Updated 2026-09-25
from the news desk
Can AI have constitutional rights? What U.S. scholarship says
Peter N. Salib argues AI lacks constitutional rights for First Amendment purposes, while Jared Schroeder examines whether the Press Clause could protect AI publishers. Margaret Hu’s Bill-of-Rights article focuses instead on constitutional…
The scholarship discussed here does not identify a U.S. constitutional right that AI currently holds. Peter N. Salib argues that, for First Amendment purposes, “AIs themselves lack constitutional rights, so their outputs cannot be their own protected speech.” Jared Schroeder treats Press Clause protection for AI publishers as a live doctrinal question, while Margaret Hu focuses on constitutional harms to people from biometric AI systems rather than on AI as the rights-holder itself.
What the current scholarship actually shows
Salib’s Washington University Law Review article addresses the most developed constitutional argument in this set of materials. Salib argues that current First Amendment doctrine does not treat AI as a speaker with its own constitutional rights, and he builds that claim around the status of AI outputs.
Schroeder’s William & Mary Bill of Rights Journal abstract shows that the question has not disappeared. Schroeder says his article focuses on whether the Press Clause, the First Amendment protection for freedom of the press, protects AI publishers not as extensions of human publishers but as non-human entities that gather and communicate information to audiences.
Hu’s article on biometrics and an AI Bill of Rights points in a different direction. Hu contends that any informed discussion of an AI Bill of Rights must grapple with biometric data collection and its integration into emerging AI systems. Hu also states that high-risk biometric AI systems can pose unprecedented challenges to criminal procedure rights and create broader constitutional threats.
One reading is that the scholarship is splitting the problem in two. One line asks whether AI can ever be a constitutional rights-holder, meaning the entity that holds the right itself. Another line asks how constitutional law protects people when AI systems are used in ways that affect speech, policing, border security, immigration enforcement, or other state activity that Hu lists.
Why the question is still live in First Amendment scholarship
Salib does not describe the opposing view as frivolous. He notes that generative systems produce language, images, and sound that look unusually close to human expression, and he writes: “Generative AI outputs are so remarkably speech-like that, it seems, they must be someone’s protected speech.”
Salib reports that Toni M. Massaro and Helen Norton contend nothing in existing law rules out the possibility that sufficiently complex AIs are already entitled to their own First Amendment rights. Salib also reports that other scholars place any protection elsewhere, arguing that AI outputs may instead be the speech of programmers, users, or corporate owners.
That disagreement matters because it keeps the constitutional question alive even for readers who reject AI personhood. If outputs are someone’s protected speech, the rights may belong to a human or corporation rather than to the system. A related explainer on free speech rights covers that narrower dispute.
Schroeder’s abstract keeps the live issue focused on the press as well as speech. Schroeder says AI publishers represent a new generation of publishers and raise substantial legal questions in areas including defamation, intellectual property, and privacy law. He then narrows his own project to the Press Clause question, asking whether AI communicators should receive protection as non-human entities that gather and communicate information.
One steelman reading of this pro-rights side is straightforward. If constitutional doctrine protects some communicative roles and institutions because they help sustain the flow of ideas in a democratic society, then an AI system that gathers and communicates information may become harder to classify as a mere passive tool. Schroeder’s abstract does not state his final answer, but it shows why the question keeps returning.
Why Salib says AI lacks its own constitutional rights
Salib’s own answer is direct: “AIs themselves lack constitutional rights, so their outputs cannot be their own protected speech.”
He gives several reasons. First, Salib argues that AI outputs are not the speech of creators or users because generative systems are designed to produce new material across a vast range of topics, not to convey a specific pre-set human message. In Salib’s account, users prompt these systems precisely because they want an answer, poem, joke, or other response that is not just a copy of their own words.
Second, Salib says the technical structure of modern generative AI weakens the claim that a human speaker is really doing the communicating. He argues that these systems self-program through training, that their resulting code is highly uninterpretable, and that humans therefore cannot directly specify or predict outputs with meaningful specificity.
Third, Salib rejects the move from human rights to corporate rights as a solution. He argues that corporate First Amendment rights are derivative of the rights of human constituents. On that view, a corporation does not turn unprotected output into protected speech just by owning the system that produced it.
That is why Salib states the rule this way: “Absent a communication from a protected speaker, there is no protected speech.”
Salib still does not treat AI regulation as constitutionally free of limits. He argues that AI outputs may receive indirect First Amendment protection when the law burdens the interests of human listeners or the human use of AI as a tool for composing or performing speech. In his examples, AI can function as an editor, proofreader, brainstormer, or medium for speech without becoming a constitutional speaker itself.
That distinction is also where the broader idea of electronic personhood becomes relevant. The personhood question asks whether law should ever recognize AI as a legal person for some purposes. Salib’s article, by contrast, argues that present First Amendment doctrine does not do so for AI speech.
Beyond AI rights: the Bill of Rights questions about people
Hu’s article shows why the constitutional debate is broader than AI’s own status. Hu says biometric AI systems serve governmental purposes including policing, border security and immigration enforcement, biometric cyberintelligence, and biometric-enabled warfare. Hu adds that these systems are increasingly categorized as high-risk when used in ways that may affect fundamental constitutional rights and human rights.
Hu further concludes that high-risk biometric AI systems, including facial recognition identification, can pose unprecedented challenges to criminal procedure rights. Her article therefore frames an AI Bill of Rights as a response to constitutional threats created by AI deployment, not as a charter of rights for AI itself.
Set beside Salib and Schroeder, Hu’s framing suggests a useful distinction. A system can be central to constitutional law without being a constitutional subject in its own right. One live question asks whether AI can ever hold rights. Another asks whether people affected by AI in policing, border security, immigration enforcement, or related settings can assert their own constitutional protections against those uses.
That is also why the constitutional question is narrower than the moral one. A reader asking whether AI should ever count as a rights-bearer in ethics can start with our explainer on whether AI should have rights. The legal materials here show a different point: moral concern, doctrinal personhood, and constitutional protection do not automatically travel together.
On this record, then, the safest answer is limited. Salib argues no for AI’s own current First Amendment rights. Schroeder’s abstract shows that press-based arguments for non-human AI communicators remain a live scholarly question. Hu’s article indicates that many of the most concrete Bill of Rights fights may concern the constitutional rights of humans affected by AI systems rather than any constitutional rights of the systems themselves.
Frequently asked questions
Does any source here say AI already has U.S. constitutional rights?
No. Salib argues that AI lacks constitutional rights for First Amendment purposes. Schroeder presents Press Clause protection for AI publishers as a question for analysis, not as an already established rule, and Hu discusses constitutional harms to people from biometric AI systems.
Is the constitutional debate only about free speech?
No. Salib and Schroeder both focus on First Amendment questions, including speech and press. Hu’s article focuses on biometric AI systems and says they can threaten criminal procedure rights and other fundamental constitutional rights when governments deploy them in high-risk settings.
Could AI output still receive some constitutional protection even if AI itself has none?
Yes, in Salib’s account. He argues that AI output can still implicate the First Amendment indirectly when regulation burdens human listening interests or the human use of AI as a tool for creating or performing protected speech.
Does an AI Bill of Rights in these materials mean rights for AI systems?
Not in Hu’s article abstract. Hu uses the phrase in connection with biometric data collection, emerging AI systems, and the constitutional threats those systems can pose to people. The abstract does not say the Bill of Rights would make AI itself a constitutional rights-holder.
caveats: The clearest direct argument about AI’s own constitutional status here comes from Salib’s First Amendment article. Schroeder’s source is an abstract that states the scope of his Press Clause article but not its full reasoning or conclusion, and Hu’s abstract addresses constitutional harms to people from biometric AI rather than AI’s own rights.