Full opinion text
GREGG COSTA, Circuit Judge: Software maker GlobeRanger obtained a $15 million judgment in a trade secret misappropriation trial against competitor Software AG. Software AG challenges that result on a number of grounds, but its principal argument is that GlobeRanger finds itself in a jurisdictional Catch-22. It argues that GlobeRanger’s trade secret claim is preempted by federal copyright law, but if not, then the result is no federal claim to support jurisdiction. Because we find that the trade secret claim is not preempted but that a dismissed conversion claim was preempted and supports federal jurisdiction, we also consider challenges to the sufficiency of the evidence, the damages award, and jury instructions. Finding no reversible error on those grounds, we AFFIRM. I. A. GlobeRanger specializes in radio frequency identification (RFID) technology, a type of wireless technology used to identify and track items and information. Those who drive on toll roads are likely grateful for the technology, even if they don’t recognize the term RFID. It is what allows drivers to speed through tollbooths while an electronic reader gathers information from a tag inside the car that is connected to the driver’s account. GlobeRanger’s software incorporates RFID for a different purpose: inventory management. Its RFID solutions filter, process, and store information from incoming inventory items in real time to help the user keep track of the items. The core of its RFID solution is its proprietary “Motion platform”, which is used in all of its inventory tracking software. This platform connects with RFID readers to quickly process transactions and implement complex workflows in real-time. For particular markets or companies with unique needs, the platform is then supplemented with add-ons, known as “solution accelerators,” that customize the product to satisfy those particular demands. For example, GlobeR-anger may combine the iMotion platform with a certain solution accelerator that helps make the RFID tracking system integrate or work well with other systems that are already used by clients in a particular industry. GlobeRanger has subcontracted on several projects creating RFID programs for Department of Defense agencies and their suppliers. This case arises from one such contract. In 2007, GlobeRanger entered into a subcontract with Science Applications International Corporation (SAIC) to build and implement a bundle of RFID technology for the Navy. The parties refer to this as the “Navy Solution.” GlobeRan-ger installed “instances,” which are fully-functioning versions of the Navy Solution, at three Navy bases. But the Navy then decided it wanted an enterprise-wide RFID system that could be run from a single location, rather than GlobeRanger’s system that required servers at each location. After considering competing proposals from GlobeRanger and Software AG (a larger, more general software company), the Navy went with Software AG’s proposal rooted in its “webMethods” software platform. The Navy ordered GlobeRanger to stop its subcontracting work and said it would convert, the three existing instances to the enterprise-wide system. Prior to receiving this Navy contract— termed RAVE (RFID Asset Visibility Enterprise) — Software AG had not implemented RFID for a client. While working on the Navy contract, Software AG accessed some of GlobeRanger’s data, manuals, and software. GlobeRanger asserts that in doing so Software AG misappropriated trade secrets in its Navy Solution. Software AG maintains that any access and use were permissible, particularly in light of federal regulations that applied to GlobeRanger’s work for the Navy. B. This litigation has a lengthy ánd messy history with both parties changing their view on whether this case should be in state or federal court. GlobeRanger initially brought suit in federal court. After Software AG objected that there was no subject matter jurisdiction, GlobeRanger voluntarily dismissed the case and refiled it in state court. In its state court filing, GlobeRanger alleged misappropriation of trade secrets, conversion, unfair competition, civil conspiracy, and tortious interference. Software AG now thought the dispute belonged in federal court. It removed the case, arguing that the Copyright Act completely preempted all of the state law claims except conspiracy. Back in federal court, Software AG filed a Rule 12(b)(6) motion to dismiss on preemption grounds. GlobeRanger sought remand to state court, arguing that none of its claims were completely preempted and thus the court lacked jurisdiction. The district court denied the motion to remand, finding that the misappropriation of trade secrets and unfair competition claims were preempted. Soon thereafter it ruled that the tortious interference and conversion claims were also preempted. As that left no remaining tort to support the derivative conspiracy claim, the court granted Software AG’s motion to dismiss in full. That dismissal was appealed, resulting in a ruling from this court the meaning of which is perhaps the most contentious issue in this second appeal. There will be more to say about that decision later, but for now we can summarize GlobeRanger I as follows. A different panel of this court overturned the dismissal, holding that at least some of the factual allegations in the trade secret misappropriation claim were outside the subject matter of the Copyright Act and therefore not preempted. GlobeRanger Corp. v. Software AG, 691 F.3d 702, 709 (5th Cir. 2012) (“GlobeRan ger I”). We also affirmed the denial of GlobeRanger’s motion to remand based on actual (or possible — this is the source of the intense debate) preemption of the conversion claim. Id. at 709-10. Although language in GlobeRanger I suggested that the district court- could have found on remand with a more developed record that the conversion claim was not preempted because it involved subject matter outside the scope of copyright, id. that issue was never reconsidered. Instead, less than a year after the remand, GlobeR-anger dropped the conversion claim. At the conclusion of discovery, Software AG moved for summary judgment on the merits of the remaining claims, without invoking its preemption defense. The district court granted summary judgment on the tortious interference claim and denied it on the trade secret misappropriation, unfair competition, and conspiracy claims. At a pretrial conference, GlobeRanger narrowed the case even more by dismissing its unfair competition claim. That left only the trade secret misappropriation and derivative conspiracy claim for trial. The district court denied Software AG’s motions for judgment as a matter of law. The jury found that Software AG misappropriated GlobeRanger’s trade secrets and awarded $15 million in compensatory damages. It found that GlobeRanger did not prove malice, and so was not entitled to punitive damages, nor did it prove conspiracy. The district court denied post-trial motions for judgment as a matter of law, a new trial, and remittitur:. Software AG appeals on the following grounds: (1) the trade secret misappropriation claim is preempted by copyright law; (2) if this claim is not preempted, then the district court lacked jurisdiction; (3) if the court did have jurisdiction, GlobeRanger failed to prove its claim or its damages; (4) the trial court’s damages award was erroneously calculated and excessive; and (5) the district court abused its discretion in formulating the jury charge. II. We turn first to the preemption and alternative jurisdictional arguments, both of which are subject to de novo review. See Franks Inv. Co. v. Union Pac. R.R. Co., 593 F.3d 404, 407 (5th Cir. 2010) (preemption by federal statute is reviewed de novo); Gilbert v. Donahoe, 751 F.3d 303, 306-07 (5th Cir. 2014) (subject matter jurisdiction is reviewed de novo). A. The different spheres of intellectual property protection can sometimes overlap. As the software code in this case illustrates, the same intellectual property can be protectable under the copyright laws or subject to trade secret protection. If the creator seeks the protection of the copyright laws, it obtains the exclusive right to make copies of the work for decades but must publicly register the work before enforcing that right through a lawsuit. 17 U.S.C. § 411(a). The creator might prefer to not publicly disclose the creation, in which case it can maintain the material as a trade secret if it takes reasonable measures to preserve secrecy. See generally, Stephen M. Drawee, Protecting Trade Secrets Through Copyright, 1981 Duke L.J. 981, 982 (1981) (describing how copyright protection is limited compared to trade secret protection in that it only covers expression and not underlying ideas, and requires disclosure of the trade secret). Depending on the business situation, one of these tradeoffs will be preferable to the other. The supremacy of federal law in the area of copyright means, however, that state protection of copyrightable subject matter must sometimes give way to its federal counterpart. Not surprisingly given the increasing importance of intellectual property to our economy, we have confronted this question of “copyright preemption” numerous times in recent years. See, e.g., GlobeRanger I, 691 F.3d 702; Spear Mktg., Inc. v. BancorpSouth Bank, 791 F.3d 586 (5th Cir. 2015); Carson v. Dynegy, Inc., 344 F.3d 446 (5th Cir. 2003); Computer Mgmt. Assistance Co. v. Robert DeCastro, Inc., 220 F.3d 396 (5th Cir. 2000); Alcatel USA Inc. v. DGI Techs., Inc., 166 F.3d 772 (5th Cir. 1999) (all addressing copyright preemption of state law claims). Although the Supreme Court has not addressed the question, in GlobeRan-ger I we joined other circuits in holding that the Copyright Act is one of the uncommon statutes (ERISA is another) that “completely preempts the substantive field” and thus “transforms a state-law complaint assei-ting claims that are preempted ... into ... a federal claim for purposes of the well-pleaded complaint rule.” 691 F.3d at 706. The Copyright Act preempts a state law claim when two conditions are met. We have already referred to this first requirement: the “work in which the right is asserted must come within the subject matter of copyright.” Alcatel, 166 F.3d at 785. If the intellectual property at issue js not subject to copyright protection, then obviously the Copyright Act would not preempt a state law cause of action protecting that uneopyrightable property. But even when the work is within the scope of copyright protection, state law is preempted only if “the right that the author seeks to protect ... [is] equivalent to any of the exclusive rights within the general scope of copyright.” Id. at 786. In other words, is state law protecting the same rights that the Copyright Act seeks to vindicate, or is it protecting against different types of interference? If state law offers the same protection, then the state law claim is preempted and must be dismissed. Copyright grants creators the exclusive right to copy their work, so this second requirement is met when the conduct for which the plaintiff is seeking protection under state law amounts to the copying that copyright law also proscribes. Spear Mktg., 791 F.3d at 598 (“Copying, communicating, and transmitting are equivalent acts to reproducing and distributing.”). In that situation, allowing a state law to provide tbe same protection as copyright law would undermine the federal copyright system, which also requires public registration of the work before a lawsuit can be brought. As a result, the Copyright Act preempts state claims that seek to provide copyrightable subject matter with the same protections that federal law provides. In GlobeRanger I, the first inquiry concerning copyrightable subject matter was the most hotly contested. If the allegations just targeted conduct relating to GlobeR-anger’s software code, the “falling within the subject matter of copyright” requirement for preemption was satisfied. But it was not then clear whether the allegations also included misappropriation of GlobeR-anger’s “procedures, processes, systems, and methods of operations,” such as how GlobeRanger “actually deployed] [the RFID program] on site,” “incorporated business process into its design of the warehouse,” and “trained sailors,” 691 F.3d at 708, 709. Given the pleading stage, we deferred to GlobeRanger’s assertions that the allegations did extend not just to theft of software but also to misappropriation of business methods beyond the scope of copyright protection. Id. at 709. That uncertainty about whether all of the creations for which GlobeRanger sought protection fall within the scope of copyright law is now gone. By the time of trial, GlobeRanger had whittled its case down to misappropriation of software that it concedes is copyrightable. That leaves equivalency of rights as the only disputed question for preemption. The test to determine equivalency is the “extra element test”: [I]f the act or acts of [the defendant] about which [the plaintiff] complains would violate both misappropriation law and copyright law, then the state right is deemed “equivalent to copyright.” If, however, one or more qualitatively different elements are required to constitute the state-created cause of action being asserted, then the right granted under state law does not lie “within the general scope of copyright,” and preemption does not occur. Alcatel, 166 F.3d at 787 (internal footnotes omitted). Whether a claim is equivalent requires looking to the actual alleged misconduct and not merely the elements of the state cause of action. Id. at 788 (equivalency is based on “the discrete facts of [a] ease”); Daboub v. Gibbons, 42 F.3d 285, 289-90 (5th Cir. 1995) (distinguishing case in which plaintiffs’ state-law claims were not preempted based on the specific acts complained of along with the elements of the state law). GlobeRanger argues that the right it seeks to vindicate under Texas trade secret law is not equivalent to the anti-copying principle of federal copyright law because the state law prevents acquisition “through a breach of a confidential relationship or ... improper means.” Tewari De-Ox Sys., Inc. v. Mountain States/Rosen, L.L.C., 637 F.3d 604, 610 (5th Cir. 2011) (quoting Phillips v. Frey, 20 F.3d 623, 627 (5th Cir. 1994)) (interpreting Texas misappropriation of trade secret law); see also, e.g., Twister B.V. v. Newton Research Partners, LP, 364 S.W.3d 428, 437 (Tex. App.—Dallas 2012, no pet.); Trilogy Software, Inc. v. Callidus Software, Inc., 143 S.W.3d 452, 463 (Tex. App. — Austin 2004, pet. denied). The evidence that Software AG accessed its Navy Solution through persons and entities that were in a confidential relationship with GlobeRan-ger means, according to GlobeRanger, that the trade secret claim involves an extra element not found in copyright law. We have never applied the extra element test to a common law misappropriation of trade secret claim. Our use of this test in cases involving other business torts provides guidance. In Alcatel, we held that a claim for misappropriation under the Texas common law of unfair competition was preempted because “the acts that form the basis of [Plaintiffs] misappropriation claim touch on interests clearly protected by the Copyright Act,” such as “reproduction” and “use” of the plaintiffs firmware, software, and manuals, and “distribution” of works derived from those materials. 166 F.3d at 789. The state law’s requirement that works be produced through “extensive time, labor, skill and money” was not an “extra element” sufficient to satisfy the equivalency test. Id. at 787-88. Because copyright protects work “in which independent creation and creativity converge,” the creator’s time, labor, and skill “are fundamental to the independent creation of a work” and thus “are necessarily contemplated in [a granted] copyright.” Id. at 789. We also found preemption in Spear Marketing, which considered claims for conversion of trade secrets as well as theft of trade secrets under the Texas Theft Liability Act (though notably no preemption defense was raised as to the misappropriation of trade secret claim). 791 F.3d at 598. As in Alcatel, the alleged misconduct involved only the “[e]opying, communicating, and transmitting” of works for which the Copyright Act also provides protection: Id.) 17 U.S.C. § 106(a)(1), (a)(2). That the Texas statute had a “knowingly” mens rea requirement did not constitute an extra element given the “well-established rule that elements of knowledge do not establish an element that is qualitatively different from a copyright infringement claim.” Spear Mktg, 791 F.3d at 598 (internal quotations omitted). As for conversion, the Court held that, although claims of conversion of physical property would not be preempted,. “to the extent [Plaintiff] allege[d] conversion of intangible ‘confidential information’ and ‘certain trade secrets,’ ” it is preempted under Glo-beRanger I, Alcatel and several other cases in which this circuit has held conversion of intangible property preempted. Id. at 597-98. In the “no preemption” column is Computer Management Assistance Co., 220 F.3d 396, 405 (5th Cir. 2000), which involved a claim under the Louisiana Unfair Trade Practice Act. Similar to Glob'eRan-ger’s allegations, a computer programming company alleged that the defendant accessed and unfairly misappropriated its materials, including trade secrets, through a shared customer. Id. at 399. In conducting the extra-element analysis, we held that because the Louisiana Unfair Trade Practices Act required that the plaintiff establish “fraud, misrepresentation or other unethical conduct,” the relief provided by the law is not equivalent to copyright and is not preempted. Id. at 404-05. The extra element in that Louisiana statute is similar to the right Texas trade secret law provides against the taking of protected information “through a breach of a confidential relationship or ... improper means.” Tewari De-Ox, 637 F.3d at 610. Indeed, akin to the “fraud, misrepresentation or other unethical conduct” that Computer Management held to be an extra element is the following evidence in this case: Software AG’s inducement of a former GlobeRanger employee to break his nondisclosure agreement in order to obtain source code; Software AG’s knowledge from technical manuals it obtained that end user agreements prohibited disclosure; and'its installment and exploration of the Navy Solution in a computer lab— given Software AG’s stated purpose of “re-plicat[ing] current GlobeRanger functionality” and GlobeRanger’s insistence that its software and license keys only be used for “direct support” of the existing Navy installments. Software AG’s actions go beyond the copying, communicating, and transmitting alleged in Alcatel and Spear Marketing. Because trade secret law protects against not just copying but also any taking that occurs through breach of a confidential relationship or other improper means, all ten circuits that have considered trade secret misappropriation claims have found them not preempted by the Copyright Act. See Seng-Tiong Ho v. Taflove, 648 F.3d 489, 503-04 (7th Cir. 2011); Stromback v. New Line Cinema, 384 F.3d 283, 302-05 (6th Cir. 2004); Dun & Bradstreet Software Servs., Inc. v. Grace Consulting, Inc., 307 F.3d 197, 217-18 (3d Cir. 2002); DSC Commc’ns Corp. v. Pulse Commc’ns, Inc., 170 F.3d 1354, 1365 (Fed. Cir. 1999); Bateman v. Mnemonics, Inc., 79 F.3d 1532, 1549-50 (11th Cir. 1996); Data Gen. Corp. v. Grumman Sys. Support Corp., 36 F.3d 1147, 1164-65 (1st Cir. 1994) abrogated on other grounds by Reed Elsevier, Inc. v. Muchnick, 559 U.S. 154, 130 S.Ct. 1237, 176 L.Ed.2d 18 (2010); Trandes Corp. v. Guy F. Atkinson Co., 996 F.2d 655, 658-60 (4th Cir. 1993); Gates Rubber Co. v. Bando Chem. Indus., Ltd., 9 F.3d 823, 846-48 (10th Cir. 1993); Computer Assocs. Int'l, Inc. v. Altai Inc., 982 F.2d 693, 716-21 (2d Cir. 1992); S.O.S., Inc. v. Payday, Inc., 886 F.2d 1081, 1090 n.13 (9th Cir. 1989). This overwhelming case law may be the reason that the defendant in Spear Marketing did not raise a preemption defense against the common law trade secret misappropriation claim even though complete preemption. of the Texas Theft Liability Act and conversion claims had been cited as the basis for federal jurisdiction. See 791 F.3d at 597-600 (finding the statutory and conversion claims preempted, but then proceeding to the merits of the trade secret claim). Software AG tries to distinguish much of this authority on the ground that a “considerable number” of the cases “held that misappropriation of trade secrets claims are not preempted because they require proof of a confidential relationship, which provides the extra element required to survive preemption.” Stromback, 384 F.3d at 303. Citing a district court case from more than twenty years ago — ancient history in the world of intellectual property litigation involving software that often raises the trade secret/copyright conundrum — Software AG says that a confidential relationship must exist between the plaintiff and defendant, which is not the case here because it was a former GlobeR-anger employee and subcontractors who allegedly breached duties of nondisclosure owed to the plaintiff. See Long v. Quality Computers & Applications, Inc., 860 F.Supp. 191, 197 (M.D. Pa. 1994) (stating that courts have not found trade secret claims preempted when they are “based upon the disclosure of material that a defendant has a duty to keep confidential”). We reject this as either an accurate characterization of the current state of preemption law or, if viewed as an original matter, as something that should make a difference in the equivalency analysis. At least three of the circuit cases finding no preemption of trade secret claims involved defendants that obtained the trade secret from a third party knowing that party owed a duty of nondisclosure to the plaintiff. See Trandes, 996 F.2d at 657, 660 (holding trade secret misappropriation not preempted in claim against defendant who “improperly acquired and used” trade secrets from a third party that had a contractual and fiduciary duty to the plaintiff); Gates Rubber Co., 9 F.3d at 830-31, 847-48 (holding trade secret misappropriation not preempted in case in which some of the defendants — former employees — had been in a confidential relationship, but at least one defendant — the corporation for which the other defendants now worked — had not); Computer Associates, 982 F.2d at 718-19 (explaining that a party that acquires a trade secret with notice that a third party stole it in violation of a duty of confidentiality can be guilty of misappropriation). Though a Second Circuit case, Computer Associates involved the same Texas trade secret cause of action at issue here. 982 F.2d at 718. It overturned a district court preemption holding that followed Software AG’s view that the defendant must owe the duty of nondisclosure to the plaintiff. Id. at 720. The court of appeals explained that Texas follows the Restatement approach, which makes a defendant liable for using a trade secret if it obtained the secret from a third party with actual or constructive notice that the third party’s disclosure violated a duty owed to the plaintiff. Id. at 718-19 (citing Restatement (First) of Torts § 757(c) (1939)). In that situation, just as when a defendant breaches a duty it owes the plaintiff, there is additional wrongful conduct beyond mere reproduction. We thus conclude that GlobeRanger’s trade secret misappropriation claim requires establishing an additional element than what is required to make out a copyright violation: that the protected information was taken via improper means or breach of a confidential relationship. Because the state tort provides substantially different protection than copyright law, it is not preempted. B. Our holding that the trade secret misappropriation claim is not preempted requires us to consider Software AG’s alternative jurisdictional out: that no federal question jurisdiction exists, so the case should have been remanded to state court. In light of our holding that the trade secret claim is not preempted, subject matter jurisdiction depends on whether some of the other, now-dismissed state law claims were completely preempted by the Copyright Act. Caterpillar Inc. v. Williams, 482 U.S. 386, 393, 107 S.Ct. 2425, 96 L.Ed.2d 318 (1987) (“Once an area of state law has been completely pre-empt-ed, any claim purportedly based on that pre-empted state law is considered, from ts inception, a federal claim, and therefore arises under federal law.”). And the answer to that turns in large part on the meaning of GlobeRanger I. Software AG contends that decision did not hold any of the claims preempted nor did the district court make such a finding on remand. Before delving into GlobeRanger I, we begin with some basic principles of federal jurisdiction that inform our reading of it. First is the “fundamental principle” that jurisdiction is determined based on the time of removal. Pullman Co. v. Jenkins, 305 U.S. 534, 537, 59 S.Ct. 347, 83 L.Ed. 334 (1939); Barney v. Latham, 103 U.S. 205, 215-16, 26 L.Ed. 514 (1880) (“The right of removal ... depends upon the case disclosed by the pleadings as they stand when the petition for removal is filed.”). Louisiana v. Am. Nat’l Prop. & Cas. Co., 746 F.3d 633, 635 (5th Cir. 2014). This removal rule is a corollary of the “time-of-filing rule” that governs jurisdictional determinations for cases originally filed in federal court. A rule of “hornbook law,” the time-of-filing rule traces at least as far back as an 1824 Supreme Court case holding that jurisdiction “‘depends upon the state of things at the time of the action brought.’ ” Grupo Dataflux v. Atlas Glob. Grp., L.P., 541 U.S. 567, 570-71, 124 S.Ct. 1920, 158 L.Ed.2d 866 (2004) (quoting Molan v. Torrance, 22 U.S. (9 Wheat.) 537, 539, 6 L.Ed. 154 (1824)). Both the time-of-filing and time-of-removal rule make sense. It would be inefficient for the court, as well as for litigants, if the power of the court to decide a case could exist at the outset, but later disappear based on changed circumstances. See Grupo Dataf- lux, 541 U.S. at 580, 124 S.Ct. 1920 (“The time-of-filing rule is what it is precisely because the facts determining jurisdiction are subject to change, and because constant litigation in response to that change would be wasteful.”). One of the changed circumstances that does not undo an initial determination of jurisdiction is the dismissal of claims. We have applied this principle before when the original jurisdiction in a removed case flowed from complete preemption. Hook v. Morrison Milling Co., 38 F.3d 776, 779-80 (5th Cir. 1994). In Hook, we explained that the plaintiffs post-removal dismissal of a wrongful discharge claim that the district court had found to be preempted by ERISA “does not divest the district court of its properly triggered subject matter jurisdiction.” Id. at 780; see also Spear Marketing, 791 F.3d at 592-93 (holding that the court must look to an original complaint, not an amended one, when determining preemption). What such a dismissal does is create a question of whether the remaining state law claims should be retained as a discretionary exercise of the district court’s supplemental jurisdiction. Hook, 38 F.3d at 780; 28 U.S.C. § 1367(c)(3). Yet Software AG never asked the district court to exercise that discretion after the dismissal of the potentially preempted claims, so it is foreclosed from raising that possibility now. Powers v. United States, 783 F.3d 570, 576-77 (5th Cir. 2015) (“ ‘While Article III jurisdiction must be considered sua sponte, review of the discretionary aspect to supplemental jurisdiction under § 1367(c) is waived unless raised in the district court.’ ” (quoting Acri v. Varian Assoc., Inc., 114 F.3d 999, 1000-01 (9th Cir. 1997) (en banc))) (alteration omitted). The sole jurisdictional question is thus whether the Copyright Act preempted any of the claims pending at the time of the removal. Software AG’s argument is that this Court departed from the time-of-removal principle in GlobeRanger I and left open the question of jurisdiction for a later determination that was never made. Its position hinges on the opinion’s following language that comes right after the statement that “for the purposes of jurisdictional analysis, the defendants make a sufficient argument to keep this case in federal court at the motion to dismiss stage.” GlobeRanger I, 691 F.3d at 709. To be clear, our conclusion on conversion is without prejudice to GlobeRan-ger’s ability to prove its claim relates to physical property.... If, with the aid of a more developed record, the district court concludes that GlobeRanger’s conversion claim relates to physical property, a different analysis will be required. If, however, the district court concludes that neither the conversion claim nor any other is preempted, it will lack jurisdiction over this case. Id. (emphasis added). As GlobeRanger dismissed the conversion claim that was the focus of this discussion after the case was back in district court, no further jurisdictional assessment occurred. We recognize some conflicting language in GlobeRanger I. But considering its context, language, and holding — as well as the background jurisdictional principles we have just discussed — our best reading is that it made a determination of federal jurisdiction over the since-dismissed conversion claim at the time of removal. Glo-beRanger I spends most of its time discussing the first question of copyright preemption — whether the state law claims cover material subject to copyright protection — in general terms applicable to all of the claims. That was a difficult determination to make at that early stage of the case because the complaint discussed both technical features of the software, which are the subject of copyright, and “procedures, processes, systems, and methods of operation” that fall outside the scope of copyright protection. 691 F.3d at 707-09. After noting that these latter allegations concerning business practices meant that “at least part of the factual basis for Glo-beRanger’s claims may fall outside of the scope of copyright,” the panel noted that a finding that none of the claims were preempted would mean “there would not be federal jurisdiction.” Id. at 709. It thus considered whether at least one claim was preempted that would support the continued pendency of the case in federal court, and focused on the best case for preemption, which was the conversion claim because that is a state law claim that essentially punishes taking or copying. See Spear Marketing, 791 F.3d at 597-98 (holding that Texas conversion claim was preempted). The discussion of the conversion claim that followed focused on whether that claim related to physical property (in which case no preemption because a conversion claim relating to physical property “is not equivalent to copyright protection”) or intangible property “such as where GlobeRanger places RFID readers” (in which case preemption because conversion of intangible property provides the same protection as the Copyright Act). GlobeRanger I, 691 F.3d at 709. Although GlobeRanger tried to argue in its brief that its conversion claim involved the taking of physical property, the panel noted that the portions of the complaint it cited actually referred to intangible property. Id. After observing that the complaint only alleged the preempted claim of converting intangible property, GlobeRanger I twice stated that it was finding federal jurisdiction based on the pleadings. Id. (“[F]or the purposes of jurisdictional analysis, the defendants make a sufficient argument to keep this case in federal court at the motion to dismiss stage”); id. at 710 (“Further, the defendants have argued enough of a basis for preemption of Glo-beRanger’s conversion claim to stay in federal court.”). As discussed, such a finding that a claim is completely preempted at the time of removal supports federal jurisdiction even if the claim is later dismissed. Giles v. NYLCare Health Plans, Inc., 172 F.3d 332, 335-36 (5th Cir. 1999) (finding appellate jurisdiction to review discretionary remand of nonpreempted claims because preemption of claims later dismissed supported federal jurisdiction). Consistent with that well-established principle as well as GlobeRanger 7’s holding that removal jurisdiction existed, we read that opinion’s language about a later jurisdictional determination as contemplating only that the conversion claim might not have ended up being governed by federal copyright law. Contrary to Software AG’s central premise, GlobeRanger I did not say that the district court had to make a renewed jurisdictional finding on remand. Instead, it gave GlobeRanger the option of “prov[ing that] its claim relates to physical property” contrary to the panel’s understanding of the claim. GlobeRan-ger I, 691 F.3d at 709. It made sense to not give Software AG that opportunity, as at that time only GlobeRanger had any interest in disproving GlobeRanger 7’s finding of preemption based on the pleadings. Recall that the parties’ positions have switched — in this case’s first visit to our court, GlobeRanger was arguing against preemption so it could go back to state court; Software AG was arguing for it so the case could stay in what was then its preferred federal forum. If GlobeRanger had accepted that invitation and produced evidence to show that its conversion claim related only to physical property so as not to be preempted, then GlobeRanger I was recognizing it could ultimately pursue that claim under state law. In that case, the copyright claim that was read into the pleading via complete preemption would have gone away. That situation would be no different than one in which á federal claim fails at the Rule 12(b)(6) stage, is found to lack evi-dentiary support at the summary judgment stage, or is voluntarily dismissed as actually happened in this case. In any of those scenarios, the court no longer has federal question jurisdiction over any pending claim, but the presence of the federal claim at the outset of the case supports supplemental jurisdiction over the remaining state claims. That is what happened here. As the complaint alleged only conversion of intangible property for which there is equivalency between the rights protected under that state tort and federal copyright law, complete preemption converted the conversion claim into one brought under the Copyright Act that supported federal question jurisdiction at the time of removal and supplemental jurisdiction after it was dismissed. III. With GlobeRanger having successfully navigated these jurisdictional challenges, we finally reach the merits of the case. Software AG argues that GlobeRanger failed to prove three different elements of a trade secret misappropriation claim: (1) what specifically constitutes the trade secrets; (2) whether Software AG acquired trade secrets improperly or with notice of impropriety, particularly in light of federal regulations governing contracts between the Navy and outside contractors; and (3) whether Software AG “used” any trade secret because there was no proof of copying. See Tewari De-Ox, 637 F.3d at 610; Twister, 364 S.W.3d at 437; Trilogy Software, 143 S.W.3d at 463 (each noting the elements of a Texas trade secret claim). As a result, it contends that it was entitled to judgement as a matter of law. Some of the arguments Software AG raises involve legal questions, which wé always review de novo. But to the extent they turn on factual findings by the jury, our standard of review “is especially deferential.” Wellogix, Inc. v. Accenture, L.L.P., 716 F.3d 867, 874 (5th Cir. 2013) (internal quotation marks omitted). A. Software AG first challenges the most foundational element of GlobeRanger’s claim: whether a trade secret existed. A trade secret is defined by Texas courts as “any formula, pattern, device or compilation of information which is used in one’s business and presents an opportunity to obtain an advantage over competitors who do not know or use it.” In re Bass, 113 S.W.3d 735, 739 (Tex. 2003) (internal quotation marks omitted). Whether a trade secret exists is a question of fact. Wellogix, 716 F.3d at 874. Texas courts weigh six factors, rooted in the Restatement (First) of Torts § 757 (1939), to determine whether a trade secret exists: (1) the extent to which the information is known outside of the business; (2) the extent to which it is known by employees and others involved in the business; (3) the extent of measures taken to guard the secrecy of the information; (4) the value of the information to the business and to its competitors; (5) the amount of effort or money expended in developing the information; (6) the ease or difficulty with which the information could be properly acquired or duplicated by others. In re Union Pac. R.R. Co., 294 S.W.3d 589, 592 (Tex. 2009). GlobeRanger produced sufficient evidence for a jury to reasonably conclude that at least some portion of its Navy Solution constituted a trade secret. Relevant to whether the software was valuable to the business and its competitors, there was testimony explaining how GlobeRan-ger’s “filtering” technology was unique and its ability to filter large amounts of information in real time added value compared to traditional inventory management systems. Both the value and the difficulty in reproducing the technology were also shown through emails and testimony from Software AG employees. .For example, when questioned about an email in which another Software AG employee asked a former GlobeRanger employee how to access the .Net source code in GlobeRanger’s solution, a testifying Software AG employee who described himself as the “technical lead” on the RAVE project explained that they needed the code because: When we were looking at the workflows in GlobeRanger, we could see step 1, then step 2, then step 3. Step 2 was a transformation. And looking at the workflow, we couldn’t see what the transformation was — we didn’t know what fields were being transformed and what business logic was being implemented, as what Navy logic was being implemented in step 2. In terms of the extent that the information is known outside the business and efforts to keep it secret, GlobeRanger offered testimony about its unique source code for Motion and other proprietary software, and how a “secret code” was required for each customer’s database. GlobeRanger’s Vice President of Software Services and Delivery detailed how the company worked to keep its software “contractually, as well as physically and electronically” protected, including: restricting physical access to the data center; requiring double password protection to access the “source code control system;” limiting access to source code only to developers rather than all employees; and licensing requirements for the software. And the unique license key required to access a particular product — including the Navy Solution software — can limit use of the product for a certain time period or to a certain number of devices. GlobeRanger also produced evidence that it reminded the Navy and RA that a license key it provided was only for maintenance of the current installation and not any other purpose. Considering this and other evidence produced at trial, a jury could reasonably conclude that at least some aspects of the Navy Solution constituted a “formula, pattern, device or compilation of information” GlobeRanger used to “obtain an advantage over competitors who do not know or use it.” In re Bass, 113 S.W.3d at 739. Software AG’s primary problem is with the “some aspects” part of what we just said. It argues that this evidence was not enough because GlobeRanger needed to show more specific, “concrete secrets.” Composite Marine Propellers, Inc. v. Van Der Woude, 962 F.2d 1263, 1266 (7th Cir. 1992) (citing Illinois trade secret misappropriation law). But it has not cited, nor can we find, any Fifth Circuit or Texas case law requiring greater specificity than what GlobeRanger provided at trial. See, e.g., Wellogix, 716 F.3d at 875 (holding in a similar posture that “Wellogix presented sufficient evidence and testimony to support the jury’s finding that Wellogix’s technology contained trade secrets,” without requiring a specific description of the trade secret). Software AG has therefore not shown that the jury’s finding that GlobeR-anger’s RFID technology contained trade secrets is unsupported as a matter of law. B. Software AG’s next contention — and one of the more complex issues in this case — is that GlobeRanger did not prove that Software AG acquired any of its alleged trade secrets improperly, and therefore failed to prove the second element of Texas trade secret misappropriation. Its arguments are rooted in application of Federal Acquisition Regulations (FAR) and the Defense Federal Acquisition Regulation Supplement (DFARS). FAR and DFARS regulate Department of Defense contracts. They impact what rights the Navy has to the technology it acquires from or pays to have developed by outside contractors like GlobeRanger. The regulations attempt to balance contractors’ interest in retaining their intellectual property with the government’s interest in obtaining sufficient rights to operate technology over its full life cycle and to retain the ability to work with other contractors on matters involving the technology. Christine C. Trend, Killing the Goose that Laid the Golden Egg: Data Rights Law and Policy in Department of Defense Contracts, 34 Pub. CONTRACT L.J. 287, 296 (2005). Resolution of this issue turns largely on which of three classifications the Navy Solution is given: “Technical Data,” “Noncommercial Computer Software and Noncommercial Computer Software Documentation,” or “Commercial Computer Software and Commercial Computer Software Documentation.” If a GlobeRanger trade secret constitutes technical data or noncommercial computer software/documentation, then the Navy possesses either “government purpose rights” or “unlimited rights” in that information, and may share it with subsequent contractors like Software AG. Defense Federal Acquisition Regulation Supplement (DFARS), 48 C.F.R. § 252.227-7013(a)-(b), § 252.227-7014(a)-(b). In that case, Software AG’s use would have been lawful and not subject to a misappropriation claim. But the third category, commercial computer software, has different default rights; namely, this software “shall be acquired under the licenses customarily provided to the public unless such licenses are inconsistent with Federal procurement law or do not otherwise satisfy user needs.” 48 C.F.R. § 227.7202-l(a). If the government requires broader rights than are normally granted in the private market, such rights must be negotiated for and specified in the contract. 48 C.F.R. § 227.7202-3(b). The terms of GlobeRanger’s general license, like those of most software companies, do not allow the end user to copy or distribute its product. The “commercial” designation signifies that the technology was initially developed in the general marketplace before its use in a military contract, which is why the contractor should be able to retain its intellectual property rights unless the military negotiates greater rights. The definition of the term thus focuses on when the technology was developed. In order to be “commercial,” it must be “software developed or regularly used for nongovernmental purposes” that meets one of the following requirements: (i) Has been sold, leased, or licensed to the public; (ii) Has been offered for sale, lease, or license to the public; (iii) Has not been offered, sold, leased, or licensed to the public but will be available for commercial sale, lease, or license in time to satisfy the delivery requirements of this contract; or (iv) Satisfies criterion expressed in paragraph (a)(l)(i), (ii), or (iii) of this clause and would require only minor modification to meet the requirements of this contract. 48 C.F.R. § 252.227-7014(a)(l). The Navy Solution was not sold, leased or licensed to the public as a package, nor was it intended to be in the future. But various products with the Motion platform were. The question thus becomes whether the Navy Solution was only a minor modification of the Motion product already sold in the marketplace. Id. § 252.227-7014(a)(l)(iv). “Minor modification” is “a modification that does not significantly alter the nongovernmental function or purpose of the software or is of the type customarily provided in the commercial marketplace.” 48 C.F.R. . § 252.227-7014(a)(13). Indeed, this was the focus of the argument at trial, with the court giving the jury a definition of minor modification that tracked the regulation. The evidence permitted a reasonable jury to find that the specification performed for the Navy was customary for all clients or that the modifications made to the underlying Motion and accelerator add-ons were minimal. Of the over 3.2 million lines of code in the installments of Navy Solution at the Hawaii bases, only about 24,000, or less than 1%, involved changes from the preexisting product. Cf. 48 C.F.R. § 2.101(b) (noting that for a “commercial item,” “the value and size of the modification and the comparative value and size of the final product” is a “factor!]” in determining if a modification is minor). The process of specializing Motion for the Navy was similar to customization provided to commercial clients, such as a large, international flower sales company. And the time and money spent on the Navy project included training on the new programs and installing them at different bases, thus countering Software AG’s argument that the $3 million price tag was too much for such “minor” changes. At a minimum, components of the Navy Solution easily qualify as commercial computer software. The source code underlying Motion or one of the accelerator add-ons is “commercial computer software,” as these components were made available to other customers. Indeed, Motion and one accelerator were even on the Government Services Association Federal Supply Schedule, which by definition lists commercial products. 48 C.F.R. § 8.402(a) (describing the Federal Supply Schedule program). In addition to arguing that the evidence showed that the Navy Solution required more than minor modifications to the preexisting Motion platform, Software AG argues that meeting the definition of commercial computer software is not enough for GlobeRanger to retain trade secret protection against future Navy contractors. It contends that as long as the Navy contributed any funding to the development of the Navy Solution, it retained a broader license in the product that allowed for Software AG’s later use. At trial and afterwards, the district court determined that funding did not matter because it interpreted the DFARS provisions to mean that if alleged trade secrets were commercial computer software, then normal licensing terms applied. Not just the language of the “commercial software” regulation quoted above, but also multiple secondary authorities support that view. Andrea B. Mayner, Understanding DFARS Technical Data Rights Regulations, Cont. Mgmt., Nov. 2004, at 16, 23, available at http://www.maynerlaw.com/ files/0C2F8_CM_Nov04_pl6.pdf (explaining that “contractors are able to make minor modifications to commercial software at government expense” and the software “will still be considered commercial”); Trend, 34 Pub. Cont. L.J. at 323 (“Minor modifications to [commercial computer] software will not deprive it of commercial status.”). On appeal, Software AG relies for the first time on the following regulation to support its assertion that funding matters for all three categories of technology, including commercial computer software: [U]se the clause at 252.227-7013, Rights in Technical Data — Noncommercial Items, in addition to the clause at 252.227-7015 [Technical Data — Commercial Items], if the Government will have paid for any portion of the development costs of a commercial item. The clause at 252.227-7013 will govern the technical data pertaining to any portion of a commercial item that was developed in any part at Government expense, and the clause at 252.227-7015 will govern the technical data pertaining to any portion of a commercial item that was developed exclusively at private expense. 48 C.F.R. § 227.7102-4(b). This regulation was not presented to the district court, which alone means we should not use it to overturn the verdict. Louque v. Allstate Ins. Co., 314 F.3d 776, 779 n.1 (5th Cir. 2002) (“Our court has held that, for obvious reasons, we will not consider evidence or arguments not presented to the district court for its consideration in ruling on the motion.”) (internal quotation marks, alterations, and emphasis omitted). In any event, there are multiple reasons to doubt the broad reading Software AG gives the regulation which would effectively nullify the commercial computer software regulation (because the government would have “funded” the minor modifications to preexisting software that section 252.227-7014(a)(1) says allow the contractor to retain the rights it would normally have under commercial computer software — that of any other licensee, 48 C.F.R. § 227.7202-l(a)). For one thing, section 227.7102^4(b) appears to apply to cost-reimbursement contracts for research and development. William C. Anderson, Comparative Analysis of Intellectual Property Issues Relating to the Acquisition of Commercial and Noncommercial Items by the Federal Government, 33 Pub. Cont. L. J. 37, 53 & n.91 (2003) (noting the regulation at issue governs “[i]f development costs are reimbursed to the contractor”). Although the government generally prefers fixed-price contracts, those cost-reimbursement contracts are preferred when it comes to research and development projects because of imprecise specifications and difficulties with estimating costs. 48 C.F.R. § 35.006(b), (c). The reimbursement process would also facilitate identification of the development for which the government is responsible and in which it obtains rights. GlobeRanger’s contract with the Navy, which was for the installation of a product, does not appear to be the type of contract to which section 227.7102-4(b) would apply. For another thing, section 227.7102-4(b) refers to a “commercial item” and the most applicable definition excludes computer software from that term. Software AG cites a definition in 48 C.F.R. § 2.101, a FAR definition provision that says “computer software” can be a “commercial item.” But the DFARS definition explicitly states that “Commercial item does not include commercial computer software.” 48 C.F.R. § 252.227-7015(a)(l). Because this definition not only comes from the more specific defense contracting regulations, but is cross-referenced in the provision at issue to govern rights in any technical data for a privately-funded commercial item, the best reading is that this provision does not cover “commercial computer software.” Finally, the provision states that the provisions giving the government broader rights will “govern the technical data pertaining to any portion of a commercial item.” 48 C.F.R. § 227.7102-4(b) (emphasis added). “Technical data” does not include computer software. 48 C.F.R. § 252.227-7013(a)(15). For this additional reason, section 227.7102^(b) does not apply. Our foray into the byzantine world of government contracting regulations thus does not find a basis for upsetting the verdict. The district court properly read the regulations to find that the Navy, and thus Software ÁG, could not lawfully use anything from the earlier Navy project that constituted “commercial computer software.” And there was sufficient, evidence for the jury to apply that classification to GlobeRanger’s technology. C. Software AG’s final argument for .why it should have been granted judgement as a matter of law is that GlobeRanger failed to prove Software AG used any trade secrets. It primarily relies on its interpretation of Spear Marketing, which it argues requires that if an alleged trade secret is in software, “use” must be “copying” and demonstrated via the “access-plus-similarity test.” Spear Mktg., Inc., 791 F.3d at 600-01. That test, also used in copyright cases, is a circumstantial way that copying can be proved in the absence of direct evidence. Id. at 601 & n.79 (citing RestateMENT (THIRD) OF UNFAIR COMPETITION § 40 emt. c (1995)). Just like the absence of evidence proving the second requirement of similarity doomed the trade secret claim in Spear Marketing, Software AG contends that the lack of similarity evidence between its software and GlobeRanger’s defeats the “use” element. We disagree that trade secret cases involving software require application of the access-plus-similarity test when direct evidence is unavailable. “Use” has a “broad” definition in trade secret law. Id. at 600. “As a general matter, any exploitation of the trade secret that is likely to result in injury to the trade secret owner or enrichment to the defendant is a ‘use[.]’ ” Wellogix, 716 F.3d at 877 (quoting General Universal Systems Inc. v. HAL Inc., 500 F.3d 444, 451 (5th Cir.2007)). Examples of use that fall within this definition include “employing the trade secret in manufacturing or production” and “relying on the trade secret to assist or accelerate research and development.” Wellogix, 716 F.3d at 877. Wellogix applied that latter type of “use” in a software case (the software helped with construction of oil wells): “the standard for finding ‘use’ is not whether [Defendant’s] templates contained [Plaintiffs] trade secrets, but whether [Defendant] “relied on the trade secrets to assist or accelerate research or development’ of its templates.” Id. (alterations omitted) (quoting Gen. Universal Sys. Inc. v. HAL Inc., 500 F.3d at 451). Nor was Wellogix our first case to include use in research and development within misappropriation. See, e.g., Gen. Universal Sys., Inc., 500 F.3d at 451 (interpreting Texas law in holding that “use” included using the plaintiffs product to “accelerate the development and marketability” of a defendant’s own product). That is the same manner in which Software AG used GlobeRanger’s software. Nonetheless, in upholding a finding of trade secret misappropriation when the software was used to assist in research and development of a new project, Wello-gix is arguably in tension with the statement a year later in Spear Marketing that “when the trade secret at issue is a technical feature of a computer program — as opposed to ‘marketing or mode of doing business trade secrets’ — ‘the issue of misuse boils down to evidence of copying.” Spear Mktg., Inc., 791 F.3d at 600-01 (alteration omitted) (quoting Plains Cotton Coop. Ass’n of Lubbock v. Goodpasture Computer Serv., Inc., 807 F.2d 1256, 1263 (5th Cir.1987)). Software AG argues Spear Marketing should control because it is “more recent.” But to thé extent there is conflict between the two, under our rule of orderliness, the earlier case controls. See, e.g., Spong v. Fid. Nat’l Prop. & Cas. Ins. Co., 787 F.3d 296, 305 (5th Cir. 2015) (“It is a well-settled Fifth Circuit rule of orderliness that one panel of our court may not overturn another panel’s decision, absent an intervening change in the law, such as by a statutory amendment, or the Supreme Court, or our en banc court”). But that begs the question whether Plains Cotton, the 1987 ease on which Spear Marketing relies, is in conflict with Wellogix. Although some of the Plains Cotton language is suggestive of a restrictive approach to software trade secrets, we conclude that, given other language and the context of the opinion, it does not preclude “use” from including reliance on trade secrets in facilitating research and development. The Court explained that “[i]f no copying occurred on any level, [plaintiff] cannot demonstrate that [defendants] misused the trade secrets,” and noted that the plaintiff must show the defendants “in any way ‘cop[ied]’ any part of appellant’s protected idea or expression” Plains Cotton, 807 F.2d at 1263 (emphases added). The use of “on any level,” “in any way,” and “any part” is notable; Plains Cotton — which analyzed whether a preliminary injunction should have been granted and did not dispose of the case altogether — compared trade secrets in software design to “alleged marketing or mode of doing business trade secrets.” Id. There is no indication that the court considered whether use in research and development qualified, as is at issue here. Nor is it clear that Plains Cotton — or Spear Marketing, for that matter — intended “copying” to exclude, for example, using actual software code to create new source code. Plains Cotton agreed with the plaintiffs that evidence of copying could include “the reproduction of the organizational structure of a software system, outlined generally in the software’s design specifications,” which is “a level broader than verbatim, line-by-line copying.” 807 F.2d at 1260. We thus do not read Plains Cotton as rejecting the well-accepted view that “use” of a trade secret for purposes of trade secret misappropriation can include using the secrets to shortcut or accelerate research and development. See Restatement (Third) of Unfair Competition § 40 cmt. c (1995). But all this parsing of our case law is not dispositive on this question of Texas trade secret law. A recent pronouncement from the Supreme Court of Texas is. See Savoie v. Huntington Ingalls, Inc., 817 F.3d 457, 464 n.5 (5th Cir. 2016) (noting that we are bound by the most recent decision of a state supreme court); Broussard v. Southern Pac. Transp. Co., 665 F.2d 1387, 1389 (5th Cir. 1982) (en banc) (explaining that we must follow our decisions interpreting state law unless a subsequent state court decision establishes that our prior decision is clearly wrong). Until recently, that court (which has not decided many trade secret cases in recent years) had not relied on the Restatement (Third) of Unfair Competition in addressing a question of trade secret use.- It did so after oral argument in this case, defining use broadly to mean “commercial use by which the offending party seeks to profit from the use of the secret” and citing our statement in a software case that such exploitation includes “relying on the trade secret to assist or accelerate research or development.” Sw. Energy Prod. Co. v. Berry-Helfand, 491 S.W.3d 699, 721-22 (Tex.2016) (quoting Gen. Univ. Sys., Inc., 500 F.3d at 450-51 (quoting Restatement (Third) of Unfair Competition § 40 cmt. c)). Southwestern Energy thus removes any doubt that Texas follows this principle of “traditional trade secret law,” which recognizes that if the concept of use “were not flexible enough to encompass modified or even new products that are substantially derived from the trade secret of another, the protections the law provides would be hollow indeed.” Mangren Research & Dev. Corp. v. National Chem. Co., 87 F.3d 937, 944 (7th Cir. 1996). GlobeRanger’s evidence is sufficient to show that Software AG used the Navy Solution in developing its own product. Indeed, much of the “use” evidence is similar to that in Wellogix, including the following: Software AG accessed implementations of the Navy Solution; received confidential system keys under the pretense of maintenance; acknowledged trying to replicate GlobeRanger’s functionality; and obtained confidential source code from a former GlobeRanger employee in order to figure out how .the technology worked, all while it was in the p