Notes: Eugene Volokh is one of the preeminent First Amendment scholars working today, and he has a long-standing interest in technology and Internet law, dating back to his extensive experience as a programmer. Note that this is a big and slightly sprawling article; many doctrinal articles are, because they attempt to cover all of the relevant issues that could arise in a lawsuit. (I’ve certainly been guilty of this myself!) As you read, try to pin down which issues are fundamental and raise genuinely new challenges, and which are likely to become non-issues over time.
Questions:
Should we treat the outputs of AIs as potentially defamatory? Aren’t they obviously the results of purely digital processes, with no semantic meaning?
Relatedly, why should anyone take what ChatGPT outputs seriously? Should the legal system presume that everyone should know that generative-AI chatbots hallucinate all the time?
Can an AI system have actual malice? What would it take to conclude that ChatGPT emitted an output “with knowledge that it was false or with reckless disregard of whether it was false or not”?
Should OpenAI have a duty to prevent ChatGPT from hallucinating defamatory lies? Should it have a duty to investigate and fix ChatGPT after someone points out a defamatory lie that it emitted?
Dan Burk, Asemic Defamation, or, the Death of the AI Speaker, 22 First Amendment Law Review 189 (2024) is the most pungent argument that AI outputs aren’t even meaningful enough to be defamatory. Another take arguing that AIs lack the intentions required in law for certain kinds of liability is Ian Ayres and Jack M. Balkin, The Law of AI is the Law of Risky Agents without Intentions, University of Chicago Law Review Online (forthcoming).
James Grimmelmann, The Defamation Machine (38th Annual Silha Lecture 2023) is my take on whether AI outputs have meaning and whether AI can have actual malice. It has pictures!
Lawrence B. Solum, Artificial Meaning, 89 Washington Law Review 69 (2014), is a philosophically sophisticated take on where the meaning, if any, in AI outputs comes from.
Peter Henderson, Tatsunori Hasimoto, and Mark Lemley, Where’s the Liability in Harmful AI Speech?, 3 Journal of Free Speech Law 589 (2023). Another recent paper, from a mixed CS+law team of authors.
Stuart Minor Benjamin, Algorithms and Speech, 161 University of Pennsylvania Law Review 1445 (2013). An older but very careful analysis of whether and when algorithmic outputs can be “speech” covered by the First Amendment. Compare Tim Wu, Machine Speech, 161 Pennsylvania Law Review 1495 (2013), from the same symposium, which takes a more generally skeptical view. James Grimmelmann, Speech In, Speech Out, in Ronald K.L. Collins & David M. Skover, Robotica: Speech Rights and Artificial Intelligence 85 (2018), is my take on some of the issues.