United States v. Yu: The Preponderance Standard for Selective Enforcement and Trade Secret Protection After Product Release

I. Introduction

The First Circuit’s decision in United States v. Yu is significant on two distinct doctrinal fronts. First, it clarifies how criminal trade secret law under the Economic Espionage Act applies to pre‑release prototype designs and to information that may be technically reverse‑engineerable from commercially available products. Second—and more importantly for constitutional and criminal procedure purposes—it squarely adopts a preponderance of the evidence standard for selective enforcement claims, rejecting the extension of Armstrong’s demanding “clear evidence” standard from selective prosecution to law-enforcement investigations.

The case arises from the conviction of Haoyang Yu, a former engineer at Analog Devices, Inc. (“ADI”), who helped translate microwave microchip designs (MMICs) from a legacy Hittite process to fabrication at Win Semiconductor in Taiwan. Yu copied ADI’s internal GDS design files, left to start a competing company (Tricon MMIC LLC), and sent GDS files to Win for manufacturing. A grand jury returned a 21‑count indictment spanning trade secret offenses, wire fraud, export control violations, visa fraud, and unlawful procurement of citizenship. After a 15‑day trial, the jury convicted Yu on only one count—unlawful possession of a trade secret under 18 U.S.C. § 1832(a)(3) relating to the design layout and GDS file for the HMC1022A microchip—and acquitted him on all other counts.

On appeal, Yu challenged:

• the sufficiency of the evidence as to multiple elements of the trade secret offense; and
• the constitutionality of his investigation and prosecution, contending that as a person of Chinese ethnicity he was subject to selective enforcement and selective prosecution in violation of the Fifth Amendment’s equal protection component.

The First Circuit (Judge Montecalvo, joined by Judges Lipez and Aframe) affirmed. In doing so, it issued a careful, structured opinion with meaningful doctrinal innovations and clarifications that will shape future trade secret prosecutions and equal protection litigation.

II. Summary of the Opinion

A. Core Holdings

The court’s principal holdings can be distilled as follows:

1. Sufficiency of the evidence. Viewing the evidence in the light most favorable to the prosecution:
   • The indictment’s reference to “the design layout and GDS file for the HMC1022A microchip” properly encompassed the prototype translation file Yu had copied from ADI, not just the final commercial version. There was no constructive amendment or fatal variance.
   • The charged file satisfied the statutory definition of a trade secret (18 U.S.C. § 1839(3)(B)):
     • On the charged date (June 14, 2019), despite the public release of the HMC1022A chip in February 2019, because reverse engineering of the design layout from a finished chip was not “readily ascertainable” and required significant cost, time, and expertise; and
     • At all relevant times, because the prototype translation design conferred independent economic value as a shortcut to producing competitive MMICs and allowed Yu to beat ADI to market.
   • The evidence was sufficient for a rational juror to infer that Yu knew the file was a trade secret, whether or not such knowledge is formally an element of § 1832—an issue the court explicitly declined to resolve.
2. Selective prosecution.
   • The court clarified the standard of reviewde novo, factual findings for clear error.
   • Applying United States v. Armstrong, 517 U.S. 456 (1996), and First Circuit precedent in United States v. Lewis, the court held that Yu failed to produce the required “clear evidence” that:
     • similarly situated non‑Chinese individuals were not prosecuted; and
     • his prosecution was motivated by discriminatory purpose.
   • The civil trade secret cases and other alleged comparators Yu cited (including a civil suit by ADI against MACOM) were not shown to be similarly situated in the legally relevant sense, especially regarding export control aspects and foreign transfer of secrets.
3. Selective enforcement.
   • The First Circuit explicitly held for the first time that:

     Selective enforcement claims are not governed by Armstrong’s “clear evidence” standard and may be proved by a fair preponderance of the evidence.

   • The court adopted the same standard of review as with selective prosecution: legal issues de novo, factual determinations for clear error.
   • On the facts, the court agreed with the district court that Yu did not carry even this lower burden:
     • He failed to show that any non‑Chinese comparators who were truly similarly situated—possessing sensitive MMIC designs with potential military use, linked to foreign manufacturing, and known to investigators—were not investigated.
     • Given the tips received, the sensitive technology, the foreign manufacturing in Taiwan, and Yu’s business efforts in China, investigators would likely have opened and pursued the same investigation even if Yu had not been of Chinese ethnicity.

B. Doctrinal Significance

The opinion thus:

• cements a two‑track evidentiary regime:
  • Selective prosecution (charging decisions by prosecutors) – Armstrong “clear evidence” standard remains;
  • Selective enforcement (investigative decisions by law enforcement) – only a preponderance of the evidence is required.
• clarifies how courts identify “similarly situated” comparators and what kind of showing is required to rebut the presumption of regularity in charging decisions;
• articulates a robust view of trade secret protection for high‑technology design files even after public release of the physical product; and
• acknowledges, but ultimately does not give dispositive effect to, concerns about implicit bias and the China‑related national security context in which the investigation arose.

III. Factual and Procedural Background

A. ADI, MMIC Designs, and the HMC1022A Project

Analog Devices, Inc. designs monolithic microwave integrated circuits (MMICs)—tiny radio frequency and microwave amplifier chips used in infrastructure, aerospace, defense, and other high‑reliability applications. In 2014, ADI acquired Hittite Microwave Corporation, inheriting Hittite’s MMIC portfolio. Some Hittite designs, including the legacy HMC1022, were manufactured at foundries other than Win Semiconductor in Taiwan.

Each foundry uses its own specialized manufacturing processes, so ADI embarked on a painstaking translation process to port Hittite designs to Win’s manufacturing rules. Translation was not mere copying: the end product had to match the performance of the original part “exactly,” neither better nor worse, and required the same iterative design pipeline used for new chips:

• Schematics (modeled by thousands of mathematical formulae);
• Simulations to refine performance;
• Physical layout design (the blueprint for fabrication);
• Creation of a GDS file (the foundry‑ready design file);
• Fabrication of prototypes at Win (costing $35,000–$50,000 and ~6 weeks per batch);
• Extensive testing and further refinement through additional “spins.”

ADI’s convention for translated parts used the prefix “HMC,” the original Hittite number, and an “A” suffix—thus, the translation project for the HMC1022 created the HMC1022A.

B. Yu’s Role and Misappropriation

Yu, originally a consumer‑chip designer, joined Hittite shortly before the ADI acquisition and then transitioned to ADI, where he learned to design industrial MMICs and participated in translating Hittite designs for Win.

He signed broad confidentiality agreements promising that all information about ADI’s past, present, and potential products was confidential, that he would not use such information for his own purposes during or after employment, and that he would return confidential materials when he left.

Around mid‑2016, Yu began copying ADI’s GDS and related design files to his personal computer, sometimes disguising them by renaming (e.g., ADI’s “K8600_TOP.gds” – containing the HMC1022A prototype – renamed as “Kids8600.jpg,” which could only be used if its extension was changed back to “.gds”).

C. Formation of Tricon and Use of ADI Designs

By early 2017, Yu was:

• contacting Win about whether it would fabricate designs for a small business; and
• reaching out to potential customers about a new MMIC design company.

In March 2017, a new company, Tricon MMIC, LLC, was registered under the name of Yu’s wife. Tricon’s marketing materials explicitly pitched its parts as substitutes for ADI/Hittite parts, including the HMC1022. In June 2017, within just two days of receiving Win’s “process design kit” (which designers typically spend weeks or months mastering), Yu sent a GDS file to Win asking that it be treated as “proprietary data.”

Win personnel perceived Tricon’s first‑run chips as “finished goods” rather than early iterations, suggesting that the usual trial‑and‑error development process had been bypassed. In July 2018, Yu sent another GDS file to Win containing designs for, among others, TM5051 and TM5052, advertised on Tricon’s website as equivalent to the “ADI (Hittite) HMC1022 with more bandwidth.” Yu received those parts in September 2018. ADI released its own HMC1022A microchip to the market on February 15, 2019.

D. Investigation, Indictments, Trial, and Sentence

In June 2019, a grand jury indicted Yu and Tricon on 15 counts (later expanded to 21 in a third superseding indictment), covering:

• trade secret theft, copying, and possession (18 U.S.C. § 1832);
• wire fraud;
• transportation of stolen goods;
• illegal export of controlled technology to Taiwan;
• visa fraud; and
• unlawful procurement of citizenship.

Federal agents executed a search warrant on Yu’s home on June 14, 2019, and discovered over 2,000 ADI files that “matched—bit‑for‑bit—those developed, owned, and still maintained by ADI.”

Yu moved pretrial to dismiss for unconstitutional selective enforcement and prosecution, claiming he was targeted and over‑charged because he is ethnically Chinese. The district court (Judge William G. Young) deferred ruling on the motion until after trial, then:

• denied the selective prosecution component; but
• allowed limited discovery on selective enforcement, ultimately rejecting that claim as well after additional fact‑finding.

At trial, after partial Rule 29 acquittal (on unlawful procurement of citizenship), the jury acquitted Yu and Tricon on all but one count, convicting Yu only of possession of a stolen trade secret – “the design layout and GDS file for the HMC1022A microchip” – based on his possession of that file on approximately June 14, 2019.

Yu was sentenced to six months’ imprisonment, 36 months of supervised release, a $55,000 fine, and restitution of nearly $200,000 representing ADI’s legal fees. He has since completed his prison term and satisfied restitution.

IV. Detailed Analysis of the Court’s Reasoning

A. Sufficiency of the Evidence on the Trade Secret Count

1. Identifying the Charged File: No Constructive Amendment or Fatal Variance

Yu’s first sufficiency argument attacked Count One’s description of the trade secret as “[t]he design layout and GDS file for the HMC1022A microchip.” He contended that:

• this language necessarily referred to ADI’s final, commercial HMC1022A design as released in February 2019; and
• because he possessed only an earlier, abandoned prototype translation design, the government’s proof at trial did not match the indictment and thus either:
  • constructively amended the indictment (unconstitutional), or
  • created a prejudicial variance between indictment and proof.

The court rejected this argument by focusing on how indictments must be read: “in a plain and commonsense manner”, not parsed like statutes. Citing United States v. Katana and United States v. Martínez, the court emphasized that one must read Count One in the context of the indictment’s extensive “General Allegations,” which:

• describe Yu’s unauthorized copying of ADI files (including GDS files) to his personal devices beginning in 2016;
• link those files directly to the Tricon parts TM5051 and TM5052 marketed as equivalents to the HMC1022; and
• detail that when agents searched Yu’s house in June 2019, they found thousands of ADI files that matched ADI’s originals “bit‑for‑bit.”

In that context, the reference to “the design layout and GDS file for the HMC1022A microchip” is best read as a functional label identifying the stolen ADI translation design file for what ultimately became HMC1022A—not narrowly as the final, post‑release version actually sold by ADI.

The court underscored common sense: Yu could not have possessed a final design file before it existed, yet the indictment clearly tied Count One to files he copied while at ADI and possessed when his home was searched. The only plausible reading of the indictment, in light of its narrative, is that Count One charged Yu with possessing the prototype translation design that ADI used in developing HMC1022A.

Once the indictment was so construed, Yu’s own concession—namely, that he possessed “a design layout and GDS file for an earlier, abandoned prototype” used in the translation process—effectively confirmed that the government’s proof matched the charge. There was therefore no constructive amendment or variance problem to ground a sufficiency challenge.

2. Trade Secret Status on the Charged Date: Product Release and Reverse Engineering

The statutory definition of “trade secret” under 18 U.S.C. § 1839(3)(B) requires, among other things, that:

the information 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.

Yu argued that, by the time of the charged possession (June 14, 2019), HMC1022A had already been released commercially (February 2019), and therefore:

• its design features were “readily ascertainable” through lawful reverse engineering; and
• the design information, including the layout, no longer enjoyed trade secret protection, citing BondPro Corp. v. Siemens Power Generation, Inc. for the proposition that once information becomes public, it ceases to be a trade secret.

The government countered with detailed expert and fact testimony about the practical difficulty, expense, and time required to reverse engineer a complex MMIC layout from a physical chip:

• Although high‑powered microscopes allow imaging of chip structures, lower layers and complex interconnects remain difficult to discern even at high magnification.
• “Counting pixels” from micrographs to deduce dimensions and placements is neither trivial nor quick; reconstructing an accurate schematic and layout can take weeks to a month of concentrated work, followed by additional time for testing and additional “tapeouts.”
• The process demands sophisticated tools, specialized software, and highly skilled engineers—implying significant cost and resource investment.

Against this evidentiary backdrop, the court refused to accept Yu’s characterization of reverse engineering as “easy” or routine for MMICs. Instead, applying the standard sufficiency test (whether a rational juror, viewing the evidence in the light most favorable to the government, could find guilt beyond a reasonable doubt), the court held that a rational juror could find the design file was not “readily ascertainable” from the finished product.

This aspect of the opinion is doctrinally important. It confirms that:

• The mere possibility of reverse engineering does not automatically defeat trade secret status; the statutory focus is on whether the information is “readily” ascertainable through proper means.
• Where reverse engineering requires substantial time, specialized skills, and expensive tools, trade secret protection can survive after product release.

Thus, even though ADI had marketed the HMC1022A chip by February 2019, the underlying proprietary design layout and GDS file retained trade secret status by June 2019 as long as the jury could conclude that reconstructing the design from the chip was not “readily” achievable.

3. Trade Secret Status at Any Time: Prototype Translation as a Valuable Secret

Yu’s third sufficiency argument was broader: he asserted that the ADI prototype file he took never qualified as a trade secret at all. Because the legacy HMC1022 had been on the market “for years” and, in his view, resembled other available competitor chips, he contended:

• the translation prototype for HMC1022A was merely an “early attempt” to replicate an already public design;
• the original HMC1022’s features were themselves readily ascertainable; and therefore
• the translation prototype did not derive independent economic value from being secret—it was simply an engineering artifact built on public information.

The government again contested both the factual premise and the legal conclusion:

• Factually, there was evidence that HMC1022’s performance was not easily matched in the market. Yu himself had written to a customer in November 2018 that “you can hardly find anything in the market that matches [the TM5051’s] performance,” suggesting that the competitive niche HMC1022 (and its translated counterpart) occupied was nontrivial.
• Legally and economically, the translation file was a substantial shortcut: it captured months of high‑value engineering work, allowing Yu to leapfrog the costly, iterative design and testing cycle and “beat ADI’s HMC1022A to market.”

The court, again deferring to the jury’s role, held that a rational juror could find that the ADI translation designs:

• represented proprietary know‑how and engineering optimization not readily duplicated from the legacy part alone; and
• consequently, “derive[d] independent economic value” from not being known or readily ascertainable to competitors.

The opinion thus underscores that prototype or even “abandoned” designs can be trade secrets if they provide economic value by compressing the R&D timeline or offering a head start in competitive markets. The subject matter need not be the exact version ultimately commercialized.

4. The Knowledge Element Under § 1832: Left Open but Easily Satisfied

Yu’s final sufficiency argument concerned mens rea. The district court had instructed the jury that the government must prove “not only that the information was in fact a trade secret, but that Mr. Yu knew it was a trade secret.” Yu embraced that instruction and argued on appeal that:

• because reverse engineering of MMICs (including competitors’ chips) was common industry practice and part of his job at ADI, he reasonably believed such designs were “readily ascertainable”; and
• at most, he knew he was violating internal confidentiality agreements, not that he was dealing with “trade secrets” in the statutory sense.

He also relied on ADI’s internal “Information Classification Policy,” which categorized “trade secret information” as “[s]ecret [n]eed to [k]now,” whereas schematics and layouts allegedly fell into a less restrictive category with the guidance: “external access to this data should be prevented, but should this data become public, the consequences are not critical.” From this, Yu argued he reasonably could have believed layout files were proprietary but not trade secrets.

The First Circuit took a two‑step approach:

1. It expressly declined to decide whether § 1832(a)(3) requires that a defendant subjectively know the information meets every element of the statutory “trade secret” definition, i.e., whether “knowledge of trade secret status” is an element as opposed to knowing the information is confidential/proprietary and misappropriated.
2. It held that, even if such knowledge is required, the evidence was sufficient to allow the jury to infer Yu’s knowledge from circumstantial evidence, including:
   • his signed confidentiality agreements promising not to use ADI product information for his own purposes and to return it upon departure;
   • his deliberate concealment of the copied files (e.g., renaming GDS files to appear as innocuous JPEGs like “Kids8600.jpg”);
   • his efforts to keep Tricon “a secret” from ADI, as he requested of a business contact; and
   • his affirmative instruction to Win to treat his own GDS files as “proprietary.”

Drawing an analogy to United States v. Martin, a prior First Circuit conspiracy‑to‑steal‑trade‑secrets case, the court reasoned that such conduct demonstrates consciousness of dealing with confidential, competitively sensitive information, sufficient for a reasonable juror to find that Yu understood the charged file’s trade secret character in the way relevant to criminal liability.

By sidestepping a definitive ruling on the legal knowledge requirement while affirming on evidentiary sufficiency, the court leaves the doctrinal question open for future cases but signals that, in practice, proving “knowledge” will often be satisfied by robust circumstantial evidence of secrecy and concealment, especially where the defendant is a sophisticated insider bound by comprehensive NDAs.

B. Equal Protection and Selective Prosecution

1. Standard of Review Clarified

The First Circuit had previously addressed the standard of review for denial of discovery on selective prosecution claims, but not clearly for denials of motions to dismiss indictments. Surveying sister circuits, it adopted the now‑standard approach:

• Legal questions (what the constitutional standard is, how Armstrong applies) are reviewed de novo.
• Factual findings (especially regarding whether comparators are “similarly situated”) are reviewed for clear error.

This aligns selective prosecution review with broader First Circuit practice on Rule 12 motions to dismiss and vindictive prosecution claims.

2. The Armstrong Framework and “Similarly Situated” Comparators

Reaffirming Armstrong and Wayte, the court restated the stringent requirements for a successful selective prosecution claim:

• There is a presumption of regularity in prosecutorial decisions, tethered to the Executive’s Article II duty to “take Care that the Laws be faithfully executed.”
• A defendant must produce “clear evidence” rebutting that presumption, showing:
  1. Discriminatory effect – in a race case, that “similarly situated individuals of a different race were not prosecuted.”
  2. Discriminatory purpose – that the decisionmaker selected or persisted in a particular course of action “because of, not merely in spite of,” its adverse effects on a protected group (quoting Feeney and Wayte).

Crucially, drawing on its prior decision in Lewis, the court again stressed that defining “similarly situated” is an intensely fact‑specific inquiry in which “no fact should be omitted”:

• Comparators must have committed roughly the same crime under roughly the same circumstances.
• Material differences may include:
  • the nature and gravity of the alleged offenses;
  • the manner in which the crimes were committed (e.g., foreign transfers, export control violations);
  • the deterrent value of a criminal prosecution in each scenario; and
  • the strength of the evidence available against each potential defendant.

3. Application: Why Yu’s Selective Prosecution Claim Fails

Yu’s selective prosecution argument hinged on several types of evidence:

• a list of state and federal civil trade-secret cases involving non‑Chinese defendants who were not criminally charged;
• the ADI v. MACOM civil case, where ADI alleged that four non‑Chinese ex‑employees misappropriated similar MMIC trade secrets but ADI pursued only civil injunctive relief against their new employer;
• alleged unlicensed exports of GDS files by ADI and Custom MMIC that were not criminally charged; and
• statistics and studies allegedly showing disproportionate criminal charging of ethnically Chinese defendants in espionage‑style cases.

The district court and First Circuit agreed that this showing fell short of Armstrong’s “clear evidence” threshold for discriminatory effect.

a. The MACOM civil case and similar matters

Yu presented the ADI v. MACOM lawsuit as a paradigmatic comparator: ex‑ADI employees allegedly took some of the same categories of information (MMIC GDS and design files) to a competitor, yet were spared criminal charges. The First Circuit, however, emphasized key differences:

• The record did not show that the allegedly misappropriated information in MACOM’s case actually reached the new employer—the employees were caught “in the attempt.”
• The quick settlement between ADI and MACOM suggested the possibility of weak or limited evidence of criminally actionable behavior, quite unlike Yu, who:
  • actually used ADI files to fabricate and sell chips competing with ADI’s MMICs; and
  • was also alleged to have engaged in export control violations and transfer to foreign entities.

Absent clear evidence that MACOM’s ex‑employees were in comparable circumstances (including similar involvement in foreign manufacturing and national‑security‑sensitive technology) and that prosecutors were actually aware of and chose not to prosecute them, the MACOM case could not serve as a valid comparator under Armstrong.

b. Alleged unlicensed exports and other civil cases

Yu also alleged that ADI and Custom MMIC exported GDS files for controlled MMICs abroad without licenses, arguing they should have been subject to similar export‑control investigations and potential criminal charges.

The government responded, and the court appeared to accept, that:

• Yu had not provided evidence, as opposed to mere assertion, that those exports violated law; and
• Custom MMIC’s exported files did not involve stolen trade secrets, as Yu’s case did.

Likewise, Yu’s list of Massachusetts civil trade secret litigations was insufficient because he offered no showing that prosecutors even knew of those matters or that the defendants there engaged in similar conduct under similar circumstances (e.g., foreign transfer, national security implications).

c. Result: no discriminatory effect under Armstrong

Given this record, the First Circuit held there was no clear error in the district court’s conclusion that Yu had not shown that “similarly situated individuals of a different race were not prosecuted.” The presumption of regularity thus remained intact, and the court did not need to formally resolve the discriminatory‑purpose prong, although it discussed context (e.g., China‑related DOJ statements) when addressing selective enforcement.

C. Equal Protection and Selective Enforcement

1. Legal Framework: Enforcement vs. Prosecution

Selective enforcement—targeting investigative efforts on one racial group but not similarly situated members of another—has long been recognized as violating equal protection. The First Circuit relied on precedents like Whren and Flowers v. Fiore to reaffirm that:

• a selective enforcement claimant must show both:
  • discriminatory effect (differential treatment of similarly situated comparators or statistical evidence of such differential treatment); and
  • discriminatory purpose (at least partial motivation by race, in the sense described in Wayte and Feeney);
• such claims may be based either on comparator evidence or on statistics robust enough to isolate race as a driving factor.

The unresolved doctrinal issue was what evidentiary standard applies: the “clear evidence” standard from Armstrong, or the ordinary preponderance standard used in most equal protection cases.

2. Doctrinal Innovation: Adopting the Preponderance Standard for Selective Enforcement

The First Circuit confronted this question head‑on and sided with the Seventh Circuit’s reasoning in Conley v. United States, holding:

Armstrong’s “clear evidence” requirement is limited to selective prosecution claims, where heightened deference to core prosecutorial discretion is grounded in separation‑of‑powers concerns. It does not govern selective enforcement claims against law enforcement officers, who are subject to greater judicial scrutiny and only qualified immunity.

Key steps in the court’s reasoning:

• Armstrong’s rationale – The heightened “clear evidence” standard protects the Executive’s core constitutional function of deciding whom to prosecute, under the presumption of regularity that attaches to prosecutorial decisions as exercises of presidential authority.
• Different institutional role of law enforcement – Agents and officers:
  • are subject to greater ongoing judicial oversight (e.g., via suppression hearings, cross‑examination, civil suits);
  • testify regularly and must justify investigative tactics; and
  • have only qualified immunity for civil damages, unlike prosecutors’ absolute immunity for core prosecutorial tasks (citing Imbler and Kalina).
• Given these differences, the separation‑of‑powers concerns and presumption of regularity justifying Armstrong’s “clear evidence” standard for prosecutors do not extend with equal force to law enforcement officers’ investigative decisions.

Accordingly, the court held that selective enforcement claims require proof by a preponderance of the evidence, i.e., more likely than not, consistent with the general standard in civil and most constitutional litigation.

This is a major doctrinal clarification within the First Circuit and contributes to a growing circuit‑level consensus recognizing a lower evidentiary threshold for selective enforcement than for selective prosecution, even if some circuits have not yet squarely confronted the issue.

3. Standard of Review for Selective Enforcement

The court extended the same review framework to selective enforcement:

• Legal questions (e.g., the applicable standard of proof, the structure of the equal protection test) – de novo review.
• Factual findings (e.g., on whether comparators are similarly situated, the weight of the evidence of bias, whether law enforcement would have acted differently but for race) – clear error review.

This matches the standard for other Rule 12 determinations and brings clarity to appellate review of such claims.

4. Application: No Discriminatory Effect

Although the district court acknowledged it found Yu’s selective enforcement motion “troubling” and saw “not insignificant evidence” of implicit bias, it ultimately concluded that Yu did not prove discriminatory effect or purpose by a preponderance of the evidence. The First Circuit found no clear error in this assessment.

a. Comparators and knowledge of law enforcement

As with selective prosecution, the court stressed that to show discriminatory effect via comparators, Yu had to demonstrate both:

• that specific non‑Chinese individuals engaged in roughly the same conduct under similar circumstances (e.g., possessing sensitive MMIC designs with potential military use, involving foreign fabrication or export control issues); and
• that law enforcement or relevant agencies were aware of those individuals’ conduct and chose not to investigate them.

Yu’s proffered comparators—chiefly the MACOM civil case and allegations about ADI and Custom MMIC’s exports—failed this dual test:

• He did not establish that investigators in his case, or relevant agencies, actually knew of those comparators, as opposed to arguing that they “could have, and should have” known about them.
• Absent such knowledge, the fact that some civil cases were not accompanied by criminal investigations cannot establish differential enforcement in the equal protection sense.

The court thus concluded that Yu had not shown that “one class is being treated differently from another class that is otherwise similarly situated,” as required for discriminatory effect.

b. Implicit bias and national security context

The district court explicitly recognized rising anti‑Asian racism since 2016 and acknowledged the possibility that Yu’s Chinese ethnicity “played a role in part in what happened here,” i.e., may have colored perceptions of risk. The First Circuit did not second‑guess that observation. But it drew a line between:

• general background concerns about implicit bias, and
• the structured constitutional requirement that discrimination must be shown to have caused differential treatment of similarly situated individuals.

On this record, the court accepted the district court’s further conclusion that, given:

• the sensitive, potentially military MMIC technology involved;
• the foreign manufacturing arrangement with Win in Taiwan;
• the specific anonymous tips that triggered the investigation;
• Yu’s travel to China and stated desire to “return to serve the country” (China) by starting a business there; and
• the overall national security environment concerning Chinese economic espionage,

law enforcement officers would likely have opened and pursued the same investigation against a non‑Chinese individual engaged in similar conduct with similar foreign ties. Thus, selective enforcement was not established by a preponderance of the evidence.

Because discriminatory effect was not proved, the court did not need to definitively resolve discriminatory purpose (though it discussed DOJ and Trump administration statements in its analysis). It also left open what the proper remedy would be if a selective enforcement claim were successfully established (dismissal, suppression, or other relief).

V. Precedents and Their Influence

A. United States v. Armstrong and the “Clear Evidence” Standard

Armstrong, the leading Supreme Court case on selective prosecution, framed the modern doctrine:

• Prosecutorial decisions enjoy a strong presumption of regularity because they are integral to the President’s constitutional duty under Article II.
• To overcome this presumption, a defendant must present “clear evidence” of both:
  • discriminatory effect (i.e., similarly situated individuals of a different race not prosecuted); and
  • discriminatory purpose (charging decisions made because of adverse impact on a protected group).

In Yu, the First Circuit reaffirms Armstrong’s rigor for prosecution claims, but limits its reach by holding that Armstrong’s “clear evidence” requirement does not govern enforcement claims. This is a careful calibration rather than a retreat from Armstrong.

B. Wayte v. United States and Feeney: Discriminatory Purpose

The court’s articulation of discriminatory purpose relies on Wayte and Personnel Administrator v. Feeney, which require that a government actor must select or continue a course of action because of (not just in spite of) its adverse effect on a protected group. The Yu opinion repeatedly invokes this formulation when evaluating arguments that DOJ and the Trump administration’s focus on Chinese espionage, and certain public statements, reflected improper racial motives.

By emphasizing that legitimate national security objectives—such as countering Chinese state‑sponsored economic espionage—can justify targeting certain conduct even if the affected population is disproportionately Chinese or of Chinese descent, the court applies Feeney’s distinction between intent to cause harm to a racial group and awareness that a lawful policy has disparate racial impact.

C. United States v. Lewis: “Similarly Situated” Analysis

Lewis is the First Circuit’s touchstone on the “similarly situated” requirement. Yu builds on Lewis by:

• reinforcing that the analysis is “fact intensive” and “no fact should be omitted” in defining the comparator pool;
• highlighting that factors such as foreign contact, export control issues, and national‑security implications are material to the comparability judgment; and
• applying these principles to reject superficial comparators (e.g., any civil trade secret case) that do not share the same overall risk and evidentiary profile.

D. Conley v. United States and the Emerging Law of Selective Enforcement

The Seventh Circuit’s decision in Conley is expressly cited and adopted in substance in Yu as to:

• the appropriateness of the preponderance standard for selective enforcement, and
• the notion that law enforcement is routinely subject to judicial scrutiny and qualified immunity, in contrast to prosecutors’ shield of absolute immunity for core prosecutorial functions.

By aligning with Conley, the First Circuit positions itself among those circuits drawing a clear line between Armstrong’s special solicitude for charging decisions and a more standard equal-protection evidentiary threshold for investigative decisions. While some other circuits have cited Armstrong in selective enforcement contexts, many have not squarely considered this distinction; Yu helps clarify the landscape.

E. Trade Secret Cases: BondPro and Martin

The opinion engages with trade secret doctrine through:

• BondPro Corp. v. Siemens Power Generation, Inc. (7th Cir.) – cited by Yu for the truism that a “trade secret that becomes public knowledge is no longer a trade secret.” The First Circuit does not dispute this proposition but emphasizes that the information here (layout and GDS files) had not become “public knowledge” in the operative sense because reverse engineering from the chip was not readily achievable.
• United States v. Martin (1st Cir.) – a prior criminal case interpreting § 1832, used by the court to analogize how knowledge and intent to deal in trade secrets can be inferred from defendants’ handling of “confidential” and “proprietary” materials, written communications, and covert behavior (e.g., referring to a source as a “spy”). Yu applies that reasoning to Yu’s concealment and protective actions regarding the files.

VI. Impact and Implications

A. For Criminal Trade Secret Enforcement

The opinion provides several important clarifications for criminal trade secret prosecutions under the Economic Espionage Act:

• Prototype and translation designs are protected. Even “abandoned” or intermediate designs can be trade secrets if they carry independent economic value by shortening development cycles or giving a competitive head start.
• Trade secrets can outlive product release. The release of a chip to the market does not automatically destroy trade secret status in its design files. The key question is whether the information is “readily ascertainable” through proper means. When reverse engineering is costly, time‑consuming, and sophisticated, secrecy (and criminal protection) may persist.
• Knowledge can often be inferred circumstantially. In cases involving sophisticated insiders bound by NDAs who actively conceal their use of proprietary files, courts are likely to infer knowledge of trade secret status even if the statutory mens rea element is not definitively resolved.

These points will make it easier for the government to prosecute misappropriation of complex R&D information in high‑technology sectors (semiconductors, defense, biotech, etc.), especially where defendants argue that reverse engineering was possible or that they only took “preliminary” designs.

B. For Equal Protection Claims in National Security and “China Initiative” Contexts

Yu is one of relatively few appellate opinions deeply engaging with claims of anti‑Chinese bias in the enforcement of economic espionage and trade secret laws. Its impact is mixed:

• On the one hand, it:
  • acknowledges the reality of rising anti‑Asian racism and the plausibility of implicit bias influencing perceptions of risk;
  • scrutinizes comparators and distinguishes legitimate national security focus on Chinese state‑linked threats from impermissible ethnic targeting; and
  • articulates a more accessible standard (preponderance) for proving selective enforcement.
• On the other hand, it:
  • sets a high bar for showing discriminatory effect through comparators, especially in complex national security contexts where individual cases may differ in subtle but legally significant ways (e.g., export‑control overlay, foreign manufacturing, specific tips); and
  • treats law enforcement’s categorization of Yu’s case as “foreign counterintelligence” and inclusion in “Chinese threat” briefings as insufficient by themselves to show that race, rather than perceived foreign ties and technology sensitivity, drove the investigation.

Future defendants raising similar claims will likely focus on building more robust comparator records (including proof that law enforcement knew of comparable non‑Chinese suspects and declined to investigate them) or statistically grounded challenges that isolate race from other confounding factors.

C. For the Law of Selective Enforcement

By expressly adopting the preponderance of the evidence standard for selective enforcement, the First Circuit:

• lowers the doctrinal barrier relative to Armstrong but still demands concrete evidence of comparably situated non‑targets or robust statistical proof;
• reinforces the distinction between:
  • prosecutorial charging decisions (highly protected, “clear evidence” required); and
  • investigative decisions (more open to judicial review, standard civil evidentiary threshold);
• aligns with a trend recognizing that equal protection scrutiny of policing should not be as restricted as scrutiny of prosecutorial decision‑making.

However, Yu also demonstrates that, even under the more accessible preponderance standard, proving discriminatory effect remains difficult without well‑documented, closely matched comparators and clear evidence that law enforcement was aware of and treated those comparators differently.

D. Implicit Bias, Judicial Acknowledgment, and Legal Standards

The district court’s candid discussion of implicit bias, and the First Circuit’s respectful acknowledgment of that discussion, signal a growing judicial willingness to recognize:

• that implicit or subconscious racial bias can influence enforcement decisions, especially in charged geopolitical contexts; but
• that such recognition does not, by itself, satisfy the constitutional threshold for equal protection violations, which still turns on demonstrable differential treatment and purpose “because of” group status.

Thus, Yu adds to the body of case law where courts take bias concerns seriously yet insist on rigorous evidentiary showings when defendants seek to invalidate prosecutions or investigations on that basis.

VII. Complex Concepts Simplified

A. What Is a “Trade Secret” Under Federal Criminal Law?

Under 18 U.S.C. § 1839(3), information is a “trade secret” if it:

1. Relates to a broad range of subjects (formulas, patterns, plans, designs, methods, techniques, programs, etc.);
2. The owner takes reasonable measures to keep it secret; and
3. It has independent economic value, actual or potential, because it:
   • is not generally known; and
   • is not readily ascertainable by proper means by others who could benefit from using it.

Key takeaways from Yu:

• Trade secret protection does not evaporate just because a product is sold; design information can remain secret if it’s hard to reconstruct.
• “Readily ascertainable” is a practical standard: if reverse engineering is expensive, time‑consuming, and complex, information may still be protected.
• Even preliminary or prototype designs can be trade secrets if they give a competitive edge.

B. GDS Files, MMICs, and Why They Matter

A few technical concepts:

• MMIC (Monolithic Microwave Integrated Circuit) – a tiny chip that amplifies or processes microwave‑frequency signals, used in radar, satellites, cell towers, and other infrastructure.
• Design layout – the precise blueprint of how all the transistors, capacitors, and other elements are placed and connected on the chip.
• GDS file – a standardized digital file format (GDSII) that encodes the layout in a form the chip foundry’s tools can use to manufacture the chip.

Think of the GDS file as the “master printing plate” for the chip—if you have it, you have the essence of the design without needing to rediscover it from the final product.

C. Constructive Amendment vs. Variance, and Why Yu’s Argument Failed

In criminal procedure:

• A constructive amendment occurs if, effectively, the government at trial charges the jury on a different crime or set of elements from what the grand jury indicted—this is forbidden.
• A variance is when the facts proved at trial differ from those alleged, but the legal elements are the same. Variances can be fatal if they prejudice the defendant, but some variances are allowed.

Yu argued that by proving his possession of a “prototype” instead of the final HMC1022A design, the government had constructively amended the indictment. The court responded that:

• Read in context, the indictment always pointed to the prototype translation file for HMC1022A as the charged secret.
• Therefore, there was no amendment and no variance: the charge and the proof matched.

D. Standards of Proof: “Beyond a Reasonable Doubt,” “Preponderance,” and “Clear Evidence”

Three different evidentiary thresholds appear in the opinion:

• Beyond a reasonable doubt – the highest standard, used for criminal guilt. The evidence must firmly convince the jury of the defendant’s guilt.
• Preponderance of the evidence – the usual civil and constitutional standard; something is more likely true than not (>50%). Yu holds this is the standard for proving selective enforcement.
• “Clear evidence” (in Armstrong) – a heightened standard, not precisely quantified but generally understood as more demanding than preponderance, reflecting strong proof that overcomes the presumption of regularity in prosecutorial decisions. This standard remains for selective prosecution.

E. Selective Prosecution vs. Selective Enforcement

Though related, these are distinct claims:

• Selective prosecution – challenges the decision to bring charges, made by prosecutors. Highly protected because it lies near the core of executive power. Governed by Armstrong’s “clear evidence” standard.
• Selective enforcement – challenges the decision to investigate or target particular individuals, made by law enforcement officers or agencies. Governed by equal protection standards and, after Yu, provable by a preponderance of the evidence in the First Circuit.

Both require proof of:

• Discriminatory effect – similarly situated members of another racial group were not treated the same.
• Discriminatory purpose – race or another protected characteristic was at least a motivating factor, in the Feeney/Wayte sense.

VIII. Conclusion

United States v. Yu is a consequential decision at the intersection of trade secret criminal law, national security, and equal protection doctrine.

On the substantive criminal side, the First Circuit affirms that:

• Prototype and translation designs can be trade secrets,
• trade secret status can survive product release where reverse engineering is not “readily” feasible, and
• knowledge of trade secret character can be inferred from sophisticated concealment and contractual secrecy obligations, without resolving the exact contours of § 1832’s mens rea.

On the constitutional side, the court:

• clarifies standards of review for selective prosecution and enforcement claims,
• maintains Armstrong’s “clear evidence” requirement for prosecutorial decisions while holding that selective enforcement claims need only be proved by a preponderance, and
• demonstrates in application how demanding the “similarly situated” comparator requirement remains in complex, national‑security‑tinged investigations, even under this lower evidentiary standard.

The opinion also illustrates the tension between a judiciary increasingly aware of implicit bias and structural racism, and the strict doctrinal demands of equal protection claims in the criminal context. By acknowledging the former but enforcing the latter, Yu both invites more sophisticated evidentiary showings in future cases and underscores that generalized concerns about bias, without concrete proof of discriminatory effect and purpose, will not invalidate federal criminal prosecutions.

In the broader legal landscape, Yu will likely serve as a key precedent for:

• federal trade secret prosecutions involving complex engineering designs and foreign manufacturing; and
• defendants challenging law enforcement targeting in cases that arise from geopolitical and national security concerns, especially where issues of race, ethnicity, and foreign origin are inextricably intertwined with legitimate investigative interests.