Download:

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7584799

Abstract:

Artificial intelligence (AI) often looks like speech, functions like speech, and causes the same types of harms as speech. Many aspects of the output of AI should therefore be considered protected speech under the First Amendment. Failure to protect this output will give the government alarming power over our ability to inform, express, and communicate. Without free speech protections, the government could prohibit AI from discussing certain topics or espousing certain viewpoints. 

This Article aims to demonstrate how our current First Amendment doctrine already provides the framework to regulate AI while preserving the free speech rights of creators and users. There is no requirement that expression be “certain” to qualify as speech, and our free speech jurisprudence will be upended if AI is not deemed speech. Even if the outputs of AI are considered speech, the government can address some of the harms caused by AI through longstanding free speech doctrine. In addition, because of the differences between AI and traditional speech, the government may have some enhanced power to compel AI creators to disclose when content is AI, or to compel disclosure of the sources of the AI output.