Abstract
In the beginning was a word—then over time, a whole civilization was built on words. Speaking, thinking, and writing have always been inherent aspects of the human experience. Until now. Artificial intelligence machines have hi-jacked words. Through focused training on our words and knowledge, they too can now write, create, and express themselves just like we do. This, then, presents a profound constitutional question flagged by Justice Barrett in Moody v. NetChoice: if the U.S. Constitution protects human expression, does it also protect AI outputs?
This article addresses that question through the derivative principle: If you squeeze an orange into juice, it is still just an orange. The juice might be something different in the end, but it is different in degree, not in kind. In the same way, current AI chatbots—those that are trained on and operate on hu-man knowledge, within preset human parameters—are derivatives of human expression. They don’t break the link to human expression; they simply extend it in a new electronic medium. Like video games and social media, they are interactive platforms operating within human-created parameters.
But there are limits. Somewhere in the recursive loops of AIs training on AI-generated data, somewhere in the emergence of entirely autonomous sys-tems—when the tool becomes a thing of its own—everything changes. If that point is ever reached, then First Amendment protection should end. After all, the U.S. Constitution was written by humans, for humans.
Rather than defining this limit abstractly, courts should resolve this fact-driven inquiry through the evidentiary record: What percentage of training data is human-created versus AI-generated? How many generations removed is the data from human sources? Can outputs be traced to human sources? Does the system operate independently of human instruction? Do human prompts meaningfully shape results? These questions answered case by case, would likely establish where constitutional protections end. For current AI chatbot models, the human link remains unbroken, but if we reach the point it is severed, the evidentiary record will likely show us—and the courts.
Recommended Citation
Mwafulirwa, Mbilike M.
(2026)
"A Free Speech Dilemma: AI Outputs and the Constitution,"
Akron Law Review: Vol. 59:
Iss.
2, Article 4.
Available at:
https://ideaexchange.uakron.edu/akronlawreview/vol59/iss2/4