Nathan Martin, J.D. University of Virginia School of Law, 2027.

The author thanks Zach Roubein (UVA Law ‘27) for his work on a prior, unpublished version of this Comment.

Introduction

NEW contours of First Amendment jurisprudence have frequently been defined and debated in connection with technological developments, and the application of the First Amendment to computer code is no exception. After the Ninth Circuit first considered First Amendment protections for encryption source code in Berstein v. U.S. Dep’t of Justice,[1] numerous circuits have also taken up the question of code’s First Amendment status to varying conclusions. Bernstein, for example, noted even as it upheld the First Amendment challenge, that some software “surely is not[]” expressive.[2] By contrast, the Sixth Circuit has taken a more expansive view, concluding that “computer source code . . . is protected by the First Amendment.”[3] Other circuits have also weighed in, and the divide on how to approach code under the First Amendment continues to persist.[4]

This longstanding struggle recently reemerged in Defense Distributed v. Attorney General of New Jersey.[5] Defense Distributed, a Texas-based company, produces and disseminates “digital firearms information” (“DFI”). DFI constitutes coded computer files ranging from text guides (.txt) and manuals (.pdf) to Computer Aided Design (“CAD”) and Computer Aided Manufacturing (“CAM”) files (.stl, .igs, .stp). These files feature technical instructions for the manufacture via 3D printers of functional firearms and firearm components. Members of the co-plaintiff Second Amendment Foundation wanted to receive these files as part of a broader political movement which advocates for self-defense rights and the decentralized manufacturing of firearms.[6]

Viewing the dissemination of files of this nature as a threat to public safety, the New Jersey legislature enacted N.J. Stat. Ann. § 2C:39-9(l)(1)-(2) (“New Jersey Statute”). This law criminalized the distribution “by any means, including the Internet, to a person in New Jersey who is not registered or licensed as a manufacturer . . . digital instructions in the form of computer-aided design files or other code or instructions . . . that may be used to program a three-dimensional printer.”[7] Defense Distributed challenged the regulation on a few bases but, pertinent to this piece, on First Amendment grounds.[8]

In upholding the New Jersey Statute, the Third Circuit added to the existing divide. The panel held that because CAD and CAM files directly execute physical manufacturing tasks (and are thus “functional”), plaintiffs must affirmatively plead specific expressive intent to trigger First Amendment protection. Because the Third Circuit felt Defense Distributed failed to properly plead, it affirmed dismissal.[9] While this context-specific analysis is reminiscent of a number of the decisions addressing the issue from the earliest days of the analysis of code as speech,[10] this decision is the first to adopt such an approach after the Supreme Court’s Sorrell decision, which this Comment contends settles the debate on code in favor of its coverage under the First Amendment without any context-specific test.[11]

Part I of this Comment briefly examines the Third Circuit’s ruling and the test it endorses. Part II considers the court’s crucial errors: (1) its misunderstanding of the impact of Sorrell on code’s status as speech, and (2) its refusal to recognize that the distribution and receipt of functional code, when tied to political advocacy, constitutes core political conduct entitled to First Amendment protection even under its own test. Finally, Part III details the proper procedure for First Amendment analyses of code, which properly shifts the question to how much protection code enjoys, not if protection exists at all.

 I. The Third Circuit’s Approach

In attempting to find a middle ground between full First Amendment protection and complete exclusion, the Third Circuit’s “fact-based and context-specific” inquiry purports to distinguish code designed to express an idea from what it labels “purely functional code with no expressive purpose, use, or intent.”[12] Adopting a requirement that plaintiffs affirmatively plead the expressive nature of any computer code in order to trigger First Amendment protections primarily rests on a finding that code is not necessarily speech.[13] If something is “speech,” it is “presumptively protect[ed] . . . against government interference.”[14] In other words, “expressiveness” is not a freestanding pleading requirement where the subject of the regulation is actual speech, but if something is not necessarily speech, an examination of its expressiveness is central in determining First Amendment protection.[15] As such, a threshold issue in the propriety of the Third Circuit’s test is whether it was correct in concluding that code is not speech.

The framework articulated by the court requires judges to evaluate the technical nature of the code, the context of its use, the intended recipient (such as “human-to-machine” versus “programmer-to-human” communication), and the purposes it serves.[16] In particular, the analysis of the intended recipient suggests possible approval of the distinction between communication to a person versus communication to a machine which the Second Circuit emphasized in its Universal City Studios, Inc. v. Corley opinion.[17] Corley said that where a programmer is communicating through code to the computer that such code is “never protected” under the First Amendment.[18] That the Third Circuit discussed that portion of Corley specifically and included the nature of the recipient in its test suggests it might view such a rule with approval.[19] However, as the Defense Distributed court itself said, it did not “go beyond recognition” of the framework it laid out because the plaintiff failed to plead sufficient factual matter to analyze protection under the test, so exactly how much weight the individual factors are given and how the test operates is not articulated.[20]

 II. Information as Speech: Sorrell, Machine Communication, and Political Expression

 In declining to analyze the code at issue under the First Amendment on the ground that the plaintiffs failed to adequately plead expressive intent, the Third Circuit made two fundamental errors.[21] The first is that threshold coverage under the First Amendment in this case is governed by Sorrell, which establishes that creating and disseminating information (which includes instructions)[22] falls squarely within the First Amendment’s definition of speech.[23] While the Third Circuit might legitimately conclude that restrictions on code of this nature are proper under First Amendment scrutiny in light of the asserted governmental interests, it cannot properly strip the medium of First Amendment coverage altogether by labeling it “purely functional.”[24] The second error is that even under the Third Circuit’s own framework, code that is expressly pleaded to function as political advocacy by promoting the decentralized manufacture of firearms and the Second Amendment rests at the very core of the First Amendment as a form of political expression, and in this case it was so pleaded.[25]

To the first point, the Supreme Court in Sorrell explicitly held that “the creation and dissemination of information are speech within the meaning of the First Amendment.”[26] Thus, these activities, by their very nature, trigger threshold First Amendment coverage. Computer code, at its core, consists of a structured set of instructions designed to accomplish a specific task. Because instructions are indisputably composed of information—without which they could not instruct a human or a machine how to do anything—Sorrell dictates that their dissemination constitutes protected speech under the First Amendment.

In an attempt to sidestep Sorrell, the Third Circuit dismissed its discussion of the creation and dissemination of information as a mere “dictum,” asserting that precedent “do[es] not support such a broad reading of First Amendment protection for ‘information’ writ large.”[27] However, the overwhelming majority of federal circuits evaluating Sorrell have treated its holding that the creation and dissemination of information constitutes speech as settled, controlling law. For example, the Second Circuit relied directly on Sorrell in holding that “the Supreme Court has made clear[] [that] ‘the creation and dissemination of information are speech within the meaning of the First Amendment.’”[28] Every other circuit to consider the issue has recognized the same,[29] and the Third Circuit stands alone in its crimped view of Sorrell and its implications. Unsurprisingly, the only other circuit to address the applicability of the First Amendment to computer code post-Sorrell easily concluded that software is protected speech (at least where comprehensible by a human), with the district court in that case explicitly noting that sharing computer code may implicate the informational speech interests recognized in Sorrell.[30]

In promulgating its multi-factor test, the Third Circuit relied on pre-Sorrell authority to conclude that context and factual circumstances must dictate whether code is expressive or merely functional.[31] While this reliance was erroneous and Sorrell dispenses with much of the Third Circuit’s reasoning on its face, the court’s test and other caselaw in the area implicate a second question: even if information is generally speech, does a listener or intended recipient have to be human for communications to constitute protected speech under the First Amendment? Both Corley and Defense Distributed suggest that the identity of the listener is at least an important factor, if not dispositive.[32] If the listener does not control protection, Sorrell directs that all code is speech because—even in those cases where it is incomprehensible to humans or is otherwise not designed for them to read—it conveys information to computers. If the listener does control protection, then a more complicated inquiry could be necessary when considering code as speech. Interestingly, the only other post-Sorrell code-as-speech case was careful to address only code which may be understood by humans in its discussion of whether code is speech, suggesting that even under Sorrell a focus on information’s utility to humans may be relevant.[33] As the artificial intelligence industry continues to boom in America, the answer to this question will only grow in importance.

Sorrell, however, does seem to shed light on the relevance of the listener. The Supreme Court observed that “[a]n individual's right to speak is implicated when information he or she possesses is subjected to ‘restraints on the way in which the information might be used’ or disseminated.”[34] A requirement of comprehensibility to humans or that the recipient be human would necessarily be such a restraint. Furthermore, recognizing that a speaker’s First Amendment interest in conveying information does not depend on whether the ultimate recipient is a machine or a human aligns directly within the First Amendment's core purpose of facilitating free speech. Sorrell alone resolves the ongoing debate among the circuits as to when code counts as speech: because code is information, its dissemination is presumptively speech under the First Amendment, unless the government can demonstrate that the code at issue falls into an established exception to First Amendment protection.[35]

Turning to the panel’s second fundamental error, even if First Amendment protection turned on contextual expressive intent, the Third Circuit ignored the explicitly political character of the file dissemination at issue in this case. Defense Distributed’s dissemination of DFI is an act of political advocacy designed to promote the decentralized manufacturing of firearms, challenge state monopolies on the sale of the firearms, and facilitate self-reliance on the Second Amendment. In fact, Defense Distributed said as much in its initial pleadings, noting that it was publishing this code “in furtherance of its mission to promote firearms knowledge and possession.”[36] Citizens United v. FEC affirmed that political speech occupies the highest tier of First Amendment protection because it is an essential component of self-governance.[37] When a company like Defense Distributed disseminates functional instructions as the primary vehicle for promoting its political goals surrounding gun ownership, it is engaged in political speech. By writing off the files as non-expressive and deeming Defense Distributed’s description of their own purpose in publishing the code as “conclusory,” the Third Circuit applied an unreasonably high pleading standard to a plaintiff’s description of their own intent.

The errors described above contravene controlling precedent, deepen an already existing divide in the circuit courts, and represent an unjustified restriction on First Amendment protections. Whatever the ultimate outcome of the case should have been, the reasoning adopted by the Third Circuit is erroneous and should be corrected.

III. Analyzing Code as Speech Post-Sorrell

Presumptive First Amendment protection, which code properly enjoys under Sorrell, does not confer immunity from state regulation; it merely subjects governmental restraints to judicial oversight in the form of tiers of scrutiny.[38] While improper to strip code of protection at the outset, courts have long provided differing levels of protection to different types of speech under the First Amendment as they have balanced private and public interests.[39] Courts should reject the Third Circuit’s approach in this case and should instead apply the longstanding tools of First Amendment analysis in order to balance the governmental interest in the regulation of code with the fundamental right of free speech.

Under the proper framework, courts must first recognize that the First Amendment presumptively protects all computer code, whether it is human-readable source code or machine-executable CAD and CAM files, as instructional and informational data. As the Supreme Court established in Sorrell, the creation and dissemination of information are speech that trigger First Amendment review.[40] Under this baseline rule, a plaintiff like Defense Distributed is not required to affirmatively plead to the expressive nature of its files at the Rule 12(b)(6) stage to invoke threshold protection—Sorrell already made that legal determination in treating information as First Amendment speech.[41] The burden remains precisely where First Amendment jurisprudence has placed it: the state must justify its restriction under the appropriate tier of scrutiny.[42] Preserving threshold protection avoids transforming constitutional litigation into a pleading contest and ensures that digital communication receives the same initial treatment as other recognized categories of speech.

Once threshold coverage is established, public safety concerns like those raised by New Jersey in this case are properly evaluated during the scrutiny and tailoring analysis where the government justifies its restriction,[43] rather than presented to justify exclusion at the pleading stage. What the proper tier of scrutiny is and what kinds of state interests will justify restricting “functional” code are questions beyond the scope of this Comment, but are instead questions best left to the courts in the first instance.

Preserving threshold protection for technical instructions provides doctrinal stability for emerging technologies far beyond DFI. As software becomes increasingly integrated into every aspect of modern life, lower courts will inevitably face challenges involving artificial intelligence and code as a general matter.[44] The question of whether code is speech is an example of one such challenge, and it is a question that courts have grappled with since it came about with the advent and widespread use of coding. Whether or not code is inherently speech, even where purely functional, was a difficult issue but is now one which has a clear answer in light of Sorrell: it is. Different methods of coding are described as “languages” for a reason—they represent different methods of communication. Today, this issue also implicates another question not yet squarely decided but is of increasing importance as AI proliferates: to what extent is speech directed to a machine “speech” under the First Amendment?[45] The case of Defense Distributed, erroneous as its reasoning may be, provides the Supreme Court with an opportunity to address both issues. Defense Distributed filed for a writ of certiorari on August 10, 2026.[46] The Supreme Court should take this opportunity to resolve the longstanding circuit split that Defense Distributed adds to and clarify the reach of Sorrell in doing so.

[1]          176 F.3d 1132 (9th Cir. 1999), reh’g granted, opinion withdrawn, 192 F.3d 1308 (9th Cir. 1999) (rehearing was ultimately mooted, but the analysis is still relevant and oft-cited, including in the case discussed in this Comment).

[2]          176 F.3d at 1145.

[3]          Junger v. Daley, 209 F.3d 481, 485 (6th Cir. 2000).

[4]          See, e.g., Universal City Studios v. Corley, 273 F.3d 429 (2d Cir. 2001); Green v. U.S. Dep’t of Just., 54 F.4th 738 (D.C. Cir. 2022).

[5]          Defense Distributed v. Att’y Gen. of N.J., 167 F.4th 65 (3d Cir. 2026) [hereinafter Defense Distributed].

[6]          Id. at 70-71.

[7]          Read in full, N.J. Stat. Ann. § 2C:39-9(l)(1)-(2) makes it a crime for:

“(1) a person who is not registered or licensed to do so as a manufacturer as provided in chapter 58 of Title 2C of the New Jersey Statutes, to use a three-dimensional printer or similar device to manufacture or produce a firearm, firearm receiver, magazine, or firearm component; or (2) a person to distribute by any means, including the Internet, to a person in New Jersey who is not registered or licensed as a manufacturer as provided in chapter 58 of Title 2C of the New Jersey Statutes, digital instructions in the form of computer-aided design files or other code or instructions stored and displayed in electronic format as a digital model that may be used to program a three-dimensional printer to manufacture or produce a firearm, firearm receiver, magazine, or firearm component.”

[8]          Defense Distributed, 167 F.4th at 71.

[9]          Id. at 84 (“These allegations fall short of plausibly alleging that Defense Distributed’s code triggers First Amendment coverage.”).

[10]          See, e.g, Bernstein v. U.S. Dep’t of Just., 176 F.3d 1132 (9th Cir. 1999), reh’g granted, opinion withdrawn, 192 F.3d 1308 (9th Cir. 1999).

[11]          Sorrell v. IMS Health Inc., 564 U.S. 552 (2011).

[12]          Defense Distributed, 167 F.4th at 82 (“To invoke the protections of the First Amendment, the proponent must show that the particular use of the code burdened by a regulation involves the expression or communication of ideas in a way that implicates the First Amendment.”)

[13]          That this is a pleading requirement is clear from the fact that the affirmed dismissal was on 12(b)(6) grounds. See Defense Distributed v. Platkin, 697 F. Supp. 3d 241, 255 (D.N.J. 2023) aff’d sub nom. Defense Distributed, 167 F.4th 65 (3d. Cir. 2026).

[14]          United States v. Alvarez, 617 F.3d 1198, 1205 (9th Cir. 2010) aff’d 567 U.S. 709 (2012).

[15]          See, e.g., Bail Project, Inc. v. Comm’r, Ind. Dep’t of Ins. 76 F.4th 569, 575 (7th Cir. 2023) (noting that the “First Amendment protects conduct only when it is ‘inherently expressive.’”) (quoting Rumsfeld v. F. for Acad. & Inst. Rts., Inc., 547 U.S. 47, 66 (2006)).

[16]          Defense Distributed, 167 F. 4th at 85.

[17]          273 F.3d 429, 449 (2d Cir. 2001).

[18]          Id.

[19]          Defense Distributed, 167 F.4th at 81, 84.

[20]          Id. at 84.

[21]          Id. at 85 (“. . . the complaint does not include sufficient allegations to assess whether that code is covered by the First Amendment.”).

[22]          See, e.g., Instruction, sense 3.a, Oxford English Dictionary,  https://www.oed.com/dictionary/instruction_n?tab=meaning_and_use#387233 (last visited Sept. 9, 2026) (“That which is taught; knowledge); Instructions, Cambridge Dictionary, https://dictionary.cambridge.org/us/dictionary/english/instructions (last visited Sept. 9, 2026) (“advice and information about how to do or use something . . .”) (emphasis added).

[23]          564 U.S 552, 570 (2011) (acknowledging that “[t]his Court has held that the creation and dissemination of information are speech within the meaning of the First Amendment.”); see also Universal City Studios, Inc. v. Corley, 273 F.3d at 446, 447-48 (computer code programs “are essentially instructions to a computer,” and “[t]he information conveyed by most ‘instructions’ is how to perform a task.”).

[24]          Defense Distributed, 167 F.4th at 82.

[25]          See Complaint at 1-2, Defense Distributed v. Grewal, No. 3:21-cv-09867 (D.N.J. July 29, 2018), decided sub nom. Defense Distributed v. Platkin, 697 F. Supp. 3d 241 (D.N.J. 2023), ECF No. 1 (Defense Distributed publishes code “in furtherance of its mission to promote firearms knowledge and possession,” as the organization is designed to operate “for the purpose of defending the civil liberty of popular access to arms . . . through facilitating access to, and the collaborative production of, information and knowledge related to the production of arms”); Morse v. Frederick, 551 U.S. 393, 403 (2007) (“Political speech, of course, is ‘at the core of what the First Amendment is designed to protect’”) (quoting Virginia v. Black, 538 U.S. 343, 365 (2003) (plurality opinion)); see also Burson v. Freeman, 504 U.S. 191, 217 (1992) (Stevens, J., dissenting) (direct regulation of political expression “implicates a core concern of the First Amendment.”).

[26]          564 U.S. at 570.

[27]          Defense Distributed, 167 F.4th at 82-83.

[28]          Upsolve, Inc. v. James, 155 F.4th 133, 141 (2d Cir. 2025) (quoting Sorrell, 564 U.S. at 570).

[29]          See Pac. Coast Horseshoeing Sch., Inc. v. Kirchmeyer, 961 F.3d 1062, 1072 (9th Cir. 2020); VoteAmerica v. Schwab, 121 F.4th 822, 837 (10th Cir. 2024); Sharpe v. Winterville Police Dep’t, 59 F.4th 674, 680-81 (4th Cir. 2023); Vizaline, L.L.C. v. Tracy, 949 F.3d 927, 928 (5th Cir. 2020); Yoder v. Bowen, 146 F.4th 516, 523 (6th Cir. 2025); Richwine v. Matuszak, 148 F.4th 942, 954 (7th Cir. 2025); Animal Legal Def. Fund v. Vaught, 8 F.4th 714, 718 (8th Cir. 2021); Price v. Garland, 45 F.4th 1059, 1067 (D.C. Cir. 2022); NetChoice, LLC v. Att’y Gen. of Fla., 34 F.4th 1196, 1216-17 (11th Cir. 2022) vacated sub nom. on other grounds by Moody v. NetChoice, LLC, 603 U.S. 707 (2024).

[30]          See Green v. United States Dep’t of Just., 54 F.4th 738, 745 (D.C. Cir. 2022) (concluding that code which is comprehensible to a human is speech with “no trouble”); Green v. United States Dep’t of Just., 392 F. Supp. 3d 68, 87 (D.D.C. 2019) (sharing computer code with others “through books, articles, or otherwise . . .arguably implicates” the First Amendment interest recognized by Sorrell with respect to information distribution).

[31]          Defense Distributed, 167 F.4th at 79-83.

[32]          Id. at 84 (factor); Universal City Studios v. Corley, 273 F.3d 429, 449 (2d. Cir. 2001) (dispositive).

[33]          Green, 54 F.4th at 745.

[34]          564 U.S 552, 568 (2011) (quoting Seattle Times Co. v. Rhinehart, 467 U.S. 20, 32 (1984)) (emphasis added).

[35]          Corley, for example, noted that some instructions have been considered to fall “outside the First Amendment,” not because “the speech took the form of instructions, but rather” because “the instructions counseled the listener how to commit illegal acts.” 273 F.3d at 447 n. 18.

[36]          See Complaint at 1-2, Defense Distributed v. Grewal, No. 3:21-cv-09867 (D.N.J. July 29, 2018), ECF No. 1

[37]          558 U.S. 310, 339 (2010).

[38]          See Frederick Schauer, The Boundaries of the First Amendment: A Preliminary Exploration of Constitutional Salience, 117 Harv. L. Rev. 1765, 1767–69 (2004) (explaining how the boundaries of First Amendment coverage determine whether a regulatory domain implicates constitutional scrutiny at all, rather than whether a covered activity ultimately survives judicial review).

[39]          See, e.g., Uber Techs., Inc. v. City of Seattle, 168 F.4th 1202, 1215-1217 (9th Cir. 2026) (acknowledging that commercial speech enjoys less protection under the First Amendment and applying a lower level of scrutiny).

[40]          564 U.S. 552, 570 (2011) (“This Court has held that the creation and dissemination of information are speech within the meaning of the First Amendment.”).

[41]          See Connick v. Myers, 461 U.S. 138, 148 n.7 (1983) (“The inquiry into the protected status of speech is one of law . . . .”).

[42]          See United States v. Playboy Ent. Grp., Inc., 529 U.S. 803, 816 (2000) (placing the burden on the government was the “proper approach” because “[w]hen the Government restricts speech, the Government bears the burden of proving the constitutionality of its actions.”) (citing Greater New Orleans Broad. Ass’n., Inc. v. United States, 527 U.S. 173, 183 (1999)).

[43]          See, e.g., Garcia v. Cnty. of Alameda, 150 F.4th 1224, 1233 (9th Cir. 2025) (noting that “[p]ublic safety is certainly a compelling interest” in considering if that justification allows a regulation to withstand constitutional scrutiny where it burdens First Amendment rights) (citing Meinecke v. City of Seattle, 99 F.4th 514, 525 (9th Cir. 2024)).

[44]          See Peter Salib, AI Outputs are Not Protected Speech, 102 Wash. U. L. Rev. 83, 124 (2024) (arguing that AI outputs are not speech, carefully drawing a distinction between regulating “an AI system’s production of certain outputs” with “a user’s request for or adoption of them” when asking the question of when regulating outputs constitutes a regulation of a person’s speech, which is consistent with a recognition of the constitutional relevance of the listener argued for in this comment) (emphases in original).

[45]          Id.

[46]          See Defense Distributed, 176 F.3d 1132 (3d Cir. 2026), petition for cert. filed, 2026 WL 2395771 (U.S. Aug. 10, 2026) (No. 26-184).