Accordingly, the court affirmed Yu’s conviction and the denial of his equal protection-based motions to dismiss.

III. Detailed Analysis

A. Factual and Procedural Context

Understanding the legal analysis requires some grasp of the technological and procedural background.

• Technology and trade secrets. ADI develops high‑performance radio frequency and microwave amplifier microchips (MMICs) used in infrastructure and military applications. After acquiring Hittite Microwave, ADI engaged in a complex “translation” process to adapt Hittite’s designs for manufacture at a Taiwanese foundry, Win Semiconductor (“Win”). Each design went through a multi‑month iterative process (schematic, simulation, layout, prototype fabrication, testing, refinement), culminating in a GDS (layout) file that could be sent to Win.
• Yu’s conduct. While employed by ADI (after previously working at Hittite), Yu:
  • signed confidentiality agreements,
  • copied ADI GDS files—including the “K8600_TOP.gds” file containing the HMC1022A prototype layout—to his personal devices,
  • renamed some files to disguise them (e.g., “Kids8600.jpg”),
  • started Tricon MMIC, a competing company registered in his wife’s name,
  • sent GDS files to Win for manufacturing under the Tricon brand, and
  • marketed Tricon chips as equivalents or improvements to ADI/Hittite chips, including the HMC1022/HMC1022A lineage.
• Indictments and trial. Over several superseding indictments, Yu was eventually charged with 21 counts. After the government’s case and renewal of Rule 29 motions, the district court acquitted him on the unlawful procurement of citizenship count but sent the rest to the jury. The jury acquitted on 20 counts and convicted on only Count 1: possession of the HMC1022A design layout and GDS file as a stolen trade secret.
• Post‑trial equal protection motions. Yu had earlier moved to dismiss for selective prosecution and enforcement based on alleged anti‑Chinese bias. The district court:
  • denied the selective prosecution claim (no “clear evidence” under Armstrong),
  • allowed limited discovery on selective enforcement (using a preponderance standard), but
  • ultimately denied that claim as well, while expressing concern about “implicit bias” and broader anti‑Asian sentiment.

B. Precedents and Authorities Cited

1. Sufficiency and indictment interpretation

• United States v. Díaz‑Rosado, 857 F.3d 116 (1st Cir. 2017): cited for the standard of review on sufficiency challenges—viewing the evidence in the light most favorable to the prosecution and determining whether a rational factfinder could find guilt beyond a reasonable doubt.
• United States v. Brandao, 539 F.3d 44 (1st Cir. 2008), and United States v. Pierre, 484 F.3d 75 (1st Cir. 2007): define and explain “constructive amendment” of an indictment—when the charges, literally or in effect, are altered after grand jury review.
• United States v. Katana, 93 F.4th 521 (1st Cir. 2024), and United States v. Martínez, 994 F.3d 1 (1st Cir. 2021): used for the method of reading indictments “in a plain and commonsense manner,” focusing on text and context to determine the scope of the charged offense and to distinguish between constructive amendments and mere variances.

2. Trade secret law under the Economic Espionage Act

• 18 U.S.C. § 1832(a)(3): criminalizes possession of a trade secret, knowing it to have been stolen or appropriated without authorization, with intent to convert it to the economic benefit of someone other than the owner, and intending or knowing the offense will injure the owner.
• 18 U.S.C. § 1839(3)(B): defines “trade secret” in relevant part as information that “derives independent economic value, actual or potential, from not being generally known to, and not being readily ascertainable through proper means by, another person who can obtain economic value from the disclosure or use of the information.”
• BondPro Corp. v. Siemens Power Generation, Inc., 463 F.3d 702 (7th Cir. 2006): for the proposition that once information becomes public knowledge, it ceases to be a trade secret—used as a framing citation, though the First Circuit emphasizes that “public” does not automatically include information that can only be extracted by expensive, complex reverse engineering.
• United States v. Martin, 228 F.3d 1 (1st Cir. 2000): an earlier Economic Espionage Act case; the court used it to analogize how knowledge of the trade secret nature of information can be inferred from behavior, such as dealing with materials marked “confidential” or “proprietary,” seeking inside information, and referring to informants as “spies.”

3. Equal protection, selective prosecution, and selective enforcement

• United States v. Armstrong, 517 U.S. 456 (1996): the anchor precedent for selective prosecution, holding that:
  • prosecutorial decisions are entitled to a presumption of regularity,
  • a defendant must present “clear evidence to the contrary” to overcome that presumption, and
  • to show discriminatory effect in a race case, a claimant must show that “similarly situated individuals of a different race were not prosecuted.”
• Wayte v. United States, 470 U.S. 598 (1985): supplies the equal protection framework; a selective prosecution/enforcement claimant must show both discriminatory effect and discriminatory purpose, with purpose meaning that the decisionmaker selected or reaffirmed a course of action “because of, not merely in spite of, its adverse effects upon an identifiable group.”
• Personnel Administrator of Massachusetts v. Feeney, 442 U.S. 256 (1979): foundational on the meaning of discriminatory purpose; merely foreseeing disparate impact is insufficient—there must be a level of purposeful targeting or indifference to discriminatory consequences that rises to intent.
• United States v. Lewis, 517 F.3d 20 (1st Cir. 2008): provides a detailed discussion of what “similarly situated” means for selective prosecution claims (comparable crimes, circumstances, deterrent value, strength of evidence) and underscores the fact‑intensive nature of the inquiry.
• Flowers v. Fiore, 359 F.3d 24 (1st Cir. 2004): a selective enforcement case; confirms that equal protection requires both (1) differential treatment of similarly situated persons and (2) motivation “at least in part” by race.
• Whren v. United States, 517 U.S. 806 (1996): cited for the proposition that the Constitution prohibits racially selective enforcement, even when objective probable cause for a stop exists.
• Conley v. United States, 5 F.4th 781 (7th Cir. 2021): a key parallel authority. The First Circuit explicitly aligns with Conley in holding that selective enforcement claims, unlike selective prosecution claims, are governed by the preponderance of the evidence standard, not Armstrong’s higher “clear evidence” burden. Conley’s reasoning about the different institutional roles and immunities of prosecutors versus law enforcement is closely tracked.
• Farm Labor Organizing Committee v. Ohio State Highway Patrol, 308 F.3d 523 (6th Cir. 2002), and Chavez v. Illinois State Police, 251 F.3d 612 (7th Cir. 2001): used to state that discriminatory effect in selective enforcement can be demonstrated via specific comparators or relevant, well‑constructed statistical evidence showing racial disparities in similar enforcement decisions.
• United States v. McGlashan, 78 F.4th 1 (1st Cir. 2023), and United States v. Bucci, 582 F.3d 108 (1st Cir. 2009): supply the general framework for the standard of review of motions to dismiss indictments (legal questions de novo, factual findings for clear error), and for analogous vindictive prosecution claims (where similar mixed review applies).
• Imbler v. Pachtman, 424 U.S. 409 (1976), and Kalina v. Fletcher, 522 U.S. 118 (1997): used to highlight the doctrinal distinction in civil immunity between prosecutors (absolute immunity when acting in a core prosecutorial role) and law enforcement officers (qualified immunity), supporting the decision to impose a lower burden of proof on selective enforcement claims.

C. Legal Reasoning

1. What exactly did Count One charge? Prototype versus final product and the indictment’s scope

Yu’s first sufficiency argument turned on a close reading of Count One. The indictment charged him with possessing “[t]he design layout and GDS file for the HMC1022A microchip.” Yu argued that this must mean the final, market‑released version of the HMC1022A, which ADI did not release until February 2019—well after he left ADI. He maintained that he had only ever possessed an “abandoned prototype” created during the translation process, and that the government’s trial proof thus varied from—or constructively amended—the indictment.

The First Circuit rejected that argument by:

• Reading the indictment in context. Rather than parsing the phrase as if it were a statute, the court followed Katana and Martínez, looking at the indictment “in a plain and commonsense manner” across all its sections. The “General Allegations”:
  • described Yu’s theft of ADI design files beginning in 2016,
  • identified how those files were used to produce Tricon chips marketed as substitutes for ADI’s HMC1022, and
  • stated that when Yu’s home was searched in June 2019, agents found thousands of files “match[ing]—bit-for-bit—those developed, owned, and still maintained by ADI.”
• Linking Count One to the specific file Yu stole. The broader narrative made clear that Count One was pegged to the specific GDS design Yu had copied (“K8600_TOP.gds”, re‑labeled “Kids8600.jpg”), which contained the HMC1022A prototype layout he used for Tricon’s TM5051/TM5052 chips—not to the inevitably evolved, final production version ADI later sold.
• Avoiding constructive amendment or variance. Because the indictment, properly read, already covered the prototype file, there was no mismatch between the charge and the evidence. The court therefore did not need to decide whether Yu’s complaint sounded in “constructive amendment” or “variance,” nor to analyze any prejudice along those lines.

This approach underscores a practical principle in complex IP prosecutions: where an indictment’s narrative background clearly ties a charge to a specific internal design file and its role in alleged theft, courts are likely to read product identifiers (like “HMC1022A”) flexibly, in light of that context, rather than limit them to final, publicly released products.

2. Was the file a “trade secret” on the charged date?

Yu’s second and third sufficiency arguments went to the core statutory requirement that the information constitute a “trade secret” under 18 U.S.C. § 1839(3)(B). He argued:

1. that once ADI released the HMC1022A commercial product in February 2019, its design was “readily ascertainable” via reverse engineering; hence, by June 2019 (the charged possession date) there was no trade secret left, and
2. that the prototype he possessed never had independent economic value because it was only a “translation” of an older HMC1022 design that he says was itself readily ascertainable and comparable to competitor chips.

The court rejected both contentions, relying heavily on evidence about the real‑world difficulty and cost of reverse engineering MMIC designs.

a. Trade secret status on the charged date

Yu invoked decisions like BondPro for the idea that information in a sold product is no longer secret. The First Circuit accepted the general proposition but drew an important distinction: information does not become “generally known” or “readily ascertainable” merely because it is embodied in a product, if extracting that information requires substantial expenditure of time, skill, and money.

The government’s expert and fact testimony established that:

• MMICs involve highly complex layouts, where the economic value lies in the precise geometry (size, shape, and placement) of multiple layers of components.
• Even with powerful microscopes, lower layers remain hard to discern; reconstructing the full design requires specialized tools and software.
• Reverse engineering is possible but would take “several weeks to a month” of dedicated design work just to prepare schematics and layout, followed by testing and new “tapeouts” (prototype runs)—all involving significant cost.

Given this evidence, the court held a rational juror could find that:

• ADI’s HMC1022A design layout was not “readily ascertainable through proper means” within the meaning of § 1839(3)(B), even after product release, and
• the prototype design file Yu took therefore still qualified as a trade secret in mid‑June 2019.

This is a significant application of trade secret doctrine to modern semiconductor technology: it reinforces that high‑complexity designs can remain trade secrets even after commercialization, where reverse engineering is non‑trivial.

b. Trade secret status at any time: independent economic value

Yu also argued the prototype never had the required “independent economic value” because it was simply an early “translation” of a legacy chip whose design he claimed was already effectively known in the marketplace.

Again, the court took the evidence in the light most favorable to the government and concluded that a rational jury could find:

• The original HMC1022’s layout itself was not readily ascertainable for similar reverse‑engineering reasons.
• The translation design (the prototype in the charged file) was itself highly valuable:
  • ADI invested months of expert effort and tens of thousands of dollars per prototype run to translate and refine the design for Win’s process.
  • Yu used ADI’s prototype to short‑circuit that investment, generating Tricon designs (TM5051 and TM5052) quickly enough to beat ADI’s HMC1022A to market.
  • Yu told a customer that one Tricon chip’s performance was so strong that “you can hardly find anything in the market that matches” it—indicating that neither the legacy nor the translation was commodity or easily replicable technology.

From these facts, the court held the jury could reasonably infer that ADI’s design files, including the prototype that Yu took, derived independent economic value from their secrecy. Yu’s “short cut” was itself proof of that value.

3. Did § 1832 require proof that Yu knew the file was a “trade secret,” and was such knowledge shown?

Yu argued that:

• the government had to prove he knew that the file not only was confidential but legally met the definition of a “trade secret,” and
• ADI’s internal information classification policy—which put “trade secret” materials in a “secret need‑to‑know” category while placing schematics and layouts in a lower category with the note that disclosure would not be “critical” if it happened—could have led him to believe layout files were proprietary but not “trade secrets.”

The court sidestepped the legal question—whether § 1832 requires knowledge that the information meets the statutory definition—by assuming such knowledge were required and holding it was supported by the evidence regardless.

Key facts supporting a finding that Yu knew he was dealing with trade secrets included:

• He had signed confidentiality agreements explicitly treating product‑related information as confidential and restricting its use “for [his] own purposes.”
• He deliberately concealed his removal and use of ADI files:
  • renaming “K8600_TOP.gds” to “Kids8600.jpg” to disguise its nature as a GDS file,
  • saving files under names of Pokémon characters, and
  • keeping Tricon a “secret” from ADI while still employed.
• He asked Win to treat his Tricon design layouts as “proprietary data,” mirroring ADI’s own proprietary treatment.

Drawing on Martin, where similar behavior supported an inference of knowledge that confidential materials were trade secrets, the court held a rational juror could conclude that Yu understood the designs’ secret and economically sensitive nature, even if the internal ADI policy labels were more nuanced.

Doctrinally, the court left open—deliberately—whether the statute requires knowledge that the information is a “trade secret” in the technical, statutory sense, or merely that it is proprietary and secret. That question remains unsettled in the First Circuit, but Yu indicates that, as a practical matter, prosecutors can satisfy even a demanding knowledge requirement through circumstantial evidence of concealment and contractual undertakings.

4. Selective prosecution: reaffirming Armstrong’s high bar

a. Standard of review clarified

The First Circuit had not previously squarely articulated the standard of review for denials of selective prosecution claims. In Yu, the court aligned selective prosecution review with its general approach to motions to dismiss indictments:

• Legal questions: reviewed de novo.
• Factual findings (e.g., comparator similarity): reviewed for clear error.
• The “ultimate ruling”—here, denial of the motion to dismiss—flows from those legal and factual determinations.

This clarification is important for future litigants: appellate challenges to the factual comparability of proposed “similarly situated” individuals will face a deferential clear error standard.

b. Substantive standard: Armstrong and “clear evidence”

On the merits, the court applied Armstrong in full:

• Prosecutorial charging decisions are entitled to a presumption of regularity.
• To rebut this presumption, a defendant must present “clear evidence” that:
  • (1) Discriminatory effect: similarly situated individuals of a different race were not prosecuted, and
  • (2) Discriminatory purpose: the prosecution was initiated in part because of the defendant’s race.
• “Similarly situated” is fact‑heavy and must be assessed considering:
  • the nature and comparability of the crimes,
  • the manner in which they were committed,
  • the deterrent value of prosecution, and
  • the strength of the evidence against each potential defendant (Lewis factors).

c. Application to Yu’s comparators and arguments

Yu pointed to:

• civil trade secret cases—including one where ADI sued MACOM—arguing that non‑Chinese ex‑employees who took MMIC‑related designs faced only civil consequences; and
• statistics and research suggesting disproportionately high criminal prosecutions against people of Chinese descent in espionage‑ and trade‑secret cases.

The court, echoing the district judge, found these comparators insufficient:

• MACOM case. Yu did not show that:
  • the former ADI employees there had actually completed theft and use of trade secrets in a way comparable to Yu’s alleged conversion and commercialization through Tricon, or
  • there were analogous allegations of export‑control violations or transfer of sensitive designs to foreign entities.
  The quick civil settlement and lack of apparent criminal referral or clear evidence of completed misappropriation undermined its comparability.
• Other civil cases. Yu did not show that:
  • the government knew about those specific defendants’ conduct at the time, and
  • the conduct involved the same kind of sensitive, potentially military‑related MMIC technology or foreign connections alleged in Yu’s case.
• Statistics and general patterns. While potentially relevant to proving discriminatory purpose in some contexts, the court viewed them as insufficient, without case‑specific comparators, to satisfy Armstrong’s clear‑evidence threshold on discriminatory effect.

Crucially, the court emphasized the particular combination of allegations against Yu: not just trade secret theft, but exports of controlled technology and a nexus to a foreign manufacturing entity (Win in Taiwan) in a national security‑sensitive industry. The absence of comparably situated non‑Chinese individuals who were spared criminal prosecution—backed by clear evidence—was fatal to his claim.

Because Yu failed on discriminatory effect, the court did not have to analyze in depth the evidence he cited of discriminatory purpose—including Trump Administration rhetoric, the timing of the case acceptance near tariff announcements, DOJ communications emphasizing his “Chinese‑born” status, and a problematic attempted peremptory strike of an Asian‑American juror. As Armstrong requires both prongs, failure on the effect prong ended the inquiry.

5. Selective enforcement: a new evidentiary standard in the First Circuit

a. Distinguishing law enforcement from prosecutors

The most significant doctrinal development in Yu is the court’s holding on the evidentiary standard for selective enforcement claims.

The First Circuit drew a clear line between:

• Selective prosecution (charging decisions by U.S. Attorneys, governed by Armstrong’s “clear evidence” standard and a presumption of regularity deeply rooted in separation‑of‑powers concerns), and
• Selective enforcement (investigative choices by law enforcement officers, including whether to open or pursue an investigation or to refer a matter for prosecution).

While acknowledging that law enforcement officers are also part of the Executive Branch and exercise discretion, the court emphasized two critical differences:

• Greater judicial oversight. Law enforcement officers:
  • regularly testify in court and must justify their investigative methods (e.g., in suppression hearings), and
  • are subject to civil suits and judicial scrutiny for alleged constitutional violations (e.g., Fourth and Fourteenth Amendment claims).
• Differing immunities.
  • Prosecutors enjoy absolute immunity for traditional prosecutorial acts (Imbler).
  • Law enforcement officers (and even prosecutors when acting in an investigative role) receive only qualified immunity (Kalina), exposing them more directly to judicial review and liability.

On this institutional backdrop, the court agreed with the Seventh Circuit’s reasoning in Conley and held that the Armstrong “clear evidence” standard does not extend to selective enforcement claims.

b. The new rule: preponderance of the evidence for selective enforcement

The court announced a clear rule:

Because law enforcement officers are subject to greater judicial scrutiny, we conclude that the reasoning laid out in Armstrong, requiring "clear evidence" that a prosecutor has violated equal protection, does not extend to selective enforcement claims. We agree with the district court and the Seventh Circuit that selective enforcement claims must be proven only by a preponderance of the evidence.

This is now binding precedent in the First Circuit. In practical terms:

• Defendants alleging selective enforcement based on race do not need to meet the “clear evidence” threshold of Armstrong.
• They must still prove:
  • (1) Discriminatory effect (they were treated differently than similarly situated individuals of a different race), and
  • (2) Discriminatory purpose (race was at least one motivating factor—“because of,” not “in spite of,” the disparate impact).
• But they need only do so by a preponderance of the evidence—showing it is more likely than not that these elements are satisfied.

The court also noted, as did Conley, that some other circuits have cited Armstrong in the selective enforcement context “in passing” without squarely rejecting a preponderance standard, so the decision does not necessarily create a sharp circuit split.

c. Application to Yu’s selective enforcement claim

The district court had allowed limited discovery on selective enforcement and, after a hearing, denied the motion to dismiss. It expressed concern about:

• “implicit bias” and a broader rise in anti‑Asian racism, and
• the role that Yu’s Chinese origin “played in part” in how events unfolded, while nonetheless concluding that law enforcement would likely have investigated someone engaging in Yu’s conduct even if the person had not been from China or of Chinese descent.

On appeal, Yu argued that the district court effectively found discriminatory motive but misapplied the law. The First Circuit disagreed, focusing on discriminatory effect.

• Comparator problem. As with his selective prosecution claim, Yu’s main comparator was the ADI–MACOM civil dispute. But for selective enforcement, he needed to show:
  • law enforcement knew of similarly situated non‑Chinese individuals engaged in comparable conduct (e.g., theft of sensitive MMIC trade secrets with foreign connections and export implications), and
  • nonetheless chose not to open an investigation.
  Yu did not show that law enforcement was actually aware of MACOM or other similar cases, let alone that agents declined to investigate them. His argument that they “could have, and should have, known” was insufficient under the fact‑intensive “similarly situated” standard.
• Facts supporting investigation. The court approved the district court’s reliance on the specific triggering factors in Yu’s case:
  • tips alerting authorities to suspected theft and export of MMIC designs with potential military uses,
  • the use of foreign foundries (Win in Taiwan),
  • Yu’s professional and personal ties involving China (including efforts to market Tricon in China and statements about returning to serve his country), and
  • the broader, legitimate concern with Chinese‑linked economic espionage.
  Given these facts, the court saw no clear error in the finding that law enforcement would have investigated a non‑Chinese engineer in the same position.

Because Yu failed to show that “one class [was] being treated differently from another class that [was] otherwise similarly situated,” the selective enforcement claim failed on discriminatory effect, and the court again did not need to definitively resolve the discriminatory purpose question.

D. Impact of the Decision

1. Trade secret and IP‑related criminal prosecutions

On the trade secret side, Yu has several practical implications:

• Commercialization does not automatically destroy trade secret protection. Particularly in complex technologies like MMICs, the fact that a product is sold into the market does not, by itself, render its design “generally known” or “readily ascertainable.” The cost, time, and sophistication required for reverse engineering are central to the analysis.
• Prototype and intermediate designs are protectable. Translation and prototype layouts—intermediate steps in a product’s evolution—can be trade secrets if they shorten development cycles or provide a “leg up” to a competitor, even if they do not perfectly match the final commercial product.
• Behavioral evidence of secrecy awareness is powerful. Emails, file‑renaming, concealment from an employer, and confidentiality agreements are all fertile ground for proving both the existence of a trade secret and the defendant’s knowledge of its sensitive status.
• Indictment drafting in technical cases can withstand imprecision. Provided the narrative sections of an indictment clarify the specific files and conduct at issue, courts are willing to interpret shorthand references to product names flexibly.

2. Equal protection and discriminatory enforcement in the First Circuit

On the equal protection side, Yu is a pivotal case in the First Circuit:

• Selective prosecution remains extremely hard to prove. Armstrong’s “clear evidence” requirement, combined with a stringent view of what counts as a “similarly situated” comparator, continues to make selective prosecution challenges to federal charging decisions rarely successful.
• Selective enforcement is doctrinally distinct and somewhat more accessible. By adopting a preponderance standard for selective enforcement claims, the First Circuit has opened, at least theoretically, more room for successful challenges to racially biased investigative practices—whether in complex white-collar matters, policing, or national security contexts.
• Comparator requirements remain demanding. Even under the lower evidentiary standard, plaintiffs and defendants must:
  • identify specific non‑minority individuals or robust statistical patterns where similarly situated persons were not investigated, and
  • tie those to the decisionmaker’s awareness and choices.
  The failure of Yu’s comparators underscores that the difference in the standard of proof does not relax the need for concrete, case‑specific evidence of differential treatment.
• Implicit bias concerns are acknowledged but not yet doctrinally decisive. The district court’s frank acknowledgment of “implicit bias” and anti‑Asian sentiment did not, standing alone, suffice to establish discriminatory effect or purpose. Yu reflects the modern tension between social science recognition of implicit bias and the legal system’s still high threshold for actionable discrimination.

In the particular geopolitical context of U.S.–China relations and DOJ’s historical “China Initiative,” the case signals that:

• Courts will allow the Executive Branch substantial leeway to prioritize investigations and prosecutions of Chinese‑linked economic espionage, especially where there is a plausible national security dimension.
• However, explicit or implicit race‑based motives remain constitutionally suspect, and future litigants now know that at least the standard of proof for selective enforcement is not as steep as for selective prosecution.

IV. Complex Concepts Simplified

1. Trade secret and “readily ascertainable”

A trade secret under federal law (18 U.S.C. § 1839) is:

• information (like a formula, design, or process),
• that has economic value because it is not generally known and not easily discoverable by fair means, and
• that the owner takes reasonable steps to keep secret.

“Readily ascertainable” does not mean “possible to figure out.” It means “reasonably easy to figure out.” If a competitor would have to spend weeks or months, hire specialized engineers, and use expensive tools to reconstruct the information from a product, courts typically treat the information as not readily ascertainable.

2. Economic Espionage Act (18 U.S.C. § 1832)

The Economic Espionage Act criminalizes various forms of trade secret theft. Section 1832(a)(3), at issue in Yu, focuses on possession of a trade secret, knowing it was stolen or wrongfully taken, with intent to convert it to someone else’s economic benefit and to harm the owner.

3. Constructive amendment vs. variance

• A constructive amendment occurs when, through the court’s instructions or the government’s proof, the terms of the indictment are effectively changed so the defendant might be convicted of something different from what the grand jury charged. This is generally reversible error.
• A variance means the government proved the same crime described in the indictment but did so using facts that differ from those alleged. Variances are only reversible if they prejudice the defendant’s substantial rights (e.g., by undermining notice or risking double jeopardy).

In Yu, the court found neither: the indictment, when read as a whole, already encompassed the prototype GDS file the government proved Yu possessed.

4. Selective prosecution vs. selective enforcement

• Selective prosecution challenges the decision to bring charges at all. It is aimed at the U.S. Attorney’s Office and is subject to Armstrong’s “clear evidence” standard.
• Selective enforcement targets law enforcement behavior before charging, such as decisions about whom to investigate or refer for prosecution. It can involve police, FBI agents, or other investigators.

Both are based on equal protection principles and require proof that:

• people of a different race, but similarly situated, were treated more leniently (no investigation or no charges), and
• the difference was motivated at least in part by race.

After Yu, in the First Circuit:

• Selective prosecution must be proven by clear evidence, and
• Selective enforcement must be proven by a preponderance of the evidence.

5. Standards of review on appeal

• De novo review: the appellate court decides the legal question fresh, with no deference to the district court (e.g., what evidentiary standard applies, whether a legal test is correct).
• Clear error review: the appellate court defers to the district court’s factual findings unless it is firmly convinced a mistake has been made (e.g., whether comparators are similarly situated in fact).

In Yu:

• Legal aspects of selective prosecution and enforcement were reviewed de novo.
• Fact‑finding—such as whether comparators were truly similar and whether agents would have investigated a non‑Chinese engineer—was reviewed for clear error.

6. “Clear evidence” vs. “preponderance”

• Preponderance of the evidence means that it is more likely than not (over 50%) that the claim is true. This is the standard for most civil cases and, now, for selective enforcement claims in the First Circuit.
• “Clear evidence” is tougher to quantify, but it means substantially more than preponderance. In practice, it means strong, convincing proof that overcomes the presumption that prosecutors acted properly—something closer to what many courts call “clear and convincing evidence,” though Armstrong does not use that exact phrase.

Under Yu, defendants alleging selective enforcement get the benefit of the lower preponderance standard, but still must marshal concrete comparators or robust statistics tightly tied to their circumstances.

V. Conclusion: Key Takeaways

United States v. Yu is significant on two fronts.

On the substantive criminal law and trade secret side, the decision:

• upholds the conviction of a single count of trade secret possession in a case where the government’s broader China‑linked narrative largely did not persuade the jury,
• clarifies that prototypes and translation layouts for sophisticated microchips can remain trade secrets even after related products hit the market, so long as reverse engineering remains costly and complex, and
• demonstrates that knowledge of trade secret status can be inferred from confidentiality agreements and efforts to conceal or protect design files.

On the equal protection and criminal procedure side, the case:

• reinforces the formidable Armstrong barrier to selective prosecution claims,
• articulates, for the first time in the First Circuit, the standard of review for selective prosecution and selective enforcement denials (legal issues de novo, factual findings for clear error),
• and, most notably, adopts a preponderance-of-the-evidence standard for selective enforcement claims, explicitly declining to extend Armstrong’s “clear evidence” requirement beyond the prosecutorial sphere.

At the same time, Yu shows that even under a lower evidentiary standard, selective enforcement claims remain difficult to win without solid comparator or statistical evidence demonstrating that law enforcement knowingly treated similarly situated non‑minority individuals differently. The court acknowledges concerns about implicit bias and anti‑Asian sentiment but insists that actionable equal protection violations turn on proof of concrete differential treatment and purpose.

For practitioners, policymakers, and scholars, Yu thus stands as an important First Circuit precedent:

• on how courts will handle high‑technology trade secrets in criminal prosecutions, and
• on how equal protection doctrine operates—both its limits and its possibilities—in the charged context of national security‑motivated investigations involving race and national origin.