–Arvind Kurian Abraham, S.J.D candidate, Harvard Law School
It is a truism that not all forms of speech and expression are protected by the First Amendment of the U.S. Constitution. One of the categories excluded from this sphere of constitutional protection is obscenity. So when a man was charged with possessing child sexual abuse material (CSAM) which was generated using AI, why did the Seventh Circuit in its recent opinion in United States v. Andereggrule that the state’s ability to act against him was constrained by the First Amendment?
Decades of jurisprudence have not questioned the need to exclude obscenity from the shield of constitutional law protection. However, what has never been settled, and what every subsequent case has had to re-litigate in some form, is how to identify obscenity without the definition enabling the government to punish whatever offends prevailing taste. This concern is best expressed by Justice Harlan’s statement that “one man’s vulgarity is another’s lyric” (although that was a case involving alleged profanity). From the early 1950s till the early 1970s, the courts ended up with what can only be referred as a doctrinal mess. The height of this doctrinal chaos was when the Justices of the U.S. Supreme Court had to watch movies, some of them pornographic, in the basement of the court to determine whether or not a movie was obscene. Justice Potter Stewart, unable to offer a workable definition of obscenity, admitted that “I know it when I see it.”
To exclude obscenity from First Amendment protection, courts first had to identify what, exactly, is wrong with it. On this matter, the Court has oscillated between two different theories: treating obscenity as simply worthless, “without redeeming social importance” (Roth), and gesturing toward a vaguer notion of harm, such as the degradation of the “quality of life,” the “total community environment,” and “tone of the society” (Paris Adult Theatres).
Some scholars like Cass Sunstein have argued that pornography does not involve the exchange of ideas and therefore should not be granted top tier protection under the free speech clause. Feminist icons Catharine MacKinnon and Andrea Dworkin attempted to offer a harm based rationale against obscenity in relation to pornography, by arguing that it promoted the subordination and exploitation of women. Their approach triggered a debate within feminist circles, with some who argued that the MacKinnon-Dworkin thesis cast women engaging in pornography as helpless victims incapable of consent. Indianapolis adopted the MacKinnon-Dworkin argument by enacting anti-porn legislation based on their thesis, only to see it struck down by the Seventh Circuit on the ground that the law engaged in viewpoint discrimination by suppressing one perspective of women, while promoting another.
Laws against child pornography avoid this pitfall entirely, because children are unable to consent to sexual activities and therefore any form of child pornography is rightly recognized as the exploitation of children. There is a concrete, identifiable injury the law seeks to prevent. The Supreme Court has recognized that a work’s artistic or literary value is simply irrelevant to this harm (Ferber). Given that children are the most vulnerable people in our society, the state may penalize even the private possession of child pornography (Osborne), a marked departure from the general rule that that a person possessing pornography within the confines of his private home cannot be prosecuted (Stanley).
However, what if the sexual material depicts a child who does not actually exist? The Child Pornography Prevention Act of 1996 extended the federal child pornography ban to “virtual” child pornography, sexually explicit images that appear to depict minors but are produced without using any real children, through either youthful-looking adult actors or computer-generated imaging. In Ashcroft v. Free Speech Coalition, the Supreme Court expressed concerns that such an overbroad law could cover mainstream films or artistic material. Shakespeare’s Juliet was thirteen years old, if a film depicted Romeo and her engaging in sexual activity, would it come under the ambit of the law? Furthermore, the government could not show evidence that there was a link between the creation of such material and the claim that it would lead pedophiles to engage in sexual activity with children. The Court reasoned that absent a real victim or a demonstrated link to real-world harm, the law reached well beyond what the obscenity exceptions to the First Amendment permit.
This precedent ended up constraining the Seventh Circuit this past August in deciding whether a person could face punishment for the private possession of CSAM material which was generated with AI, without the involvement of any real children. The court concluded that in light of Ashcroft, as no actual child was used in the production of the sexually explicit materials, the private possession of such material could not be a basis of criminal prosecution without violating the First Amendment.
Justice Thomas, the only sitting Justice who sat on the Ashcroft bench, while upholding the Texas law which imposes an age-verification compliance requirement to access pornographic material, noted that opinions such as Aschroft were deliberately narrow because the Court was wary of stifling the growth of what was then a nascent internet. Now that internet access is ubiquitous and online pornography has grown far more extreme than anything the Court could have anticipated in 2002, Thomas suggested that First Amendment doctrine in this area must adapt accordingly.
Similarly, the advent of AI and its ability to generate sexual content involving children requires the Court and Congress to rethink Ashcroft. One approach could be to adopt the path that Thomas offered in his concurrence in Ashcroft. With considerable foresight, Thomas had reasoned that a ban on sexual material involving virtual children could be constitutionally permissible if the advancement in technology reached a stage which made virtual pornography so indistinguishable from real child pornography that it became “impossible to enforce actual child pornography laws.” Such an approach can avoid the difficulty of having to show that an actual child was harmed in the generation of CSAM using AI. Even if one is not satisfied with that approach, one thing is clear- the First Amendment needs to adapt to a new era being shaped by AI.
Suggested citation:Arvind Kurian Abraham, Virtual Children, Real Problems: AI and the First Amendment in the United States, Int’l J. Const. L. Blog, Nov. 8, 2026, at: http://www.iconnectblog.com/virtual-children-real-problems-ai-and-the-first-amendment-in-the-united-states//
