UNITED STATES v. DE SOUZA PRADO (2025)

A jury convicted Thiago de Souza Prado of wire fraud, conspiracy to commit wire fraud, and aggravated identity theft. The convictions arose from a scheme to defraud rideshare and food delivery companies and, to further that scheme, steal or misappropriate the identities of third parties. The district court sentenced Prado to seventy months in prison. Prado appeals, challenging his convictions and sentence. We affirm.

I.

The following facts are uncontested. Prado was born in Brazil. In 2003, Prado entered the United States on a tourist visa, but he is not presently authorized to be in the country. In January 2019, Prado began participating in the criminal scheme for which he was convicted. A network of Brazilian nationals living in the United States carried out the scheme by creating and using fraudulent accounts so as to permit ineligible individuals to drive for Uber, Uber Eats, Lyft, DoorDash, Instacart, and Grubhub. Prado was himself ineligible to drive for these companies because he had numerous disqualifying driving infractions; others were ineligible because, among other reasons, they were not legally authorized to work.

Prado initially drove under fraudulent accounts created by others and paid rental fees to use these accounts. Eventually, however, Prado began creating fraudulent accounts to use himself and to rent to others. To establish these accounts, Prado surreptitiously obtained driver's license images and Social Security numbers from individuals who could pass the necessary driver background checks. Prado acquired these images and numbers from several sources: an associate who worked at a night club and took pictures of patrons’ licenses; employee records from a painting company that he once owned; an associate who acquired Social Security numbers from the dark web; software Prado purchased that enabled him to access the dark web and acquire Social Security numbers himself; and associates who delivered alcohol for grocery delivery companies and took photographs of patrons’ licenses, purportedly to confirm their ages.

Prado also used a different software program, called a “drone” or “bot,” to defraud Uber and Lyft. The program spoofed GPS information on the companies’ apps to make it appear that the fake drivers on Prado's fraudulent accounts had completed rides that they never actually provided. The program also made completed trips look longer than they had been, thus inflating driver fees. Prado bought the drone program, shared it with other participants in the scheme, and sold it to other drivers for hundreds of dollars.

Finally, Prado defrauded Uber and Lyft by having the fake accounts he created refer other fake accounts as “new” drivers. These fake referrals generated for Prado hundreds of dollars in rewards paid by Uber and Lyft for referring new drivers. Prado also used an account in his wife's name to receive the referral rewards.

On May 17, 2021, a grand jury indicted Prado and seventeen others in connection with the scheme described above. Over two years later, the grand jury returned a third superseding indictment charging Prado with one count of conspiracy to commit wire fraud, 18 U.S.C. § 1349 (Count One); three counts of wire fraud, 18 U.S.C. § 1343 (Counts Two to Four); and three counts of aggravated identity theft, 18 U.S.C. § 1028A (Counts Five to Seven). Only Prado went to trial. After a seven-day trial, a jury convicted him on all counts. The district court later sentenced Prado to concurrent forty-six-month prison terms on Counts One through Four and concurrent twenty-four-month prison terms on Counts Five through Seven to run consecutively, thereby producing a total term of seventy months in prison.

II.

Prado challenges his convictions by asserting that he was prejudiced by an amendment of the third superseding indictment during trial and by the district court's refusal to disqualify the prosecution team from involvement in his case. Alternatively, Prado contends that his sentence was procedurally and substantively unreasonable.

A.

We first address Prado's prejudicial amendment argument. The third superseding indictment detailed how and when Prado defrauded some of his victims and misappropriated and misused others’ identities. In so doing, the document's narrative section employed generic pseudonyms (for example, “Rideshare Company A” or “Victim 1”), rather than the corporate or individual victims’ names. On the fourth day of trial, the district court suggested that the government submit a revised version of the third superseding indictment replacing the pseudonyms with the victims’ names. The government did so shortly thereafter. The court took this action to assist the jury deliberations by making the indictment more intelligible.

Days later, following the close of evidence, Prado raised a pro se objection that he had not received notice of the names of the victims referenced in the third superseding indictment until the district court allowed the document's revision mid-trial, despite having asked his attorney for these names. Defense counsel did not join in the objection and confirmed that, in the months prior to trial, the prosecution had twice provided him with the victims’ names. Defense counsel also confirmed that he knew the names of the victims identified in each count. The court took the matter under advisement.

Following the verdict, the district court allowed Prado to pursue his lack-of-notice objection by means of a pro se motion. Prado responded with a motion for acquittal under Federal Rule of Criminal Procedure 29 and an affidavit reasserting that, despite having sought the information from his attorney, he did not learn the victims’ names until the sixth day of trial. Prado argued that this late notice violated his right to a constitutionally adequate indictment. Prado further argued that this violation was structural error requiring the court to set aside the jury's verdicts and dismiss the indictment.

By a written memorandum and order, the district court denied the motion. After noting that Prado's challenge to the unrevised third superseding indictment should have been brought as a motion to dismiss under Federal Rule of Criminal Procedure 12, the court denied the Rule 29 motion on three grounds: (1) the government had provided Prado's defense counsel with actual notice of the victims’ identities, which was imputed to Prado; (2) the third superseding indictment's factual allegations were in any event sufficiently specific to give Prado constructive notice of the victims’ names, both by themselves and as supplemented by the case discovery; and (3) even if the third superseding indictment were deficient in some way, Prado had not demonstrated prejudice from any deficiency.

The district court also opined that, to the extent Prado wished to raise a claim of ineffective assistance of counsel arising out of any failure to provide him with the victims’ names, the claim was meritless. The court expressed doubt that, in fact, Prado had requested the victims’ names from his counsel with no response. But even if that had happened, the court concluded that Prado failed to show how his defense would have differed if he had sooner known the victims’ specific identities. The court also emphasized that there was overwhelming evidence incriminating Prado and that there was no reasonable possibility that the trial outcome would have been different if defense counsel had shared with Prado the names of the victims. See Porter v. McCollum, 558 U.S. 30, 40, 130 S.Ct. 447, 175 L.Ed.2d 398 (2009) (stating that, to establish prejudice as part of an ineffective assistance of counsel claim, a defendant must show “a reasonable probability that, but for counsel's unprofessional errors, the result of the proceeding would have been different” (quoting Strickland v. Washington, 466 U.S. 668, 694, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984))).

In appealing the district court's denial of his motion, Prado focuses on whether the revision of the third superseding indictment worked an impermissible direct amendment, a constructive amendment, or a prejudicial variance requiring reversal of his convictions. Following the parties’ lead, we assume (without deciding) that Prado preserved his challenge, thus triggering de novo review. See United States v. Dowdell, 595 F.3d 50, 66 (1st Cir. 2010).

Prado has not demonstrated that the district court erred in allowing the victims’ names to be added to the third superseding indictment. Certainly, these revisions did not work a constructive amendment or a prejudicial variance. “A constructive amendment occurs when the government's evidence or arguments or the court's jury instructions alter the terms of an indictment such that the defendant is effectively charged with a different offense than the one returned by the grand jury.” United States v. Katana, 93 F.4th 521, 530 (1st Cir. 2024). “A variance, by contrast, does not involve a change in the offense charged in the indictment. Rather, a variance occurs when the government relies at trial on different facts than those alleged in the indictment to prove the same offense.” Id. (citation omitted).

Here, neither the evidence nor the jury instructions supported a conviction for any offense other than the ones charged in the third superseding indictment. And Prado has identified no discrepancy between the facts proved at trial and those alleged in that indictment. Indeed, the addition of the names merely introduced details consistent with those already disclosed by the government. Therefore, there was neither a constructive amendment of the third superseding indictment nor a prejudicial variance between the facts alleged and the facts proven at trial.

That leaves us to consider whether the district court erred in allowing the government to directly amend the third superseding indictment to include the victims’ names. There was no error. While the Presentment Clause of the Fifth Amendment bars a court or prosecutor from amending an indictment without grand jury involvement, it “does not extend to alterations that are ‘merely a matter of form.’ ” Dowdell, 595 F.3d at 67 (quoting Russell v. United States, 369 U.S. 749, 770, 82 S.Ct. 1038, 8 L.Ed.2d 240 (1962)).

Thus, for example, we have declined to find reversible error where courts have permitted corrections to mistakes in indictments such as a reference to “cocaine” when “cocaine base” was intended, id. at 66-69; a reference to the wrong penalty provision that would apply to a charged count, see United States v. Eirby, 262 F.3d 31, 37-38 (1st Cir. 2001); the transposition of the names of two unindicted co-conspirators within a charged count, see United States v. Rivera-Ruiz, 244 F.3d 263, 271 (1st Cir. 2001); and the specification of an incorrect date on which the offense was committed, Jervis v. Hall, 622 F.2d 19, 22-23 (1st Cir. 1980). When a correction “leaves the substance of the charge unaffected, the switch does not usurp the prerogative of the grand jury.” Dowdell, 595 F.3d at 68 (cleaned up).

Here, adding the victims’ names to the third superseding indictment worked no change to the substance of the charged crimes. Indeed, it did not even correct a mistake; it only replaced pseudonyms with real names. Prado does not seek to explain how those changes might have negatively affected him other than to say, without elaboration, that earlier notice might have aided him in planning trial strategy or deciding whether to pursue a trial at all. But Prado offers no explanation -- nor can we conceive of one -- for how earlier notice of these victims’ identities would have substantively affected his trial strategy. And absent any such explanation as to how the revisions “bear[ ] on the substance of the charges,” id., we will not reverse the district court's decision to allow them.

In sum, we reject Prado's challenge to his convictions on grounds of impermissible direct amendment, constructive amendment, or variance.

B.

We next address Prado's disqualification argument. On June 1, 2021, shortly after Prado was indicted, attorney Joshua Levy entered an appearance on Prado's behalf as appointed counsel. About two months later, Levy moved to withdraw as Prado's counsel because he had decided to leave the private practice of law. The court granted Levy's motion. At some point thereafter, Levy joined the U.S. Attorney's Office for the District of Massachusetts. On January 11, 2022, he became the First Assistant U.S. Attorney for the District of Massachusetts.

Four days prior to Levy's appointment as First Assistant U.S. Attorney, in anticipation of Levy's appointment, the office's ethics advisor notified attorneys in the office that Levy would be “conflicted off” any case that he had handled in private practice. This meant that Levy would not supervise or discuss, receive any status reports, or be involved in any briefings or weekly updates regarding such a case. Prado's case was among those listed in the ethics advisor's notification as one from which Levy would be “walled off.”

On May 19, 2023, Levy became the Acting U.S. Attorney for the District of Massachusetts. Almost two months later, the Department of Justice (“DOJ”) recused the U.S. Attorney's Office for the District of Massachusetts from Prado's case, reassigned the case to the U.S. Attorney's Office for the District of Rhode Island, and authorized Zachary Cunha, the U.S. Attorney for the District of Rhode Island, to supervise and direct the case. But the recusal order also authorized the specific prosecutors and staff assigned to the case from the District of Massachusetts to remain trial counsel under U.S. Attorney Cunha's direction and supervision.

On August 9, 2023, Prado filed a pro se motion to dismiss the third superseding indictment. Prado asserted that, in disregard of the order walling off Levy from his case, Levy had communicated to the prosecution team certain directions and information that would be used against him. About a month later, the district court held a hearing on Prado's motion. At the conclusion of the hearing, at which the court heard from, among others, Prado and one of the prosecutors (whom the court swore in as a witness and permitted Prado to cross-examine), the court denied the motion. In doing so, the court found that the wall separating Levy from the prosecution team had not been breached and that Levy had not disclosed to the team any confidential information.

In appealing this ruling, Prado asks that we treat his motion to dismiss as a de facto motion to disqualify the District of Massachusetts prosecution team for a conflict of interest. The government agrees with this characterization and consents to us treating the issue as preserved. Accepting this framing, we conclude that Prado's argument nonetheless fails.

Prado does not challenge the district court's finding that Levy was effectively walled off from the prosecution team. Moreover, Prado explicitly acknowledges that disqualification of government counsel has been described as a “drastic measure” that courts should be hesitant to impose except where necessary. United States v. Bolden, 353 F.3d 870, 878 (10th Cir. 2003) (citation omitted). He also explicitly acknowledges authority establishing that, because the disqualification of government attorneys implicates the separation of powers, “the generally accepted remedy” is to disqualify only the conflicted government attorney and not the entire office. Id. at 879. Indeed, Prado even acknowledges that “[e]very circuit court found to have reviewed the court-ordered disqualification of an entire United States Attorney's office has reversed the disqualification.” E.g., United States v. Whittaker, 268 F.3d 185, 187, 196 (3d Cir. 2001); United States v. Vlahos, 33 F.3d 758, 759 (7th Cir. 1994); United States v. Caggiano, 660 F.2d 184, 185 (6th Cir. 1981); see also Bolden, 353 F.3d at 879 (making the same observation).

Immediately after recognizing this authority, Prado asserts, in a single sentence, that it is inapplicable here because “[t]he Department of Justice has already decided that [recusal] was necessary.” It appears to us that Prado is suggesting that the present situation should be governed by an all-or-nothing rule whereby, as a matter of law, no member of a U.S. Attorney's office may work on a case if the DOJ has recused the person's home office from supervising and directing the case. But Prado provides no support for such a rule, which is in evident tension with the cases we have just mentioned. Without more, we have no basis for setting aside Prado's convictions on this ground.

We therefore reject Prado's challenge to his convictions on the ground that the prosecution team should have been disqualified from the case.

C.

We next address Prado's argument that his sentence was procedurally unreasonable. The district court, which adopted the presentence report without change, determined Prado's sentence as follows. Prado's convictions on Counts One through Four, the conspiracy to commit wire fraud and substantive wire fraud violations, yielded a base offense level of seven. See U.S.S.G. §§ 2B1.1, 2X1.1. The court added twelve points to this base offense level because it concluded that Prado's offense conduct caused a loss (determined to be $400,224.27) of more than $250,000 but not more than $550,000. See id. § 2B1.1(b)(1)(G). The court also added two points because it concluded that the offenses involved ten or more victims, see id. § 2B1.1(b)(2)(A)(i), and two additional points because it concluded that the offense conduct involved sophisticated means, see id. § 2B1.1(b)(10)(C). These determinations combined to yield a total offense level of twenty-three.

The district court next calculated Prado's criminal history category. Although Prado had four prior convictions, the court determined that only one -- a 2015 Massachusetts conviction for operating a motor vehicle with his license revoked for being a habitual traffic offender, see Mass. Gen. L. ch. 90, § 23 (2024) -- qualified as a felony and thus gave rise to a single criminal history point, see U.S.S.G. §§ 4A1.1(c), 4A1.2(c). Prado's criminal record therefore fell within criminal history category I and, given a total offense level of twenty-three, Prado faced a guideline sentencing range (“GSR”) of forty-six to fifty-seven months on Counts One through Four. See id. ch. 5, pt. A. The court then imposed concurrent sentences of forty-six months of imprisonment -- the low end of the GSR -- on each of those counts.

Prado's convictions on Counts Five through Seven, the aggravated identity theft convictions, required sentences of two years of imprisonment that could run concurrently with each other, see 18 U.S.C. § 1028A(b)(4), but had to run consecutively to the sentences imposed on Counts One through Four, see id. § 1028A(a)(1), (b)(2); see also U.S.S.G. § 2B1.6 (stating that, for a conviction under 18 U.S.C. § 1028A, “the guideline sentence is the term of imprisonment required by statute”). The district court chose to make the two-year sentences on Counts Five through Seven concurrent to each other but, as required, consecutive to the forty-six-month sentences it imposed on Counts One through Four. Thus, the court imposed a total sentence of seventy months of imprisonment.

Prado says that this sentence was procedurally unreasonable for two reasons. First, he contends that the district court erred in finding that his conduct caused losses of between $250,000 and $550,000. In Prado's view, there was little to no loss caused by his conduct because customers got their rides and grocery deliveries, and the rideshare and grocery delivery companies were paid and thus profited from his scheme. Second, Prado argues, the court erred in assigning him a criminal history point because the evidence relied on by the presentence report to determine that his 2015 Massachusetts conviction as a habitual traffic offender qualifies as a felony -- a police report -- was insufficient to support the report's conclusion by a preponderance of the evidence. Even if we assume, solely for argument's sake, that the district court committed errors in calculating loss and Prado's criminal history, any such errors would be harmless.

With respect to the loss issue, the district court made the following, alternative determination at Prado's sentencing hearing:

I will say that if there was no loss, I would depart upward ․ because the offense level would substantially understate the seriousness of the offense, and I would depart upward to the level that the loss, actual loss, as I calculate it, would get to as the starting point or the guideline range that I'm calculating.

Thus, even if the court had accepted Prado's loss-calculation argument, it still would have used a GSR of forty-six to fifty-seven months as the basis for determining Prado's sentence on Counts One through Four.

Moreover, the district court made clear that it would not impose a sentence of fewer than forty-six months on these counts (and, therefore, a total sentence of not fewer than seventy months, when the mandatory consecutive twenty-four-month sentences on Counts Five through Seven are factored in):

And I'll tell you, if I have any concern about this sentence, it's that it's too low, not that it's too high ․ I was inclined to give you a longer sentence than 70 months. That's the low end of the guideline range which is, in this case, in my view, the minimum that's reasonable ․

We have repeatedly held that a district court's statements to the effect that it, “cognizant of the dueling guidelines calculation,” would impose the same sentence regardless of how a contested guideline calculation is resolved renders harmless an error in the calculation. United States v. Ouellette, 985 F.3d 107, 110 (1st Cir. 2021) (citation omitted); see also United States v. Ahmed, 51 F.4th 12, 22 (1st Cir. 2022); United States v. Ayala, 991 F.3d 323, 326-27 (1st Cir. 2021); United States v. Arif, 897 F.3d 1, 12 (1st Cir. 2018). Here, the court made such statements when it indicated that it would depart upward and use the same GSR regardless of how the loss-calculation issue was resolved, and when it further indicated that the total sentence selected was the minimum that would be reasonable under the circumstances. Prado's argument that it is not sufficiently clear that the court would impose the same sentence ignores the court's statement that seventy months was “the minimum that's reasonable.” Moreover, Prado makes no argument that a hypothetical upward variance to seventy months of imprisonment would have been impermissible. Thus, any potential error in the court's loss-calculation was harmless.

With respect to the criminal history issue, we note that Prado did not raise below the insufficiency argument that he makes on appeal. Nor does he argue on appeal for application of the plain-error review standard. He has thus waived the issue. See United States v. López-Felicie, 109 F.4th 51, 57 (1st Cir. 2024). And in any event, there was no plain error. We have held that “a court does not commit plain error when it incorrectly calculates and applies a criminal history score that nonetheless results in the defendant being placed in the correct Criminal History Category.” United States v. Delgado-Sánchez, 849 F.3d 1, 12 (1st Cir. 2017); see also United States v. Albanese, 287 F.3d 226, 229 n.1 (1st Cir. 2002) (per curiam) (similar, noting that “[a] remote possibility that the outcome could have been affected” in a situation such as this does not establish the prejudice that must be shown under plain-error review). Here, the district court's assignment of a single criminal history point for Prado's 2015 Massachusetts conviction as a habitual traffic offender did not affect his criminal history category, which remained category I. Thus, Prado has not demonstrated that the assignment of the point was prejudicial even if the assignment was in error.

We therefore reject Prado's challenges to the procedural reasonableness of his sentence.

D.

Finally, we address Prado's challenge to the substantive reasonableness of his sentence. Prado's seventy-month sentence was twenty-five months longer than any of the sentences imposed on his co-defendants, even those co-defendants who, according to Prado, had greater culpability. Prado contends that this disparity shows that the district court violated his due process rights by punishing him for going to trial. Prado also contends that the disparity shows that the court did not properly discharge its statutory obligation to “consider ․ the need to avoid unwarranted sentence disparities among defendants with similar records who have been found guilty of similar conduct.” 18 U.S.C. § 3553(a)(6).

The government concedes that Prado has preserved the right to challenge the length of his sentence as unreasonable, and that our review is therefore for an abuse of discretion. United States v. Coplin-Benjamin, 79 F.4th 36, 43 (1st Cir. 2023). But it contends that Prado has forfeited the right to make arguments for why his sentence was substantively unreasonable -- the disparity between his sentence and those of his co-defendants -- by not properly raising them before the district court. Prado responds that his trial counsel sufficiently raised the question of whether Prado was being penalized for going to trial to avoid the application of plain-error review.

We need not decide the preservation/standard-of-review issue. We assume for the sake of argument that Prado's arguments are preserved and, therefore, subject to review for an abuse of discretion. We also assume, although the government contests it, that Prado was similarly situated to co-defendants who received shorter sentences and, in fact, was less culpable than some of them. Even so, the district court did not abuse its discretion in sentencing Prado to seventy months of imprisonment.

At the sentencing hearing, the district court stated several times that it was not penalizing Prado for going to trial and that it would be imposing a sentence at the low end of the applicable GSR -- a sentence shorter than the court thought was otherwise warranted -- precisely because it wished to minimize the disparity between Prado's sentence and those of his co-defendants. The court noted, however, that Prado's decision to go to trial had led it to learn details about the conspiracy and offense conduct of which it was unaware when the co-defendants pleaded guilty and the court sentenced them. See Alabama v. Smith, 490 U.S. 794, 801, 109 S.Ct. 2201, 104 L.Ed.2d 865 (1989) (observing that, during a trial, a judge may come to learn facts about a defendant that undermine arguments for leniency that the defendant might plausibly have pressed in connection with a guilty plea). Indeed, the court opined that, if it had known at the time it accepted the co-defendants’ guilty pleas what it learned during Prado's trial, it might have imposed longer sentences on at least some of them.

Our review of the record persuades us that the district court was appropriately sensitive to the disparity issue, and that it fashioned a sentence which reasonably balanced the need for case-specific justice with comparative fairness considering the differing context in which the various defendants appeared before the court for sentencing. The court thus acted within its discretion in sentencing Prado to seventy months of imprisonment and did not punish him for going to trial.

We therefore reject Prado's challenge to the substantive reasonableness of his sentence.

III.

For the reasons we have stated, we affirm Prado's convictions and sentence.

AFRAME, Circuit Judge.