United States v. Francisco Colorado Cessa

861 F.3d 121, 2017 U.S. App. LEXIS 11303, 2017 WL 2742277

Opinion

STEPHEN A. HIGGINSON, Circuit Judge:

Defendant-Appellant Francisco Colorado Cessa (“Colorado”) appeals his conviction for conspiracy to commit money laundering under 18 U.S.C. § 1956(h). Colorado was convicted of participating in a. scheme to launder drug proceeds for Los Zetas, a Mexican drug cartel. On appeal, Colorado raises six issues: (1) whether the district court erred when it declined to order disclosure of certain Brady and Giglio material; (2) whether the district court erred by instructing the jury that it could infer intent to conceal funds from commingling legitimate and unlawful assets; (3) whether the prosecutor’s closing statement on the commingling inference requires a new trial; (4) whether the district court erred in finding that double jeopardy principles did not bar Colorado’s retrial; (5) whether the district court erred when it declined to dismiss the indictment based on alleged prosecutorial misconduct before the grand jury; and (6) whether the district court erred in ordering the forfeiture of some of Colorado’s property or in entering a money judgment against him. We remand for further findings on the Brady and Giglio claim. In light of our remand on the Brady claim, we do not reach Colorado’s challenge to the forfeiture order. We otherwise reject Colorado’s arguments.

I

Colorado is a businessman from Mexico who owns an oil-services business called ADT Petro Services (“ADT”). Around 2004, Colorado became associated with the Zetas. The Zetas import drugs from Colombia and export them to the United States. Colorado’s association with the Zetas arose out of his close friendship with Efrain Torres, a leader in the organization, who was also known as “Zeta 14” (the 14th member of the Zetas) or “La Chispa.” Colorado’s association continued, however, after Torres was murdered at the direction of Miguel Trevino (also known as “Zeta 40”) in 2007.

The Zetas engaged in a money-laundering operation that involved purchasing quarter horses — a type of racehorse — in the United States. The scheme was designed to conceal illegal drug money by repeatedly buying and reselling horses to “straw purchasers and shell companies”— a process that generated “clean” money, the origin of which was difficult to trace.

Colorado and others were first indicted in 2012. Colorado was charged with one count of conspiring to launder drug proceeds in violation of 18 U.S.C. § 1956(h). The indictment alleged that the conspiracy began “in or about 2008” and that its objective was “to launder U.S. currency gained from the sale of illegal controlled substances by Los Zetas to purchase, breed, train, and race quarter horses in the United States and Mexico.” Colorado and other conspirators were convicted in 2013, but this Court reversed Colorado’s conviction due to an improper jury instruction. See United States v. Cessa, 785 F.3d 165, 187 (5th Cir. 2015) (“Cessa I”).

On August 4, 2015, a grand jury returned a second superseding indictment, which again charged Colorado with violating § 1956(h) by conspiring to commit money laundering with the Zetas. The second superseding indictment mirrored the first indictment in almost every respect. However, the new indictment expanded the dates of the alleged conspiracy, this time alleging that the conspiracy began “in or about 2004.”

A jury found Colorado guilty after a nine-day trial. The district court sentenced Colorado to 200 months in prison, followed by three years of supervised release. The district court also ordered forfeiture of Colorado’s personal property and a $60 million money judgment. Colorado timely appealed.

II

Colorado first argues that the district court erred under Brady and Giglio by failing to order the Government to turn over certain interview memoranda related to Carlos Nayen, a cooperating Government witness who testified at the second trial. Government agents interviewed Nay-en at least nine times before trial. The agents generated FBI Forms 302 (official interview memoranda) for each of the nine interviews, the last occurring on December 18, 2013, almost two years before trial began. Before Nayen testified, Colorado, citing Brady and Giglio, moved for the Government to produce “all FBI-302s, DEA-6s, and similar interview memoran-da” related to Nayen to the court for in-camera review. 1 The next day, the district court granted the motion. Following Nay-en’s direct testimony; Colorado’s counsel asked the court to review the produced 302s with a “careful eye” in light of Nay-en’s direct testimony. The court responded by stating that nothing in the 302s was “helpful” to the defense, and accordingly, denied Colorado access to the interview memoranda. On appeal, we granted, over the Government’s objections, Colorado’s motion to view the 302s. Colorado now argues that the 302s contain material, exculpatory and impeachment evidence requiring that we vacate his conviction under Brady.

“We generally review whether the government violated Brady de novo, although even when reviewing a Brady claim de novo, we must proceed with deference to the factual findings underlying the district court’s decision!.]” United States v. Brown, 650 F.3d 581, 589 (5th Cir. 2011) (internal quotation marks and citations omitted). “But we have an exception to our general rule of de novo review: Where ... ‘a district court has reviewed potential Brady material in camera and ruled that the material was not discoverable, we review [that] decision only for clear error.’ ” Id. (alteration in original) (quoting United States v. Skilling, 554 F.3d 529, 578 (5th Cir. 2009), vacated in part on other grounds by Skilling v. United States, 561 U.S. 358, 130 S.Ct. 2896, 177 L.Ed.2d 619 (2010)). “The district court’s finding is clearly erroneous if, on the entire evidence, [the Court is] left with a ‘definite and firm conviction’ that a mistake has been committed.” Id. (quoting United States v. U.S. Gypsum Co., 333 U.S. 364, 395, 68 S.Ct. 525, 92 L.Ed. 746 (1948)).

Under Brady v. Maryland, a defendant’s due process rights are violated when the prosecution suppresses evidence that is exculpatory. 373 U.S. 83, 87, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963). The principle also applies to evidence that could be used to impeach prosecution witnesses. Giglio v. United States, 405 U.S. 150, 152-54, 92 S.Ct. 763, 31 L.Ed.2d 104 (1972). “To establish a Brady violation, a defendant must show: (1) the evidence at issue was favorable to the accused, either because it was exculpatory or impeaching; (2) the evidence was suppressed by the prosecution; and (3) the evidence was material.” United States v. Dvorin, 817 F.3d 438, 450 (5th Cir. 2016). “Evidence is material if there is a reasonable probability that, had the evidence been disclosed to the defense, the result of the proceeding would have been different.” Id. at 451 (internal quotation marks and citation omitted).

Although the Government claimed at oral argument that the district court considered all three prongs of the Brady analysis, the record does not support that proposition. The district court denied Colorado’s request for the 302s, saying: “I’ve read everything with the idea of how could any of that help you, and I couldn’t find anything that would help you. Found a lot that could hurt you but not much that could help you. Nothing that could help you.” On the face of the district court’s statement it considered only favorability. This is not a case where the district court reviewed all potential Brady material in-camera and denied discoverability without giving reasons or with reasons that could reasonably be read to reach all three Brady prongs. In that type of case, we might (in the absence of contrary evidence) presume that the district court did a full Brady analysis. Here, however, the district court gave reasons for denying discovery and couched its holding in terms of favora-bility only.

In addition, the timing of the district court’s decision indicates that the court did not assess materiality. The district court determined that the 302s were not discoverable immediately following Nayen’s direct examination. But much of what we focus on in assessing materiality is how the suppressed evidence relates to cross-examination. For example, we have asked whether the suppressed impeachment evidence is the “only avenue of impeachment” and whether the suppressed evidence impeaches an already impeached witness. United States v. Sipe, 388 F.3d 471, 483 (5th Cir 2004); accord United States v. Hughes, 230 F.3d 815, 819-22 (5th Cir. 2000). Further, even when our materiality jurisprudence extends beyond the scope of cross-examination, it asks questions that often depend on a full review of the trial evidence. For example, we ask whether the suppressed evidence impeaches a witness at the “heart of the government’s case,” whether the case is close on any element of the crime, and whether the testimony that could be attacked with the suppressed evidence is strongly corroborated. Sipe, 388 F.3d at 490, 492; United States v. Alexander, 681 Fed.Appx. 391, 399, 2017 WL 1031281, at *6 (5th Cir. 2017) (unpublished). These questions implicate the totality of the evidence presented at trial. Indeed, we have made clear that “[t]he materiality of Brady material depends almost entirely on the value of the evidence relative to the other evidence mustered by the state.” Banks v. Thaler, 583 F.3d 295, 321 (5th Cir. 2009) (quoting Smith v. Black, 904 F.2d 950, 967 (5th Cir. 1990), vacated on other grounds 503 U.S. 930, 112 S.Ct. 1463, 117 L.Ed.2d 609 (1992)). Accordingly, based both on its own words and the context of the trial, we conclude that the district court assessed the undisclosed interview memoranda for only fa-vorability.

The district court clearly erred in finding that the 302s were not favorable to the defense. As noted, evidence is favorable under Brady if it is either exculpatory or impeaching. Sipe, 388 F.3d at 477 (citing Strickler v. Greene, 527 U.S. 263, 281-82, 119 S.Ct. 1936, 144 L.Ed.2d 286 (1999)); see also United States v. Dillman, 15 F.3d 384, 390 (5th Cir. 1994) (“Although exculpatory and impeachment evidence fall within the purview of Brady, neutral evidence does not.”).

Some of Nayen’s statements in the 302s were exculpatory. At trial, Colorado’s defense theory was that he bought horses for the Zetas using his own (or his company’s) money. He explained that he spent millions of dollars on horses for the Zetas because he feared them. Nayen’s statements in the 302s supported both points. For example, one 302 reports that “Nayen claimed that Colorado only gave horses to Trevino and Lazcano [another Zeta leader] as a gift. Nayen said that when Trevino and Lazca-no would inquire as to the price of a horse, Colorado would give it to them and say it is a gift.” This statement corroborates Colorado’s defense theory that he did not buy horses using Zeta money. The 302s also suggest that Colorado feared the Zetas, reporting that “Nayen described Colorado as being in a constant state of anxiety after meeting with Trevino.” This statement supports Colorado’s theory that he gave horses to the Zetas — not to aid in the money laundering conspiracy — but out of fear. Because the 302s lent support to Colorado’s trial theory, we find that they were favorable.

The Government, however, argues the 302s are not truly exculpatory because nothing in the 302s goes to the heart of Colorado’s guilt. On the Government’s theory, Colorado’s fear¡ of the Zetas or the Zetas’ coercion of Colorado could be relevant only if Colorado requested a jury instruction on the affirmative defense of duress. Because Colorado disclaimed a duress instruction, the Government reasons that evidence of fear and coercion cannot be relevant and, therefore, cannot be exculpatory.

This argument misunderstands Colorado’s defense theory. Colorado did not concede that he joined the conspiracy but only under duress. Instead, Colorado argued that he did not join the conspiracy at all, claiming that he gave the Zetas gifts using his own money because he feared them. This argument is directed at elements of the money laundering conspiracy charge— namely the allegations that Colorado joined the conspiracy and did so with specific intent. See United States v. Threadgill, 172 F.3d 357, 366 (5th Cir. 1999) (noting that to prove a money laundering conspiracy, the government must show “(1) that there was an agreement between two or more persons to commit money laundering; and (2) that the defendant joined the agreement knowing its purpose and with the intent to further the illegal purpose”).

Defendants charged with participating in a conspiracy to launder money may argue that they did not “joint ] the conspiracy[J” Cessa I, 785 F.3d at 175; see also United States v. Blessing, 727 F.2d 353, 356 (5th Cir. 1984) (reversing conspiracy conviction when the government failed to prove that the defendant joined the conspiracy). In Cessa I, we held that evidence indicating that a defendant trained horses for the Zetas, was paid with Zeta drug money, and knew that he was paid with Zeta drug money, was insufficient to support a money laundering conspiracy charge. 785 F.3d at 175-80. In doing so, we made clear that involvement with the Zetas, even involvement with Zeta drug money, was insufficient to support a money laundering conviction because it did not show that the defendant knowingly joined the Zetas’ laundering conspiracy. Id. Colorado pursued a similar argument in this case. He argued that although he knew the Zetas engaged in drug trafficking, he did not join the money laundering conspiracy. Instead, Colorado claimed that his only transactions with the Zetas were not money laundering transactions, but gifts. And he explained the gifts by saying that he gave millions to the Zetas out of fear. If the jury believed Colorado, it could have acquitted on that theory. See id.; cf., e.g., United States v. Espinoza-Seanez, 862 F.2d 526, 538 (5th Cir. 1988) (“It is not enough that the defendant merely associated with someone who was knowingly participating in a conspiracy .... ”); Panci v. United States, 256 F.2d 308, 312 (5th Cir. 1958) (reversing conspiracy conviction when evidence “amount[ed] to no more than that [defendant] was seen associating with characters of low repute”). 2

Moreover, even if the Government were correct that evidence of fear and coercion was not relevant absent a duress instruction, we would still find that evidence supporting duress was favorable to the defense at Brady step one. A defendant’s decision to pursue or disclaim an affirmative defense instruction is not made in a vacuum; defendants must evaluate the evidence available to them and determine whether seeking the instruction is likely to help or harm their cause. Denying defendants access to evidence they are entitled to under Brady can significantly change this risk calculus. For this reason, courts routinely find that evidence supporting an affirmative defense is exculpatory and, therefore, favorable under Brady. See, e.g., Mahler v. Kaylo, 537 F.3d 494, 504 (5th Cir. 2008) (finding Brady error when the prosecution withheld evidence that supported self-defense). Accordingly, evidence may be favorable under Brady if it would have supported an affirmative defense— even if the defendant, acting without the Brady evidence, later disclaims the defense. 3 And that is exactly what happened here: the district court denied the defense access to the alleged Brady material before ruling on whether to give a duress instruction. We therefore hold that the evidence contained in the 302s is exculpatory and thus is favorable under Brady.

Furthermore, because Nayen testified inconsistently with the 302s, the 302s could have impeached Nayen’s testimony. “It is a well-established principle of evidence that prior inconsistent statements of a witness are admissible to impeach that witness.” United States v. Palacios, 556 F.2d 1359, 1363 (5th Cir. 1977). At trial, Nayen 'testified inconsistently with the 302s in two respects.

First, Nayen testified that on three occasions the Zetas delivered millions of dollars to Colorado, which Colorado used, in part, for horse payments. The 302s tell a somewhat different story. In the first 302, Nayen contradicts his trial testimony by saying that Colorado bought horses for the Zetas as gifts — implying that Colorado used his own money, not Zeta money, to buy the horses.

Second, Nayen testified that Colorado was “friends” with Zeta 40 (also called Trevino), and accordingly, Zeta 40 would not have threatened or coerced Colorado. Again, the 302s tell a different story. There, Nayen states that Zeta 40 said that he “was not [Colorado’s] friend” and that Zeta 40 “always wanted to kill [Colorado].” Nayen further states that Colorado was “in a constant state of anxiety after meeting with [Zeta 40].”

The Government questions the impeachment value of the 302s, arguing that “in its proper evidentiary context, the ‘suppressed’ evidence of Nayen’s statements about Miguel Trevino [Zeta 40] did not impeach Nayen’s trial testimony.”

Preliminarily, the Government does not argue that Nayen’s statements about the gifts are not favorable impeachment evidence. Accordingly, even if the Government were correct that the friends-statements are not favorable, we would still find the 302s as a whole contained favorable impeachment evidence.

Regardless, we hold that the friends-statements were also favorable to Colorado. According to the Government, Colorado and Zeta 40 may have been “not friends” before early 2007, but that changed sometime after Colorado and Zeta 40 met in Tampico in March 2007. In early 2007, Zeta 40 had Torres (Zeta 14 or “La Chispa”), who was friends with Colorado, murdered. Following Torres’s death, Nay-en, Colorado, and Colorado’s business partner, Tavo, were summoned to Tampico to discuss their friendship with Torres with Zeta 40. Nayen missed the meeting and Tavo was killed there. The Government concedes that before the Tampico meeting, Colorado may have feared Zeta 40. But the Government solicited evidence from a number of witnesses, including Nayen, that they saw Colorado and Zeta 40 behaving like they were friends following the Tampico meeting. Based on those witnesses’ testimony, the Government argues that “Nayen’s testimony about the friendship between the defendant and Trevino, when placed in its proper context, addressed their relationship after the Tampico tension had dissipated.”

The Government’s contextualization of Nayen’s statements in the 302s does not make the 302s unfavorable under Brady. Nayen’s statements in the 302s are broad enough to encompass the time period following the Tampico meeting: Nayen said that Colorado was in a “constant state of anxiety after” the Tampico meeting and that Zeta 40 “always wanted to kill” Colorado “through the course” of their relationship. Those statements impeach Nayen’s testimony that Colorado and Zeta 40 were friends and that Zeta 40 never threatened Colorado. Maybe, if confronted with his statements in the 302s,. Nayen would have explained that his statements in the 302s concerned the time period before the Tam-pico meeting, and maybe the jury would have credited that explanation. But it would have been favorable to the defense to be able to demand that explanation and test it with the jury. See, e.g., United States v. Campos, 20 F.3d 1171, 1994 WL 144866, at *13-15 (5th Cir. 1994) (unpublished) (evidence of prior inconsistent statements is favorable to the defense under B rady); United States v. Weintraub, 871 F.2d 1257, 1260 (5th Cir. 1989) (same). 4

The district court’s analysis ended at the favorability step — declining to address either suppression (the Government appears to have arguments that some, but not all, of the statements in the 302s were not suppressed) 5 or materiality. We hold that the district court clearly erred in concluding that the 302s were not favorable.

In this posture, we often proceed to analyze the suppression and materiality ourselves, but we are not required to do so. Compare Smith v. Dep’t of Corr., 572 F.3d 1327, 1348-49 (11th Cir. 2009) (“We could conduct the cumulative materiality analysis involving the six pieces of withheld evidence ourselves, but we think it would be beneficial to remand the case so that the district court can do that in the first instance. The district court has not yet performed that analysis and in the circumstances of this case we think that resolution of the issue will benefit from a two-tiered consideration.”), with Hughes, 230 F.3d at 819-22 (conducting a materiality inquiry in the first instance). We believe that remand is the proper course here for several reasons.

First, the Government acknowledged at oral argument that a remand would be proper to allow the district court to consider the interview notes corresponding to the 302s. This acknowledgment makes sense because there is reason to believe that the interview notes may require a more probing Brady analysis. 6 Interview notes may be discoverable under Brady if they are inconsistent with the content of the corresponding 302s. See United States v. Brown, 303 F.3d 582, 592-93 (5th Cir. 2002). At oral argument, the Government indicated that it had not thoroughly looked through all of the interview notes. Although Special Agent Scott Lawson authored each of the 302s, and a post-oral argument affidavit indicates that Lawson was present at each of the interviews, 7 the Government separately confirmed that the “only” note takers in the interviews were the prosecuting attorneys. This might not be problematic, but here, Lawson did not generate the 302s, produced as “interview memoranda,” until more than six months after the interviews: the first seven interviews occurred between November 27, 2012 and February 12, 2013; Lawson did not begin to generate 302s until July 2013. And in urging that the 302s would not be material, the Government itself disclaimed the accuracy of the 302s, arguing that impeaching Nayen with the 302s would have been ineffective because the 302s were “written long after the interviews.” Accordingly, in light of the unique circumstances surrounding the interview memo- randa here, including the Government’s disclaiming the accuracy of the 302s and acknowledging that it had not thoroughly reviewed the interview notes taken by the prosecuting attorneys for use a half year later in Lawson’s 302s, we conclude that remand for full assessment of the Nayen interview memoranda is necessary. Because Brady errors must be assessed cumulatively, see Skilling, 554 F.3d at 590, remand is the best course to allow the district court to consider all Brady material at the same time.

The district court is also in a better position to conduct the required suppression and cumulative materiality analysis because the district court had the opportunity to observe the trial. See, e.g., United States v. Rodriguez, 496 F.3d 221, 227 (2d Cir. 2007) (remanding to the district court to make a materiality and suppression determination in a Brady case because “[t]he district court is far better positioned to make t[he] appraisal”); Smith, 572 F.3d at 1348-49 (noting that materiality assessment would benefit from “two-tiered consideration”).

Accordingly, we remand the Brady claim to the district court to conduct a suppression and materiality analysis. On remand, the district court may; but need not, review the contemporaneous prosecutor notes corresponding to the later-generated agent 302s in camera. 8

III

Colorado next argues that a new trial is needed because the district court gave an improper instruction concerning what inferences the jury could draw about Colorado’s intent from his commingling illegal proceeds with legitimate funds (the “commingling instruction”). We find no error.

This is not the first time we have considered Colorado’s objections to the district court’s instructions on the commingling of assets. At Colorado’s first trial, the district court instructed: “the commingling of illegal proceeds with legitimate business funds is evidence of intent to conceal or disguise. Therefore, it would not be a defense that legitimate funds were also involved in a transaction involving illegal and legitimate funds combined.” Cessa I, 785 F.3d at 184. Colorado challenged the instruction on appeal. We noted that “ ‘[e]vi-dence that the defendant commingled illegal proceeds with legitimate business funds is sufficient to support a conviction under § 1956,’ particularly when the defendant also subsequently used the commingled funds to purchase equipment to further illegal activities.” Id. at 185 (quoting United States v. Rodriguez, 278 F.3d 486, 491 (5th Cir. 2002)). However, we held that the instruction was an abuse of discretion because it did not “make clear that the inference is permissive and not mandatory.” Id. at 186. Finding that the error was not harmless, we vacated and remanded. Id. at 187.

At Colorado’s second trial, the district court gave an amended commingling instruction:

With respect to the fourth element, the commingling of illegal proceeds with legitimate business funds may constitute evidence of intent to conceal or disguise. If you find that someone engaged in such commingling in this case, then you may find that the person did so with the intent to conceal or disguise the nature, the location, the source, the ownership, or the control of the proceeds. You are not obliged to so find, however. You are the exclusive judges of the facts proved, and in deciding what inference to draw, if any, from any such commingling, you should carefully consider all of the evidence in this case, including any explanatory or mitigating circumstances, and giving the evidence whatever weight you think it deserves.

Jury instructions are reviewed “for abuse of discretion, affording substantial latitude to the district court in describing the law to the jury.” United States v. Wright, 634 F.3d 770, 774 (5th Cir. 2011) (citation omitted). This Court must determine whether the “charge, as a whole, is a correct statement of the law and whether it clearly instructs jurors as to the principles of law applicable to the factual issues confronting them.” United States v. Mendoza-Medina, 346 F.3d 121, 132 (5th Cir. 2003) (quoting United States v. Lara-Velasquez, 919 F.2d 946, 950 (5th Cir. 1990)). A jury instruction must be “legally accurate” and “factually supportable” — a district court “may not instruct the jury on a charge that is not supported by evidence.” Cessa I, 785 F.3d at 185 (quoting Mendoza-Medina, 346 F.3d at 132). “In assessing whether evidence sufficiently supports a particular jury instruction, this Court views ‘the evidence and all reasonable inferences that may be drawn from the evidence in the light most favorable to the Government.’ ” Id. (quoting Mendoza-Medina, 346 F.3d at 132).

Colorado challenges the amended commingling instruction on three grounds. First, he argues that the instruction was not supported by the evidence, claiming that “there was no evidence that drug proceeds went into any account from which horse payments came.” Second, he contends that “the commingling instruction improperly highlighted one inference among many the jury could draw.” And third, he claims that the instruction was constitutionally infirm because commingling, at least here, does not make an intent to conceal more likely. We reject each claim.

First, the instruction was supported by the evidence. Evidence of commingling “illegal proceeds with legitimate business funds is sufficient to support a conviction under § 1956.” Rodriguez, 278 F.3d at 491 (citing United States v. Willey, 57 F.3d 1374, 1386 (5th Cir. 1995); United States v. Jackson, 935 F.2d 832, 840 (7th Cir. 1991)). Indeed, “[t]he government is under no duty to trace the individual funds and ‘it is not necessary that a transaction be examined wholly in isolation if the evidence tends to show that it is part of a larger scheme that is designed to conceal illegal proceeds.’ ” Id. (quoting Willey, 57 F.3d at 1386). Therefore, if there was evidence that Colorado commingled drug proceeds with his ■ legitimate business funds, the instruction was appropriate.

There was. For example, Nayen testified that when Colorado agreed to hold $10 million for Zeta 40, Colorado told Zeta 40 “this time ... I’m going tó handle it the way I need to handle it. I don’t want it to be like the last time. I’m going to put it in my account, I’ll handle it.” The jury also heard significant evidence indicating that the Zetas deposited drug money into ADT controlled accounts as loans and that ADT funds were commingled with Colorado’s personal funds. There was also ample evidence that Colorado purchased horses for the Zetas. Colorado’s investment adviser and an IRS investigator corroborated these loans and purchases. For example, Colorado’s investment adviser in the United States testified that ADT’s financial documents reflected the mixing of business accounts and personal expenses. An IRS investigator testified that Colorado’s balance sheets showed he would not have been able to fund ADT and purchase over $10 million in horses from 2008 to 2012, as ADT merely “broke even” during that time. Accordingly, we find that sufficient evidence supported the commingling instruction.

Second, our precedent forecloses the argument that the commingling instruction unfairly highlighted one inference to -the jury or was constitutionally infirm under Leary v. United States, 395 U.S. 6, 36-37, 89 S.Ct. 1532, 23 L.Ed.2d 57 (1969), which held that the Due Process Clause requires that permissive inferences be based in fact. In Cessa I we explicitly approved of a permissive inference instruction on commingling, so long as the instructions “included ... language clarifying that the inference was permissive and not mandatory.” Cessa I, 785 F.3d at 186.

Moreover, even if we were to reexamine Cessa I, we would not find that the commingling instruction was an abuse of discretion. The commingling instruction here did not unfairly highlight one inference to the jury. Colorado points to the Ninth Circuit’s opinion in United States v. Rubio-Villareal, 967 F.2d 294 (9th Cir. 1992) (en banc). There, the court held that a permissive inference instruction that allowed the jury to infer knowledge of drugs from the fact that the defendant was the driver of the vehicle containing the drugs was an abuse of discretion because (1) “the instruction constituted an intrusion on the jury’s deliberative process because it effectively told the jury in this case that the judge thought there was sufficient evidence to convict the defendant!,]” and (2) the instruction allowed the jury to ignore evidence outside the permissive inference. Id. at 295, 299-300 However, subsequent cases from the Ninth Circuit have distinguished Rubio-Villareal, and approved of permissive inference instructions that (1) emphasize that the jury is the ultimate fact finder; (2) remind the jury that explanatory or mitigating facts may erode the power of the permissive inference; and (3) remind the jury .to consider all of the evidence. See, e.g., United States v. Beltran-Garcia, 179 F.3d 1200, 1205 (9th Cir. 1999); United States v. Houser, 130 F.3d 867, 870 (9th Cir. 1997); United States v. Warren, 25 F.3d 890, 899 (9th Cir. 1994). Even assuming that Rubio-Villareal and its progeny apply, but see United States v. Parker, 32 F.3d 395, 402 (8th Cir. 1994) (rejecting Rubio-Villareal), the instruction here did all that the Ninth Circuit requires. The instruction told the jury (1) that they were “the exclusive judges of the facts proved”; (2) to consider “explanatory or mitigating circumstances”; and (3) to “consider all of the evidence ... giving [it] whatever weight you think it deserves.”

Neither did the commingling instruction run afoul of Leary’s requirement that permissive inference instructions be based in fact. “A permissive inference violates the Due Process Clause only if the suggested conclusion is not one that reason and common sense justify in light of the proven facts before the jury.” Francis v. Franklin, 471 U.S. 307, 314-15, 105 S.Ct. 1965, 85 L.Ed.2d 344 (1985); see also Barnes v. United States, 412 U.S. 837, 845, 93 S.Ct. 2357, 37 L.Ed.2d 380 (1973) (noting that permissive inferences may be based in “common sense and experience”); Smith v. Stalder, 26 F.3d 1118, 1994 WL 286165, at * 1 (5th Cir. 1994) (unpublished) (noting that permissive inference instructions are unconstitutional only if “reason and common sense do not justify the suggested conclusion”). The inference that commingling signals an intent to conceal is based in reason and common sense; indeed, both we and other courts have relied on commingling of assets to reject sufficiency challenges to money laundering convictions. See, e.g., United States v. Shepard, 396 F.3d 1116, 1121 (10th Cir. 2005) (“This circuit has noted that deposit ing illegal proceeds.into the bank account of a legitimate business may support the inference of an intent to conceal.”); Rodriguez, 278 F.3d at 491; United States v. Bowman, 235 F.3d 1113, 1117 (8th Cir. 2000) (“Depositing unlawfully obtained money into a seemingly legitimate business account and passing the money off as funds of a legitimate business is sufficient evidence to allow a reasonable jury to conclude that Mr. Bowman intended to conceal the stolen nature of the money and its true ownership.” (internal citation, quotation marks, and alterations omitted)); Willey, 57 F.3d at 1386 (“Evidence that the defendant commingled illegal proceeds with legitimate business funds has been held to be sufficient to support the design element.”); United States v. Termini, 992 F.2d 879, 881 (8th Cir. 1993) (“We reject Termini’s contention that the evidence of commingling of legal and illegal funds was insufficient to support a jury finding of concealment.”); United States v. Posters N Things Ltd., 969 F.2d 652, 661 (8th Cir. 1992) (“By conducting the financial end of her drug paraphernalia business in this way — commingling in one account, in an indistinguishable way, legitimate business receipts and illegitimate receipts — a reasonable juror could infer that Acty’s actions were designed to disguise the nature or the source of the proceeds from her unlawful business.”), aff'd 511 U.S. 513, 114 S.Ct. 1747, 128 L.Ed.2d 539 (1994); Jackson, 935 F.2d at 840 (“Indeed, the commingling in this case is itself suggestive of a design to hide the source of ill-gotten gains that the government must prove under § 1956(a)(1)(B)(i).”).

Accordingly, even were we able to reconsider Cessa I, we would hold that the district court did not abuse its discretion in giving the commingling instruction.

IV

Colorado next argues that the Government’s statements at closing require a new trial. During closing, the Government said “[a]nything that comes out of ADT that’s been capitalized by drug proceeds, I submit to you that once it gets commingled with dirty funds, anything that comes out of that company is, in fact, money laundering.” Colorado argues that this statement both converted the permissive instruction into an unlawful mandatory one and impermissibly broadened the scope of the commingling instruction by shifting the presumption from one element of money laundering — intent to conceal or disguise — to the entire crime and by extending the inference to cover a commingled account even after the unlawful funds had dissipated. Colorado objected to the prosecutor’s statement at trial and asked for the district court either to not give the commingling instruction or for “some corrective language.” We review for abuse of discretion. See United States v. Gracia, 522 F.3d 597, 600 n.2 (5th Cir. 2008); United States v. Stouffer, 986 F.2d 916, 926 (5th Cir. 1993).

We have a two-step process for analyzing alleged prosecutorial misconduct. See United States v. Fields, 483 F.3d 313, 358 (5th Cir. 2007). “First, we assess whether ‘the prosecutor made an improper remark.’” Id. (quoting United States v. Insaulgarat, 378 F.3d 456, 461 (5th Cir. 2004)). “In assessing whether statements made by a,prosecutor were improper, it is necessary to look at them in context.” United States v. Gallardo-Trapero, 185 F.3d 307, 320 (5th Cir. 1999); see also United States v. Delgado, 672 F.3d 320, 335 (5th Cir. 2012) (“[C]ontext is crucial to determining the effect of the statement”). To properly gauge context, we “consider[ ] a prosecutor’s closing argument as a whole and in the context of the trial.” United States v. Boyd, 773 F.3d 637, 645 (5th Cir. 2014).

If we conclude that a prosecutor’s statement was improper, “then we ask whether the defendant was prejudiced. ... [T]he prejudice step of the inquiry sets a high bar[.]” Fields, 483 F.3d at 358. “The determinative question is whether the prosecutor’s remarks cast serious doubt on the correctness of the jury’s verdict.” United States v. Holmes, 406 F.3d 337, 356 (5th Cir. 2005) (quoting United States v. Iredia, 866 F.2d 114, 117 (5th Cir. 1989)). “We generally look to three factors in deciding whether any misconduct casts serious doubt on the verdict: ‘(1) the magnitude of the prejudicial effect of the prosecutor’s remarks, (2) the efficacy of any cautionary instruction by the judge, and (3) the strength of the evidence supporting the conviction.’ ” Fields, 483 F.3d at 358 (quoting United States v. Mares, 402 F.3d 511, 515 (5th Cir. 2005)).

The prosecutor’s statement did not improperly convert the permissive inference instruction into a mandatory one. The prosecutor indicated, in the challenged statement itself, that she was inviting the jury to make an inference, and therefore, mimicked, rather than obfuscated, the commingling instruction that the jury would later hear. By prefacing her evidentiary conclusion with the phrase “I submit to you,” the prosecutor indicated to the jury that she was making an inferential argument using the evidence adduced at trial. See United States v. Perez-Solis, 709 F.3d 453, 467-68 (5th Cir. 2013) (“The challenged excerpt does not say T think that’ or T believe.’ It says T submit to you’ — which a prosecutor making an argument from evidence necessarily does. The statement was mere argument — not misconduct^encouraging a common-sense inference from the record.”); cf. United States v. Zabaneh, 837 F.2d 1249, 1260-61 (5th Cir. 1988) (examining a closing argument statement that began “I would submit to you” and concluding that the statement “advis[ed] the jury to consider” certain facts “in deciding how much weight” to give the evidence). That is, all the challenged statement communicates is the conclusion that the prosecutor asked the jury to draw from the evidence of commingling. Because that conclusion was permissible, we find no error. See United States v. Morris, 568 F.2d 396, 402 (5th Cir. 1978) (“[A]n attorney properly may state, T believe that the evidence has shown the defendant’s guilt[.]’ ”).

Reading the challenged statement in context confirms this innocuous reading. Immediately before the challenged statement, the prosecutor told the jury that she “anticipated an instruction on commingling.” And immediately surrounding the challenged statement, the prosecutor summarized the trial evidence indicating that Colorado commingled drug proceeds with ADT money. This context suggests that the prosecutor was alerting the jury to the permissive inference blessed by the commingling instruction. Accordingly, we conclude that the challenged statement did not improperly convert the permissive instruction to a mandatory one.

Neither did the challenged statement impermissibly broaden the scope of the permissive inference. The commingling instruction allowed the jury to infer a single element of money laundering, 9 intent to conceal or disguise, from commingling unlawful proceeds with legal funds. Colorado argues that because the challenged statement says “I submit to you that once it gets commingled with dirty funds, anything that comes out of that company is, in fact, money launderingl,]” the prosecutor’s statements broadened the permissible inference from going to a single element, to going to the whole crime. Context belies Colorado’s suggestion. Indeed, immediately after the challenged statement, the prosecutor made clear that she was talking about intent. Mqreover, the remainder of the closing argument makes clear that the jury could not convict based on only commingling, outlining the numerous elements that the Government needed to prove, that proving money laundering was not enough to convict on money laundering conspiracy, and multiple pieces of evidence beyond the commingling. Likewise, the prosecutor’s suggestion that the jury could infer intent from commingling did not require the jury to find intent even after unlawful funds had dissipated. Indeed, throughout the closing, the prosecutor highlighted the continued replenishment of illegal proceeds into ADT account. Taking the closing as a whole, we find no misconduct.

In any event, we would find any error in the closing harmless. With regard to prejudicial effect, the prosecutor’s statement that the case would “really come down to one thing” which was whether Colorado “intend[ed] to join th[e] conspiracy,” highlights the prosecution’s burden to prove that element of the offense and tends to rebut Colorado’s argument that the' jury would have assumed any evidence of commingling should result automatically in a guilty verdict. Moreover, the jury instructions made explicit that any inference drawn from the commingling evidence was permissive. And, during.closing argument, the prosecutor advised the jury that the judge’s “word is law,” and, in the event of a conflict between the lawyer and the trial judge’s articulation of the law “it’s his law and his words that you go with.”

V

Colorado next argues that under principles of double jeopardy and collateral es-toppel, he should not have been retried on the theory that he knowingly used illegal proceeds to purchase horses. Colorado argues that because in Cessa I we stated that the Government failed to submit sufficient evidence that he bought horses using illegal proceeds, the Government should have been barred from arguing the same theory on retrial. We disagree.

In Cessa I, we denied Colorado’s sufficiency challenge:

Colorado also argues that insufficient evidence supports his conviction because the government failed to prove that Colorado knowingly used illegal proceeds to purchase horses. Although we agree with Colorado’s argument that the government failed to introduce sufficient evidence on that point — which was the government’s theory of the case — we hold that the government’s evidence supports Colorado’s conviction on a different theory; • namely, that Colorado purchased the horses with the knowledge that they would be used as part of a larger money-laundering conspiracy. This knowledge is some circumstantial evidence that he voluntarily joined a conspiracy knowing its purpose was to conceal the source or nature of illegal funds. However, the issue is close, as the government offered weak circumstantial evidence based largely on Colorado’s personal relationships with Los Zetas’ leaders that Colorado’s knew the conspiracy’s purpose was to conceal the nature or source of illegal proceeds. As we explain ... the erroneous jury instruction on commingling, viewed in light of the sparse evidence of Colorado’s guilt, requires that Colorado’s conviction be vacated.

Cessa I, 785 F.3d at 182 (internal citations omitted).

The doctrine of double jeopardy bars retrial when the defendant has been acquitted, or when an appellate court reverses “on grounds of evidentiary insufficiency.” Burks v. United States, 437 U.S. 1, 16, 98 . S.Ct. 2141, 57 L.Ed.2d 1 (1978). However, double jeopardy does not apply when a “conviction is set aside because of an error in the proceedings leading to conviction.” Id. at 14, 98 S.Ct. 2141 (quoting United States v. Tateo, 377 U.S. 463, 465, 84 S.Ct. 1587, 12 L.Ed.2d 448 (1964)).

Relatedly, “[t]he doctrine of collateral estoppel is one aspect of the protection afforded by the double jeopardy clause.” United States v. Price, 750 F.2d 363, 365 (5th Cir. 1985). Collateral estop-pel — or issue preclusion — is the principle that “when an issue of ultimate fact has once been determined by a valid and final judgment, that issue cannot again be litigated between the same parties in any future lawsuit.” Ashe v. Swenson, 397 U.S. 436, 443, 90 S.Ct. 1189, 25 L.Ed.2d 469 (1970). “This court has interpreted Ashe to require a twofold inquiry for analyzing double jeopardy claims. First, the court must determine what, if anything, the jury necessarily decided in the first trial.” United States v. Tran, 433 Fed.Appx. 227, 230 (5th Cir. 2011) (unpublished) (citing Bolden v. Warden, W. Tenn. High Sec. Facility, 194 F.3d 579, 583-84 (5th Cir. 1999)). “Second, a court must determine ‘whether the facts necessarily decided in the first trial constitute essential elements of the offense in the second trial.’ ” Id. (quoting Bolden, 194 F.3d at 584). Colorado “bears the burden of demonstrating that the factual issue allegedly barred by collateral estoppel “was actually decided in the first proceeding.’ ” Garcia v. Dretke, 388 F.3d 496, 501 (5th Cir. 2004) (quoting Dowling v. United States, 493 U.S. 342, 350, 110 S.Ct. 668, 107 L.Ed.2d 708 (1990)). “This burden requires a defendant to prove that a second jury necessarily made a finding of fact that contradicted a finding” made at the first trial .Id. “The question of whether [the] prosecution violates the Double Jeopardy Clause is one of law and is thus reviewed de novo.” United States v. Cihak, 137 F.3d 252, 257 (5th Cir. 1998).

Collateral estoppel did not bar Colorado’s retrial, even on the theory that he used drug proceeds to buy horses. “The doctrine of collateral estoppel as delineated in Ashe ... deals with facts not theories.” United States v. Ragano, 520 F.2d 1191, 1198 (5th Cir. 1975); see also Wright v. Whitley, 11 F.3d 542, 546 (5th Cir. 1994) (“[T]he Ashe holding only bars relitigation of a previously rejected factual allegation where that fact is an ultimate issue in the subsequent case.”). No facts (pertaining to Colorado) were conclusively decided in Cessa I. True, we said in Cessa I that the Government did not introduce sufficient evidence to show that Colorado used drug proceeds to buy horses, but that fact was not part of our holding. 785 F.3d at 182. Indeed, we held that the Government put forward sufficient evidence to convict, and reversed on only instructional error. Id. at 182, 187. Likewise, whether Colorado used drug proceeds to purchase horses was not necessarily at issue at the second trial. As we said in Cessa I, the jury could convict on the theory that “Colorado purchased the horses with the knowledge that they would be used as part of a larger money-laundering conspiracy.” Id. at 182. And accordingly, the Government repeatedly, and correctly, eschewed the burden to trace drug funds through Colorado and then back to the Zetas, in the form of horses. Because no facts actually decided in Cessa I were at issue in Colorado’s trial, double jeopardy does not apply.

VI

Additionally, Colorado argues that the district court should have dismissed the Second Superseding Indictment because of prosecutorial misconduct before the grand jury. Because we find no material misconduct, we disagree.

Colorado alleges three types of prosecu-torial misconduct before the grand jury. First, he argues that the prosecutor improperly informed the grand jury of Colorado’s prior indictment 'and conviction. Specifically, the prosecutor told the grand jury that the appeals of all the original defendants in the conspiracy “were affirmed except for one guy, Francisco Antonio Colorado Cessa. And so the Court came back and said, ‘Even though your evidence is sufficient on this guy and all the other guys, you had an improper jury instruction.’ ” He also argues that many statements made by the prosecutor were “prejudicial and misleading.” Finally, Colorado points out several leading questions that he contends functioned as “unsworn assertions of fact thinly disguised as questions to the [witness].”

These were essentially the argu-. ments Colorado made below. The district court denied Colorado’s motion to dismiss the indictment. “The denial of a motion to dismiss the indictment for Governmental misconduct is reviewed de novo.” United States v. Forte, 65 Fed.Appx. 508, 2003 WL 1922910, at *5 (5th Cir. 2003) (unpublished) (citing United States v. Johnson, 68 F.3d 899, 902 (5th Cir. 1995)). However, we review any district court factual determinations concerning prosecutorial misconduct before the grand jury for clear error. United States v. Bourgeois, 950 F.2d 980, 984 (5th Cir. 1992).

“[A]s a general matter, a district court may not dismiss an indictment for errors in grand jury proceedings unless such errors prejudiced the defendant ].” Bank of Nova Scotia v. United States, 487 U.S. 250, 254, 108 S.Ct. 2369, 101 L.Ed.2d 228 (1988). “[Dismissal of the indictment is appropriate only ‘if it is established that the violation substantially influenced the grand jury’s decision to indict,’ or if there is ‘grave doubt’ that the decision to indict was free from the substantial influence of such violations.” Id. at 256, 108 S.Ct. 2369 (quoting United States v. Mechanik, 475 U.S. 66, 78, 106 S.Ct. 938, 89 L.Ed.2d 50 (1986) (O’Connor, J., concurring)). 10 Put another way, “[w]hether or not prosecutorial misconduct prejudiced a defendant depends on whether it affected the grand jury’s decision to indict.” United States v. Whitfield, 590 F.3d 325, 359 (5th Cir. 2009).

First, although the prosecutor mentioned Colorado’s first trial, he explic itly instructed the grand jury: “your job is still to look at this case fresh.” Moreover, the prosecutor explained that the dates on the new indictment had been expanded after the conviction was reversed, and the prosecutor highlighted that each indictment had “to be considered independently.” Overall, as the district court concluded, it appears that the prosecutor “repeatedly emphasized to the grand jury that they were more than a rubber stamp.” Given this repeated emphasis of the grand jury’s independent role, mentioning Colorado’s prior conviction and appeal did not substantially influence the grand jury’s decision to indict. See Nova Scotia, 487 U.S. at 256, 108 S.Ct. 2369.

Second, the prosecutor did not knowingly present material false and misleading information to the grand jury. We have “refuse[d] to adopt the proposition that, absent perjury or government misconduct, an indictment is flawed simply because it is based on testimony that may later prove to be questionable.” United States v. Sullivan, 578 F.2d 121, 124 (5th Cir. 1978). Indeed, even perjured grand jury testimony does not demand dismissal of the indictment under the district court’s supervisory power unless the perjury was “knowingly sponsored by the government.” United States v. Strouse, 286 F.3d 767, 772 (5th Cir. 2002). Accordingly, when a defendant claims that the prosecution put false information before the grand jury, we ask two questions (1) did the government “knowingly “sponsor[]” false information before the grand jury and (2) was that information material, that is, was the information “capable” of influencing the grand jury’s decision. See, e.g., id. at 770-72; Forte, 2003 WL 1922910, at *5-6.

Colorado challenges the following statements made to the grand jury as being materially false and misleading: (1) that the panel in Cessa I “affirmed except for one guy”; (2) that the panel in Cessa I found the evidence against Colorado sufficient; (3) that the expansion in the dates in the indictments was “based on some stuff’ the Government discovered between the first - and second indictments; (4) statements citing Agent Fernald’s financial analysis, including that ADT was not legitimate because it was begun with drug proceeds and that Colorado had “unexplained wealth.” For all but the first challenged statement, the Government did not knowingly put false information before the grand jury because it had an evidentiary basis for the statements. See United States v. Stearman, 981 F.2d 1256, 1992 WL 386839, at *3 (5th Cir. 1992) (unpublished) (finding no prosecutorial misconduct when the government witness had a factual basis for their grand jury testimony).

The first challenged statement was immaterial. In Cessa I, we did not “afflrm[ ] except for one guy”; we reversed both Colorado and Eusevio Huitron’s convictions, finding that the Government failed to carry its burden to show that Huitron, a horse trainer who worked with the Zetas, joined the money laundering conspiracy. Cessa I, 785 F.3d at 180. However, we fail to see how the fact that Huitron’s conviction was reversed was “capable” of influencing the grand jury’s decision to indict Colorado, especially because (1) Huitron and Colorado were charged with participating in the money laundering conspiracy in very different ways — Huitron as a horse trainer and Colorado as a major financial player and (2) the' prosecutor repeatedly told the grand jury that they had to independently consider the evidence, despite the prior verdicts.

Third, we reject Colorado’s objections to the prosecutor’s leading questions. Leading questions are permitted in grand jury proceedings, so long as “a duly-sworn witness actually testified] to the factual correctness of all the [leading] questions asked.” United States v. Brown, 872 F.2d 385, 388 (11th Cir. 1989); accord McKethan v. United States, 439 U.S. 936, 938, 99 S.Ct. 333, 58 L.Ed.2d 333 (1978) (“In grand jury proceedings, the ordinary rules of evidence do not apply. Leading questions and multiple hearsay are permitted and common.”); United States v. Weiss, 752 F.2d 777, 786 (2d Cir. 1985) (“We perceive no error in the prosecution’s use of leading questions before the grand jury.”); United States v. Galindo, No. 04-053 DAE, 2008 WL 918405, at *5 (D. Haw. Apr. 4, 2008) (“Defendant asserts that the excessive use of leading questions was improper. Leading questions alone are not grounds for dismissal of an indictment.”). And here, a sworn agent responded to each of the prosecutor’s leading questions. Accordingly, we find no error in the prosecutor’s use of leading questions before the grand jury.

VII

For the foregoing reasons, we remand Colorado’s Brady claim to the district court for further findings on the suppression and materiality elements. We do not reach Colorado’s forfeiture claim because it is contingent on a valid conviction. We otherwise reject Colorado’s arguments. Because this is a limited remand, “[w]e retain jurisdiction of this appeal pending the district court’s compliance with our limited remand.” Wheeler v. City of Columbus, 686 F.2d 1144, 1154 (5th Cir. 1982); accord Smitherman v. Bayview Loan Servicing, L.L.C., 675 Fed.Appx. 428, 429 (5th Cir.2017) (unpublished); In re Schlumberger Tech. Corp., 648 Fed.Appx. 420, 421 (5th Cir. 2016) (unpublished); Hornbeck Offshore Servs., LLC v. Salazar, No. 10-30585, 2010 WL 3219469, at * 2 (5th Cir. Aug. 16, 2010) (unpublished).

1 . Colorado's specific request for the Nayen interview memoranda followed the defense's more general pre-trial request for all Brady and Giglio material.

2 . True, Colorado’s substantive defense may have had factual similarities to a duress de fense. To show duress, a defendant must prove:

(1) that the defendant or a member of his family was under an unlawful and present, imminent, and impending threat of such a nature as to induce a well-grounded apprehension of death or serious body injury; (2) that he did not recklessly or negligently place himself in a situation in which it was probable that he would be forced to choose the criminal conduct; (3) that he had no reasonable legal alternative to violating the law, that is, he had no chance to refuse to do the criminal act and to avoid the threatened harm; and (4) that there was a direct causal relationship between the criminal action taken and the avoidance of the threatened harm.

United States v. Montes, 602 F.3d 381, 389 (5th Cir. 2010) (quoting United States v. Willis, 38 F.3d 170, 175 (5th Cir. 1994)). Facts that would support the threat element of duress likely overlap with the facts underlying Colorado’s explanation for why he gave gifts to the Zetas. But this overlap neither required Colorado to seek a duress instruction nor improperly shifted the burden of disproving duress onto the Government. See, e.g., Martin v. Ohio, 480 U.S. 228, 234, 107 S.Ct. 1098, 94 L.Ed.2d 267 (1987) ("It may be that most encounters in which self-defense is claimed arise suddenly and involve no prior plan or specific purpose to take life. In those cases, evidence offered to support the defense may negate a purposeful killing by'prior calculation and design ....”); United States v. Haischer, 780 F.3d 1277, 1282 (9th Cir. 2015) (holding that the district court erred by excluding evidence that the defendant had been coerced in a wire fraud case, even when the defendant did not request a duress instruction, because evidence of coercion supported the defendant's trial theory "that she signed the documents without knowing what they said because of pressure from [her abuser], such that she lacked the required knowledge and intent”).

3 . Of course, finding that evidence supports an affirmative defense and is therefore favorable at Brady step one does not mean that the evidence is material at Brady step three.

4 . Nayen was a key Government witness at the second trial — his testimony was a significant evidentiary addition in that trial. Nayen had been interviewed by three government agencies over a two year period. And yet, the defense had limited discovery material with which to challenge Nayen. And, on cross-examination, Nayen was able to repel lines of questioning by claiming that defense counsel’s premises were lies. Had the 302s been disclosed before trial, the defense may have been able to pursue additional lines of impeachment.

5 . It is not at all clear that the Government's arguments on the suppression prong would apply to all the alleged Brady and Giglio material. For example, many of the 302s indicate that notes are attached to the document, but the 302s in the record do not have attached notes.

6 . We recognize that "Brady violations not considered in district court are not properly before our court, and should not be considered.” Banks, 583 F.3d at 329. And accordingly, if we were to reach an ultimate conclusion on the Brady claim we could not consider the notes. However, because we find clear error in the district court’s favora-bility analysis, and because potential Brady issues surrounding the noted arose on appeal, the best course is to let the district court decide itself how to handle the notes on remand.

7 .We note that although the 302s list the individuals present at each interview, Lawson is not listed as present for any of the 302s.

8 . On appeal, Colorado also sought disclosure of any “documents reflecting interviews with Nayen after December 18, 2013.” Below, neither the defense motion seeking Brady material nor the district court’s order requiring that the 302s be produced was time limited. Nonetheless, the Government, claiming that no additional interview memoranda existed, produced no interview memoranda after December 18, 2013. Brady obligations are continuing throughout trial, and are neither dependent on a request from the defendant nor the form of the Brady material. See, e.g., United States v. Bagley, 473 U.S. 667, 681-83, 105 S.Ct. 3375, 87 L.Ed.2d 481 (1985) (no requirement that defendant requests Brady material); Sipe, 388 F.3d at 486 (evidence in inadmissible form may still be Brady material); Jackson v. Johnson, 194 F.3d 641, 649 n.18 (5th Cir. 1999) (Brady obligations continuing). In light of our remand, we do not reach the issue; instead, we leave it to the district court in the first instance to consider Colorado’s request for additional disclosures.

9 . To be clear, the intent to conceal or disguise element is an element of the substantive money laundering offense, not the charged conspiracy offense. See Threadgill, 172 F.3d at 366 (indicating that money laundering conspiracy does not require proof of the substantive elements of money laundering).

10 . Nova Scotia also makes clear that certain structural errors in the grand jury (such as the exclusion of women or racial minorities from the grand jury) may lead to the presumption of prejudice. 487 U.S. at 256-57, 108 S.Ct. 2369. Colorado makes no argument of structural error here.