Intentionally Unintentional: GenAI Exceptionalism and the First Amendment

2025-06-05

AI Summary

This paper challenges the prevailing assumption that generative AI outputs (e.g., from GPT-4, Gemini) qualify as “speech” protected under the U.S. First Amendment, arguing that such outputs lack the subjective intent essential to constitutional speech under established case law and theory. Method: Drawing on legal hermeneutics, constitutional theory, and philosophy of AI, the paper conducts the first systematic critique of AI “speech” claims, analyzing doctrinal requirements for protected expression and interrogating the ontological status of AI-generated content. Contribution/Results: It establishes *intentionality* as a necessary jurisprudential threshold for First Amendment protection, demonstrating that AI outputs do not satisfy this condition—and that users possess no derivative free-speech right in receiving them. The analysis refutes dominant “AI speech rights” positions, revealing that constitutional recognition would undermine regulatory legitimacy, exacerbate disinformation, and entrench technological power asymmetries. The framework provides a foundational theoretical basis for constitutionally sound governance of generative AI.

Technology Category

• Philosophy and Ethics of AI: Artificial General Intelligence
• Natural Language Processing: Generation
• Humans and AI: AI for Accessibility

Application Category

• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

This paper challenges the assumption that courts should grant First Amendment protections to outputs from large generative AI models, such as GPT-4 and Gemini. We argue that because these models lack intentionality, their outputs do not constitute speech as understood in the context of established legal precedent, so there can be no speech to protect. Furthermore, if the model outputs are not speech, users cannot claim a First Amendment speech right to receive the outputs. We also argue that extending First Amendment rights to AI models would not serve the fundamental purposes of free speech, such as promoting a marketplace of ideas, facilitating self-governance, or fostering self-expression. In fact, granting First Amendment protections to AI models would be detrimental to society because it would hinder the government's ability to regulate these powerful technologies effectively, potentially leading to the unchecked spread of misinformation and other harms.

Problem

• Challenges First Amendment protections for GenAI outputs
• Argues AI outputs lack intentionality and legal speech status
• Warns extending free speech rights harms societal regulation

Innovation

• Challenges First Amendment for AI outputs
• Denies AI outputs as protected speech
• Warns against AI free speech rights