Helen L. Norton (University of Colorado Law School) has posted Nonhuman Speakers and the Constitution on SSRN. Here is the abstract:
Long-standing free speech doctrine sometimes does—and sometimes doesn’t—turn on speakers’ human character. This Essay explores how these doctrinal threads might shape the constitutional doctrine applied to generative artificial intelligence’s (AI) speech outputs if we assume that a human speaker cannot be identified as the creator of those words and images—and that those words and images are nevertheless considered “speech” for First Amendment purposes.
For example, First Amendment doctrine sometimes requires proof of a speaker’s culpable mental state before that speaker can be held liable for what would otherwise be unprotected speech. These requirements seek to prevent the chilling of human speakers’ valuable expression and to ensure the punishment only of morally blameworthy human speakers. But because generative AI’s speech outputs do not pose these concerns, these doctrinal features requiring a human speaker should not apply to the government’s regulation of generative AI’s outputs, lest those outputs be insulated from liability for endangering human users.
Other long-standing First Amendment doctrines apply to the government’s regulation of speech regardless of the speaker’s human or nonhuman character. This is the case, for example, of commercial speech doctrine, which protects commercial speech that furthers the interests of human listeners but not commercial speech that frustrates those humans’ interests. These doctrinal threads show that extending First Amendment coverage to nonhuman speakers’ expression need not foreclose the government’s regulatory efforts to protect human users from grave harm. At the same time, these doctrinal threads also show that First Amendment coverage of the words and images produced by generative AI can provide important guardrails against the government’s self-interested or clumsy regulation of those outputs despite their value to human users.
Finally, the Court’s government speech doctrine generally remains indifferent to governmental speakers’ human or nonhuman character. Here, too, governments’ increasing use of generative AI for expressive purposes calls for an appropriately listener-centered government speech doctrine: one that recognizes the inevitability (and often the value to human listeners) of government speech, while constraining the limited but important universe of government speech that interferes with its (human) targets’ exercise of their constitutional rights.
Recommended!
To receive new posts from Legal Theory Blog by email, get a free subscription to Legal Theory Stack.
Lawrence Solum
