I will not be surprised if someday, maybe someday soon, they find their way back here.