dissenting:
I concur in the majority opinion in its statements of the controlling law and conclusions as to the National Stolen Property Act (“NSPA”) and the Jury Instructions. I respectfully dissent however, as to Part II.A.2.a, because I believe that the majority’s discussion of the Electronic Espionage Act (“EEA”) directly conflicts with our decision in United States v. Aleynikov, 676 F.3d 71 (2d Cir.2012). The majority ignores the factual similarities of Aleynikov and its narrow construction of the EEA, only months after the decision was rendered, in order to, in effect, retroactively apply Congress’s statutory changes made during the interim period. For this rea*263son, applying the principle of stare decisis, I must dissent.1
I. Aleynikov and the Economic Espionage Act
Under the EEA, a violation occurs when, “[wjhoever, with intent to convert a trade secret, that is related to or included in a product that is produced for or placed in interstate or foreign commerce, to the economic benefit of anyone other than the owner thereof, and intending or knowing that the offense will, injure any owner of that trade secret ... without authorization ... transmits ... or conveys such information.” 18 U.S.C. § 1832(a) & (2) (emphasis added). As the majority correctly states, this highlighted language creates two requirements: (1) a product requirement, that the product “is produced for or placed in” interstate or foreign commerce and (2) a nexus requirement, that the trade secret is “related to or included in” that product.
In Aleynikov, under a self-admittedly narrow construction of the statute’s two requirements, we reversed the defendant’s EEA conviction. We concluded that “produced for or placed in” created a statutory limitation, under which Goldman’s HFT System did not qualify as a “product.” 676 F.3d at 82. Because Goldman created the System to be used internally, as a company tool, rather than to be “produced for” distribution in interstate commerce, we concluded it did not satisfy the product requirement. Additionally, we concluded that the stolen trade secret, Goldman’s HFT code, was not sufficiently “related to” any plausible product because it did not make anything that ultimately would be placed in the stream of commerce. Id. at 81 n. 5. See also id. at 82.
The present case, with a nearly identical fact pattern,2 confronts similar challenges. From the outset of this case, the government, just as it did in Aleynikov, alleged that the HFT System was the “product” *264which was “related to” the stolen code. Therefore, Agrawal’s EEA count should have failed just as it did in Aleynikov. Instead, the majority reaches the opposite conclusion, upholding Agrawal’s EEA conviction. It adopts the government’s new theory — advanced for the first time on appeal — that the securities, not the HFT System were the alleged “product,” and that the code was sufficiently “related to” those securities. No doubt the majority’s misapprehension of both law and fact is in part driven by its conviction that the defendant is a “thief’ and its wish to retroactively apply Congress’s amendment to the EEA. Maj. Op. at 241, 244 n. 7.3 However, whether or not Agrawal’s EEA count would stand under the new statute,4 we are bound by precedent and not by the benefits of hindsight.
1. Product Requirement
As to the first requirement, in Aleynikov, we decided that Goldman’s HFT System was not a “product” that was “produced for or placed in” interstate commerce because the phrase was of limited reach. Observing the legislative history that “produced for or placed in” had not been in the original statute, we concluded that its inclusion must have implemented a limitation. See Aleynikov, 676 F.3d at 79-81. A plain reading of the text, construing the two categories in relationship to one another, also revealed “produced for” was distinct and narrow as compared to “placed in.” Id. at 79-82. We stated,
Products that have not yet been “placed in” commerce but are still being developed or readied for the marketplace can properly be described as being “produced for,” if not yet actually “placed in,” commerce. Reading the statute in this way gives effect to both categories of product (those “produced for” commerce and those “placed in” commerce), without making one a subset of the other.
Id. at 80.
We further determined that the district court read the phrase “produced for” too broadly when it found Goldman’s HFT System was a “product.” Id. Just because Goldman used the System “to rapidly execute high volumes of trades in various financial markets” and “[t]he Trading System generates millions of dollars in annual *265profits,” was not enough to find that the HFT System was “produced for” interstate commerce. Aleynikov, 676 F.3d at 80. We stated, these comments by the district court about the market and the profitability of the System, evaluated the product requirement “in a vacuum.” Id. Under this “untenable” interpretation, “every product actually sold or licensed is by definition produced for the purpose of engaging in commerce, every product that is ‘placed in’ commerce would necessarily also be ‘produced for’ commerce — and the phrase ‘placed in’ commerce would be a surplusage.” Id.
Conversely, we held that a product “produced for or placed in interstate or foreign commerce” could not include an organization’s internal tools such as Goldman’s HFT System. Only a product actually “produced for” interstate commerce could qualify. Therefore, because “[Goldman] went to great lengths to maintain the secrecy of its [S]ystem” and the System’s “enormous profits ... depended on no one else having it,” we concluded that “the HFT [S]ystem was not designed to enter or pass in commerce,” and thus it did not qualify as a “product” under the statute. Id. at 82.
Here, the same facts apply. Since the initial stages of this case, the government identified SocGen’s HFT System as the product. Indictment at ¶ 19. The plain words of the indictment clearly allege under Count One, that SocGen’s product was “the Financial Institution’s high frequency trading business” and the stolen trade secret was the Institution’s “proprietary computer code for [that System].” Indictment at ¶ 19.5 The indictment further alleged, “the Financial Institution spent millions of dollars to develop and maintain a computer system that was used in high-frequency trading (the ‘Trading System’). [And t]he Trading System is generally composed of a network of computers and computer code (the ‘Code’).” Indictment at ¶ 5.
The government continued to assert its theory, that the HFT System was the product and the code was its trade secret. At the district court level, the government belabored the point that the HFT code was “the building blocks of [the] software program ” and it was “the system [that] made a huge number of trades per day, which added up to millions of dollars in profits a year.” Transcript at 19. In summation the government stated, “the whole point of stealing [the code] was to turn it into something that Tower could use,” into a product that “Tower want[ed]” and that was SocGen’s “system.” Transcript at 1257-58 (emphasis added). Even the district court echoed the government’s HFT System-as-product theory when it instructed the jury that “Soeiété Générale’s property [was] some or all of the computer code used by Soeiété Générale in its high frequency trading operation.” Transcript at 1315.
In its brief on appeal, the government once again asserted that,6 “SocGen’s HFT *266system, which traded securities on multiple markets in the United States, was ‘produced for ’ the very purpose of engaging in interstate commerce.” Appellee’s Br. at 58. It continued, “The trading system was created to trade securities in various markets located in the United States, such as the Chicago Mercantile Exchange, from SocGen’s offices located in Manhattan. Indeed, to engage in interstate commerce was not just the primary purpose of the system; it was the only purpose of SocGen’s HFT system. Thus, the trading system is ‘produced for’ interstate commerce under any commonsense and ordinary reading of those terms.” Id. at 58-59 (internal citations omitted). Because this Court decided in Aleynikov, an HFT System is not a product, we should conclude Agrawal’s EEA count similarly fails.
Instead, the majority tries to distinguish Aleynikov by adopting the government’s new theory. Presented for the first time on appeal, the government asserted in its supplemental briefing before this Court that the securities, not the HFT System were the alleged product. The majority adopts this argument and writes, “in [ Aleynikov ] ... the government and the court elsewhere specifically identified the trading system as the relevant product. Where, as here, no pleading, argument, or charge ever labeled SocGen’s trading system a product.” Maj. Op. at 246. Instead, the majority alleges that the government’s “argument is more reasonably understood to identify the stocks and futures bought and sold on national exchanges as the products placed in interstate commerce.” Id. It relies on the indictment’s statement “that SocGen engaged in ‘high-frequency trading in securities’ ” and the district court’s jury instruction which required proof that, “the purpose of the computer code ... was to effectuate securities trades, at least some of which were in interstate or foreign commerce.” Id. at 246. “This,” the majority claims, “effectively identified securities as products traded in interstate commerce.” Id. at 246.
However, the majority plainly admits, “Agrawal’s indictment did not state this [securities-as-product] theory.” Id. This admission attempts to gloss over the indictment’s clear identification of the HFT System as the product and the fact that the government’s newly minted securities-as-product theory is nowhere revealed in the record. It attempts to evade this fact by pointing to the district court’s charge where it instructed the jury that the EEA conviction may be upheld “if the government proves that purpose of the computer code was to effectuate securities trades.” Transcript at 1315 (emphasis added). However, this jury charge did not identify the securities as the product under the EEA. Instead, it identified the security trades as a mere result of using the HFT System. In essence, the court’s instructions also identified the System as the product, which was used to effectuate the trades. In no way does this mere mention of the word “securities” reverse the government’s and district court’s consistent adherence to the HFT System-as-produet theory. In no way does it undermine the clear words of the indictment. Thus, the majority’s heavy reliance on this jury charge is misplaced. We therefore should *267conclude the HFT System-as-product theory does not satisfy the statute’s requirement, as we did in Aleynikov7
2. Nexus Requirement
Even assuming the product requirement is satisfied, the majority’s opinion still incorrectly upholds the EEA conviction because it misconstrues the nexus requirement. Under the EEA’s nexus provision, a trade secret must be “related to or included in” the product. Like the product requirement, the meaning of this phrase is governed by the “doctrine of statutory interpretation which instructs that words in a statute are known by the company they keep.” Aleynikov, 676 F.3d at 80. Similar to our determination that the “produced for or placed in” posited two separate protections, so too does “related to or included in” reach two distinct characteristics. Thus, “related to” must be interpreted in such a way that it does not render “included in” obsolete.
In Aleynikov, we gave one hint as to what might constitute “related to” when we held that “[bjecause the HFT system was not designed to enter or pass in commerce, or to make something that does, Aleynikov’s theft of source code relating to that system was not an offense under the EEA” 676 F.3d at 82 (emphasis added). Under this interpretation, the phrase “related to” is most naturally read to deal with things like a piece of specialized machinery, which itself is not intended to enter the stream of commerce, but which makes the product that does so. In contrast, “included in” refers to those items that, like a can of Coke, are part of the product in a physical sense. Under this reading of the statute, the relationship between the HFT System and the securities is too attenuated for them to be “related to” one another.
Both the government and the majority fail to meaningfully address this language from Aleynikov. While the government states that the “make something” language in Aleynikov is limiting, it simply dismisses it as an “inconsistent” reading. And while the majority admits that “[i]n Aleynikov, this court ruled that a connection between a confidential trading system and publicly traded securities was legally insufficient to make the system itself a product,” it goes on to dismiss this language. Maj. Op. at 251. Conversely, the majority writes, “we conclude that the term ‘related to,’ as used in the EEA’s nexus provision is intended to reach broadly rather than narrowly.” Maj. Op. at 248. Thus, it concludes, “the stolen code [was] related to *268traded securities[.]” Maj. Op. at 247. It reasons that, “the confidential code was valuable only in relation to the securities whose interstate trades it facilitated.” Maj. Op. at 248.
This construction of “related to” makes several missteps. First, it improperly creates a direct link between the stolen code and the securities, when in fact the stolen code was not only valuable in relation to the securities but was only a portion of the code, which composed the larger HFT System, which only then, was used to make the securities trades. As the government stated, “the whole point of stealing [the code], was to turn it into something the Tower could use ... the competing system.” Transcript at 1258. Thus, the code is far more properly characterized as being “related to” the System than to the ultimate securities, which is probably why both here and in Aleynikov, all parties assumed the HFT System was the product. The majority’s leap-frog connection also strains the legal demands of Aleynikov’s nexus requirement beyond this Circuit’s parameters. As stated previously, Aleynikov concluded that “related to” meant to “make something” that would be put into the stream of commerce. Under the majority’s interpretation, I can think of no example of a trade secret that would not be covered by the EEA.
This reading of “related to” is also so stretched that it affronts Supreme Court precedent. The Supreme Court has previously affirmed that “where ‘related to’ is used in legislation creating a discrete exception to a general rule, it may not be construed so expansively as to swallow the general rule.” Maj. Op. at 247 (citing N. Y. Conf. of Blue Cross & Blue Shield Plans v. Travelers Ins. Co., 514 U.S. 645, 655, 115 S.Ct. 1671, 131 L.Ed.2d 695 (1995)). Disregarding this rule, the majority continues “we need not delineate the outer limits of that reach because we easily conclude that SocGen’s HFT code related to publicly traded securities in such a way as to bring the theft of the HFT code within the EEA.” Maj. Op. at 248. In addition this expansive reading also goes against general principles of statutory construction obligating us to read Congress’s statutes narrowly. See Fed. Commc’ns Comm’n v. AT & T Inc., — U.S. -, 131 S.Ct. 1177, 1184, 179 L.Ed.2d 132 (2011) (“[C]onstruing statutory language is not merely an exercise in ascertaining ‘the outer limits of [a word’s] definitional possibilities’ .... ” (quoting Dolan v. U.S. Postal Serv., 546 U.S. 481, 486, 126 S.Ct. 1252, 163 L.Ed.2d 1079 (2006))).
Finally, the majority’s interpretation also offends the legislative history of the statute which narrowly construes “related to.” In both the House and Senate Reports on the EEA, examples of trade secrets that “relate to” products included production processes, bid estimates, production schedules, manufacturing specifications or fermentation processes. See S. Rep. 104-359, at 6, 8-9; H.R. Rep. 104-788, at 4, 8-9. All of these listed trade secrets bear a much closer relationship to a product than the relationship between the securities and the code asserted by the majority. In fact, all of these relationships adhere to the description of “related to” we gave in Aleynikov that the trade secret “make[s] something that [results in a product].” For example, applying Aleynikov’s definition, one can see how both manufacturing specifications and fermentation processes are clearly much more closely “related to” their products than parts of a code are “related to” the securities, because the prior examples “make” or give detailed instructions on how to create the product.
In contrast, Agrawal’s stolen code was only part of a larger computer code that *269helped make an HFT System, which in turn made the trades of the securities. The stolen code cannot be said to have directly made the securities. Supreme Court precedent, longstanding rules of statutory interpretation, and the legislative history all direct us to the same outcome: that the securities were not properly “related to” the code under the narrow construction of the EEA.
III. Conclusion
Aleynikov’s interpretation was narrow by its own terms, and in the meantime, before Congress clarified the language, Aleynikov’s interpretation governed. Regardless of Congress’s subsequent change to the statute, we are compelled to follow a decision of an earlier panel “unless it has been called into question by an intervening Supreme Court decision or by one of this Court sitting in banc,” United States v. Santiago, 268 F.3d 151, 154 (2d Cir.2001), or “unless and until its rationale is overruled, implicitly or expressly, by the Supreme Court, or this court in banc,” In re Sokolowski, 205 F.3d 532, 535 (2d Cir. 2000). See also Square D Co. v. Niagara Frontier Tariff Bureau, Inc., 476 U.S. 409, 424, 106 S.Ct. 1922, 90 L.Ed.2d 413 (1986) (quoting Burnet v. Coronado Oil & Gas Co., 285 U.S. 393, 406, 52 S.Ct. 443, 76 L.Ed. 815 (1932) (Brandéis, J., dissenting)) (stating that for statutory determinations, “it is more important that the applicable rule of law be settled than that it be settled right.... This is commonly true, even where the error is a matter of serious concern, provided correction can be had by legislation.” (internal quotation marks omitted)); id. at n. 34 (quoting NLRB v. Int’l Longshoremen’s Ass’n, 473 U.S. 61, 84, 105 S.Ct. 3045, 87 L.Ed.2d 47 (1985)) (“ ‘[W]e should follow the normal presumption of stare decisis in cases of statutory interpretation’ ”); Illinois Brick Co. v. Illinois, 431 U.S. 720, 736, 97 S.Ct. 2061, 52 L.Ed.2d 707 (1977) (“[W]e must bear in mind that considerations of stare decisis weigh heavily in the area of statutory construction, where Congress is free to change this Court’s interpretation of its legislation”). In order to circumvent Aleynikov, decided just months prior to oral argument in this case, the majority attempts to distinguish the present facts through mischaracterizations, while simultaneously stretching Aleynikov and disregarding the principle of stare decisis.
For this and other reasons explained above, I must dissent.
. The NSPA makes it a crime to “transport[ ], transmití], or transferí] in interstate or foreign commerce any goods, wares, merchandise, securities or money, of the value of $5,000 or more, knowing the same to have been stolen, or converted or taken by fraud.” 18 U.S.C. § 2314. In Aleynikov, we asked what qualified as "goods,” “wares,” or "merchandise,” 676 F.3d at 77, and determined that stolen code, in Aleynikov’s circumstances, did not qualify as a good, ware, or merchandise. In general, we found that stolen code could qualify under one of the three categories, but because Aleynikov had downloaded it to a server and then electronically transferred it from that server to his own computer, the digital code did not qualify as a good, ware, or merchandise. Here, a slight but important factual distinction exists: Agrawal transferred the code onto a hard drive at work and then printed out thousands of sheets of paper and then brought those papers home to New Jersey. Indictment at ¶ 11-12. Agrawal tries to apply Aleynikov to his circumstances — however as the majority points out, this "argument fails because it ignores Aleynikov’s emphasis on the format which intellectual property is taken.” Maj. Op. at 251. "Had Agrawal stolen the code in intangible form, as the defendant in Aleynikov had done, and only later copied it onto paper or some other tangible medium, that would not be enough ... to state an NSPA offense.” Id. at 252-53. This conscientious application of Aleynikov’s law to these particular facts is why I concur as to the majority’s discussion as to the NSPA and am, however, even more confounded by the majority’s misapplication of Aleynikov as to Agrawal's EEA count.
. Both Aleynikov and Agrawal were employed by their companies and pursued by competing companies because of their knowledge and expertise in the HFT System. Agrawal and Aleynikov also both stole sections of the HFT code, and shared it with their future employers. Both Agrawal and Aleynikov's indictments alleged under the EEA counts that the "trade secret” was the code for the System and the "product” was the HFT System. Indictment at ¶ 19; see Aleynikov, 676 F.3d at 75.
. When we decided these two issues in Aleynikov, we did it with the understanding that our construction of the EEA was narrow. See Aleynikov, 676 F.3d at 82. Judge Calabresi's concurrence asked Congress to clarify the EEA’s vague parameters. Id. at 83 (Calabresi, J. concurring) ("I wish to express hope that Congress will return to the issue and state, in appropriate language, what I believe they meant to make criminal in the EEA.”). Following Aleynikov, but prior to Agrawal’s conviction, Congress amended Section 1832(a) of Title 18, United States Code, to reverse our reading of the EEA. See Maj. Op. at 244 n. 7 (quoting Theft of Trade Secrets Clarification Act of 2012, Pub.L. No. 112-236, 126 Stat. 1627). However, Congress did not hint at — let alone explicitly state — that such an amendment was to apply retroactively. Thus, as there is a "well-established presumption against the retroactive application of legislation, including amendments[,]” this Court has an obligation withhold those amendments from taking effect retroactively. Velez v. Sanchez, 693 F.3d 308, 325-26 (2d Cir.2012).
Although I recognize that Congress’s new, more expansive amendment to the EEA might hold defendants like Aleynikov and Agrawal liable going forward under the EEA — that is of no application to this case. The majority concedes this point, see Maj. Op. at 237, but it wholly ignores Aleynikov's facts and reasoning as to the EEA in order to uphold a conviction of what it deems a "thief." Maj. Op. at 237.
. Moreover, Congress’s change to the statute only further demonstrates that Aleynikov’s outcome should govern this case or otherwise such changes would have been needless.
. In fact, the word "securities” is almost nowhere mentioned in the indictment. The word is used on the first page to state that "The Financial Institution engaged in among other financial activities, high-frequency trading in securities markets.” Indictment at ¶ 1. Additionally, in reference to the NSPA Count the indictment mentions that Agrawal transferred, "goods, wares, merchandise, securities, and money, of the value of $5,000 and more.” Id. at ¶21 (emphasis added). However, these two cursory mentions of the word “securities” do not undermine the fact that the government alleged all along that the HFT System was the product. In fact, that same NSPA section of the indictment identifies the HFT System as the product or "proprietary computer code.” Id.
. In its brief on appeal, the government highlighted the profitable nature and secret treat*266ment of SocGen’s HFT system, underscoring its product-like nature. For example it stated, “Both of these systems were made up of highly complicated computer code that had been written and refined over the course of years by employees of SocGen. ADP and DQS were highly profitable, generating well more than $10 million of profit for SocGen in each of the years 2007, 2008, and 2009. ADP and DQS cost close to $10 million to develop.” Appellee’s Br. at 8-9 (internal citations omitted).
. Even if we accept that the securities were the alleged "product,” we still cannot uphold the majority’s conclusion. In Aleynikov, we wrote, that "[bjecause the HFT system was not designed to enter or pass in commerce ... Aleynikov’s theft of source code relating to that system was not an offense under the EEA.” 676 F.3d at 82. This reading of "product” as something designed demands the product be the result of some amount of "physical labor or intellectual effort.” See Maj. Op. 246 n. 8 (citing to Webster's 3d New Int'l Dictionary 1810 (1986)). Here, SocGen’s securities, had not been labored over, invested in, developed, designed, or produced. Unlike the System which had been "written and refined over the course of years,” which had cost the company $10 million to develop it, and which had employee compliance manuals, password protections, code storage, special passwords and surveillance cameras installed to maintain it, SocGen had no hand in the making, maintenance or protection of the securities. The securities existed entirely outside of SocGen’s System. Thus, even applying the government’s newly invented securities-as-products theory we still cannot uphold Agrawal’s EEA conviction, where this Circuit’s caselaw cannot support a securities-as-product theory. To conclude otherwise renders the word "product” meaningless — as SocGen had no hand in any securities's creation.