Does the First Amendment protect code from the government’s regulation? More specifically, does the Free Speech Clause protect the digital files that enable 3-D printers to manufacture “ghost guns” that cannot be traced by law enforcement (because they have no serial number)?
In a short and snappy decision earlier this year, the Third Circuit answered: “it depends.” The court offered an analytical approach that may prove attractive to other courts deciding whether and when the First Amendment protects code from governmental regulation—including, but by no means limited to, the code used to manufacture ghost guns.
Human programmers create code, which can then be packaged into software that instructs a machine to perform tasks. Code is everywhere. It powers search engines, appliances, autonomous vehicles, browsers, word processors, and countless other technologies. Sometimes code is directed not only to machines, but also to human beings—users of those technologies or other programmers. Most, maybe all, code is functional. Some, but not all, code is expressive.
Under the Third Circuit’s approach, that distinction matters at the threshold: “[C]omputer code can be covered by the First Amendment. But we also hold that coverage cannot be assumed because code is inherently functional. 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. Purely functional code with no expressive purpose, use, or intent is simply not covered by the First Amendment.”
The case: Defense Distributed v. Attorney General of New Jersey, 167 F.4th 65 (3d Cir. 2026)
New Jersey enacted a criminal law prohibiting unlicensed persons from using 3-D printers to manufacture certain firearms or firearm components. More important for First Amendment purposes, it also prohibited any person from distributing—to New Jersey residents “not registered or licensed” as firearms manufacturers—“digital instructions in the form of computer-aided design files or other code or instructions stored and displayed in electronic format” that may be used to program a 3-D printer to produce a firearm or firearm component.
Defense Distributed challenged the law as violating its rights under the Free Speech Clause. According to its complaint, Defense Distributed develops and publishes what it calls “digital firearms information,” through digital files that included Computer-Aided Design (CAD) and Computer-Aided Manufacturing (CAM) files, as well as .stl, .igs, .sldprt, .skp, .stp, .txt, and .pdf files. The files were by no means alike. For example, some concerned firearm components and their assembly, while others involved plain-text materials discussing federal firearms statutes.
The Third Circuit did not decide that the challenger’s files are—or are not—speech covered by the First Amendment. Instead, the court affirmed dismissal of the complaint with prejudice because the complaint did not allege enough facts to permit the court to determine 1) which of its digital files were actually regulated by the statute; nor 2) whether the regulated files were used for expressive rather than purely functional purposes, when the court held this inquiry to be key to determining whether the code is even covered by the First Amendment.
In so holding, however, the Third Circuit offered plenty of guidance for litigation yet to come.
The Third Circuit’s opinion drew two critical distinctions
The Third Circuit first emphasized the fundamental distinction between First Amendment coverage and First Amendment protection (for more on this distinction, see here and here). The “coverage” inquiry asks whether the regulated activity is sufficiently expressive to trigger Free Speech Clause analysis at all. Only after that threshold has been crossed does a court then engage the separate question of whether the government’s regulation survives the relevant First Amendment doctrinal analysis—in other words, whether speech that is “covered” by the First Amendment is nevertheless “protected” from the government’s regulation. The Third Circuit expressly left the latter question unresolved, focusing only on whether and when code is covered by the First Amendment.
The Third Circuit then drew a second distinction between code’s expressive and functional characteristics, emphasizing that code is not even covered by the First Amendment absent some showing of its “expressive purpose, use, or intent.” This requires an inquiry into the technical nature of the code: how it’s used; who (or what) is its intended recipient; what’s its purpose; and what, if anything, it communicates. In other words: what is this particular code doing, and with whom?
Who’s reading the code: a machine or a human?
The court’s focus on the recipient is especially important, in my view. The Third Circuit drew on the Second Circuit’s earlier decision in Universal City Studios, Inc. v. Corley that distinguished programmer-to-programmer, programmer-to-user, and programmer-to-computer interactions. A human programmer may use code to communicate with another human programmer. A programmer may use code to communicate with a human user. Or the programmer may use code for communicating only with a machine.
To be sure, code does not become speech based on any interaction with a human. A user may have to click a button or otherwise take a small action to deploy software—and this, by itself, doesn’t turn the software’s underlying instructions to a machine into expression. The more important question is whether the code itself is being read, studied, discussed, modified, taught, or otherwise used by humans in human discourse—or whether its value lies instead in getting a machine to make something happen.
Why Defense Distributed lost
According to the court, Defense Distributed’s complaint suffered from two related pleading failures.
First, the complaint did not adequately identify which of its numerous digital files were actually restricted by New Jersey’s law. The statute regulated digital instructions that “may be used to program” a 3-D printer to manufacture firearms or components. A plain-text file discussing the National Firearms Act, for example, did not fit that description. Yet the complaint did not sufficiently identify which files did.
Second, even assuming that some of its files were regulated by the statute, Defense Distributed did not provide enough technical or factual detail for the court to evaluate those files’ expressive character, if any. Defense Distributed identified CAD, CAM, and other file formats, but it did not adequately explain what each file contained, how each worked, whether it was ready for insertion into a 3-D printer, whether it required substantial human manipulation, how users interacted with it, or what ideas it communicated. The court was unwilling to treat all of that material as a single constitutional category merely because the plaintiffs called it “digital firearms information.”
The Third Circuit therefore had “no occasion” to decide whether the CAD or CAM files at issue deserved First Amendment coverage. The problem was that Defense Distributed had not built a sufficiently concrete factual record about what the files actually were and what people actually did with them.
What next?
The lesson for future litigation (at least under the Third Circuit’s approach) is straightforward: explain the technology. What does the file do? How does it work? Does the file execute automatically? Does it require conversion, interpretation, or substantial human intervention?
Then ask who reads the file and why. Engineers? Programmers? Students? Researchers? Hobbyists? Collaborators? The stronger case for coverage is one in which humans routinely inspect, study, modify, compare, critique, annotate, or discuss the files. Revision histories, comments, educational materials, peer review, technical discussions, and collaborative design environments could all help establish that the files engage human communication.
Ask, too, about the code’s purposes: “to perform a function, to express an idea, or some combination thereof . . . and what, if anything, the code communicates.”
A First Amendment challenge to the regulation of digital files should thus identify which files are actually regulated, what each file contains, how each file works, how humans interact with it, and what ideas it communicates. The government, in turn, should emphasize the extent to which the regulated code directly produces machine execution, operates automatically, or has little independent communicative function.
A prohibition on using a 3-D printer to manufacture a firearm is conceptually different from a prohibition on distributing information to humans about firearms or manufacturing. A statute aimed at machine-ready manufacturing files has a stronger constitutional footing when it identifies that category specifically rather than sweeping in manuals, historical material, commentary, diagrams, or other material that cannot itself program a printer.
In adopting its context-sensitive flexible standard, the Third Circuit expressly declined to adopt the Sixth Circuit’s categorical bright-line rule in Junger v. Daley that “source code” (which represents computer instructions in a specialized programming language readable by humans) is protected by the First Amendment (in contrast to “object code” that represents computer instructions in a sequence of binary digits directly executed by a machine). For the Third Circuit, the mere possibility that code could be used for expressive purposes is not enough. The relevant question is how the code at issue is actually being used—or is intended to be used—in the context of the challenged regulation.
The Third Circuit’s decision directly engaged a decades-long scholarly debate about whether and when “code is speech.” Its opinion was transparently influenced by Orin Kerr’s pragmatic warning against categorical coverage: if every program becomes speech because programs contain information or are written in programming languages, courts risk constitutionalizing huge areas of ordinary technological regulation.
Some code is expressive (and thus covered by the First Amendment). Probably not most code.
This is, in my view, the most useful takeaway from the Third Circuit’s decision. Code is not a single constitutional category. Some code is plainly used to communicate ideas among humans: academic research, cryptographic publications, open-source discussions, demonstrations of algorithms, instructional materials, and code published as part of technical debate. Other code exists primarily to execute a task: software controlling a pump, air conditioner, or generator are straightforward examples.
In my view, a CAD or CAM file sent to a printer to manufacture a gun looks much more like the latter, at least when its relevant function is simply to cause the printer to manufacture the object. That doesn’t mean the file can never be used expressively. It means that a challenger’s First Amendment argument needs to be grounded in the particular use at issue rather than simply on the abstract proposition that humans wrote the file or that the file contains information of some sort.
The Third Circuit’s decision is therefore less a final answer about “ghost-gun code” than a framework for asking certain questions. It did not hold that Defense Distributed’s files are outside the First Amendment. It did not hold that the First Amendment generally doesn’t cover CAD files or CAM files. It did not decide what level of scrutiny would apply to the government’s regulation of code determined to be expressive and thus covered by the First Amendment. And it did not say that every file capable of producing a gun is necessarily functional rather than expressive.
What it did say is that courts cannot answer those questions from labels alone. They need a factual record showing what the files are, what they do, who interacts with them, for what purpose, and what they communicate—if anything. Purely functional code with no expressive purpose, use, or intent is outside First Amendment coverage. Code demonstrated to be covered expression moves to a different stage of analysis.
That approach is useful well beyond guns manufactured by 3-D printers. The same line can do valuable work in a wide range of disputes involving algorithms, artificial intelligence, autonomous systems, and other technologies that both communicate information to humans and direct machines to make things happen. The Third Circuit’s answer is not that code is speech, or that code is conduct. It is that context matters—which, for code, means paying close attention to who (or what) is on the other end.
