Our Children Are Not For Sale

Children are not—and never should be—for sale. By failing to categorize virtual child pornography as an unprotected class of speech under the First Amendment, American jurisprudence provides a disservice to American children, and a service to the child exploitation industry. Quite simply, virtual child pornography is child pornography.

Artificial intelligence is developing at an alarming rate. AI-image generators are even being trained using explicit photos of children. In fact, Stanford researchers found more than 1,000 such instances in the dataset known as LAION 5B. Deepfakes, and deepfake porn specifically, are metastasizing across the cyber domain (deepfakes are different from virtual child pornography, where an adult actor is used). This is creating more demand for an industry that exploits children.

As a result, several social media companies like Meta have AI-governed reporting systems that catch this type of content. However, for U.S. law enforcement to view the image in question, they must obtain a search warrant unless the content has been reviewed first by a social media company. In one shared database, international investigators have flagged 13,500 AI-generated images of sexual abuse—and that is just the beginning.

A pedophile who was tried and convicted in Canada for virtual child pornography had more than 540,000 files on his computer dedicated to sexual abuse. Among those was a series of images showing one girl being abused over the span of seven years, the AI-generating her different ages from ages seven to 14. It is only a matter of time before a case like this reaches the American courts.

How the U.S. can regulate AI-produced images without violating the First Amendment is an ongoing legislative question. Currently, American law is definitionally “missing the forest for the trees,” heavily focusing on safeguarding free speech while leaving little room to combat virtual child abuse.

Is protecting unbridled free thought worth sacrificing our children’s protection from predators’ using their playground pictures to feed AI-generated child porn? The answer to that question should be obvious.

Because the law has not yet provided the necessary protections for our children in this new AI age, state attorneys general have called upon Congress for legislation that will prevent the exploitation of our children through AI. In February of 2023, several adult sites and Meta implemented an online tool called “Take It Down” that allows teenagers to report explicit images and videos of themselves from the internet. These private company actions and legislative efforts show how the legal field is abdicating its responsibility.

The same reasoning that has made obscenity and child pornography unprotected speech under the law is the same reasoning that should bar the creation and distribution of virtual child pornography. There are two premises underlying this conclusion.

First, under Chaplinsky v. New Hampshire (1942) and its progeny, child pornography is categorically obscene and therefore legally prohibited, being clearly outside the boundaries of the First Amendment. Second, obscenity not only includes the evils of child pornography but also other considerations of an unprotected category under the First Amendment such as (1) an underlying crime; (2) the message versus the manner of the content; and (3) the value of the message and legal standards like strict liability. When there exists an underlying crime, the message overrides the manner of the content, and the valuelessness of the message are met, there is potential for strict liability.

Virtual child pornography is becoming both more prevalent and easier to spot. In the Canadian case, police were able to identify the images because they have a “digital fingerprint” that shows they were virtually created, an attribute that does not somehow override its pornographic nature. Thus, this exceptional categorization is inherently faulty. The compelling government interest outweighs the freedom to make such content. Both cases on the books, cases waiting to be heard, and statutes yet to be written should all be interpreted to ensure that virtual child pornography falls in the category of unprotected speech.

When creating and refining current laws, three important standards in American jurisprudence are: 1) no vagueness, 2) a lack of overbreadth, and 3) no unbridled discretion. For a law not to be vague, it must be clearly defined by objective criteria and explicit standards such that a person of common intelligence would not have to guess, and such that it does not delegate policymaking subjectively. The best way to craft such a law is to outline specific guidelines that identify virtual child pornography correctly, which could include a more comprehensive and correct reading of 18 U.S. Code §1466A, for example. Additionally, the Supreme Court could choose to interpret its ruling in New York v. Ferber (1982), which found that states can ban material depicting children engaged in sexual activity, to cover virtual child pornography as well. This would allay vagueness concerns by giving the legal framework that will help define the meaning of the term “virtual child pornography.”

One reason a court sets a law aside is that it overregulates what should be constitutionally protected under the First Amendment. Overbreadth can be avoided by making sure that the definition of virtual child pornography is adhered to, and that the First Amendment is not infringed upon. In practice, this would look like a law that adds virtual child pornography to its list of contraband and has a well-crafted definition of what virtual child pornography is and what it is not.

Unbridled legal discretion means that there are no explicit limits and a lack of neutral criteria. The purpose of such discretion is to cut off the problem before it gets to content discrimination (restricting particular types of speech) or viewpoint discrimination (restricting particular views on a subject matter).

Weakening the child exploitation industry is a compelling government interest. As such, caselaw and statutes must be narrowly crafted to protect children and allow true AI developers to flourish, thereby preventing viewpoint discrimination and protecting against certain types of virtual creations being erroneously stamped as “virtual child pornography.”

Just because virtual child pornography is virtual, it does not definitionally become less pornographic. The law has the authority and the ability to protect children. It will take a nuanced, thoughtful, and swift-acting justice to ensure that it does just that.