Friends in our phones? Legal responses to AI companions

AI companions - chatbots marketed or used as friends, confidants, and romantic partners - have moved rapidly from a source of novelty to a subject of litigation, legislation and regulation. This article considers three foundational chatbot cases: two involving lawsuits against chatbot providers and a Federal Trade Commission complaint against a third. While some concerns about AI focus on elements of dishonesty and deception, the providers in these cases were transparent about the true nature of the chatbots, and the users were under no illusions (at least initially) as to their nature. We argue, however, that deception and delusion remain plausible risks even under conditions of transparency. Drawing on the concepts of ‘dishonest anthropomorphism’ and ‘fake intimacy’, we argue that risks of deception can arise even without any misperception of facts. Psychological dependency and the displacement of human relationships can harm users even when they hold accurate beliefs about their chatbot interlocutors. Applying this to existing and emerging legal responses, including companion-specific legislation in New York and California, we find reason to be sceptical that mandated reminders of nonhuman status would have averted the tragedies that motivated them.