Tamara Dobler

Speech liberalism holds that the linguistic output of generative AI is continuous with human speech and should therefore be treated as meaningful in broadly the same way. Strong versions of this view take the fluent behaviour of chatbots to justify attributing genuine linguistic and cognitive agency, while nevertheless advocating caution when questions of moral or legal responsibility arise. I argue that this gives too much evidential weight to successful conversational interaction, where speaker intentions largely recede into the background. Once AI-generated speech contributes to serious harm, however, those intentions become central. More broadly, the paper challenges a way of theorising about AI speech that builds classifications of linguistic agency primarily from successful conversational interaction. I develop two independent but complementary arguments. The first is ethical–legal. I argue that classifying chatbot output as speech is not normatively neutral: within existing free-speech doctrine, it risks generating responsibility and liability gaps in cases of AI-mediated harm. The second is epistemic. I argue that the behavioural evidence supporting strong speech liberalism is systematically distorted by anthropomorphic bias, reinforced by AI design practices. The paper concludes that theories of AI speech should not be built exclusively from successful conversational exchanges and that, until a broader evidential case is made, a more conservative approach to AI speech is warranted in high-stakes contexts.

Should conversational AI be treated as constitutionally protected speech? Recent litigation, most notably Garcia v. Character Technologies, Inc., presents this question in a new form. While the plaintiff characterises Character.AI as a commercially distributed product whose design gives rise to ordinary principles of safety and liability, the defendants argue that chatbot outputs constitute protected speech under the First Amendment. We argue that Garcia presents a problem of legal underdetermination between competing concepts of speech rather than a dispute resolvable by analogy alone. Drawing on conceptual ethics, teleological approaches to legal interpretation, and recent work on legal construction, we develop a normative methodology for choosing between competing legal concepts. Applied to conversational AI, this methodology requires asking which concept of speech best advances the constitutional purposes of the First Amendment. We argue that those purposes include not only expressive freedom, listeners’ interests, and innovation, but also responsibility and legal redress, and that they do not all support the same concept of speech. Extending constitutional speech coverage to conversational AI places pressure on doctrines governing harmful speech that continue to presuppose intentional human speakers, while offering comparatively little additional protection for the constitutional values traditionally served by the First Amendment. We therefore argue that courts have good reason to prefer a concept of speech that does not initially treat AI-generated outputs as constitutionally protected speech but instead allows disputes concerning conversational AI to be addressed through product law.

ABSTRACT This paper examines Wittgenstein’s rule-following argument in relation to Chomskyan generative grammar, focusing on selectional restrictions. While Wittgenstein views meaning determining rules as context-dependent and socially embedded, Chomsky treats linguistic rules as computational and internal to the mind. I argue that Wittgenstein’s critique does not apply to most syntactic rules but poses a challenge for selectional restrictions, which rely on world knowledge. If lexical features depend on external factors, selectional restrictions cannot be purely internalist. To preserve internalist semantics, I propose a modification of Pietroski’s (2018) model in which meaning rules function automatically, eliminating interpretative choice. This approach reconciles internalism with linguistic flexibility while maintaining its explanatory power within a generative framework.