Tarmio Frei (Bucerius Law School) & Greta Sparzynski (Bucerius Law School) have posted Hot Singles in Your Area (May Be Chatbots) Comparing EU and New York Approaches to AI Companion Transparency on SSRN. Here is the abstract:
AI companions – whether purpose-built systems or general-purpose AI systems (GPAIS) repurposed for companionship – are increasingly used as friends, partners, or even therapists. While such systems promise benefits including emotional support, accessibility, and personalized interaction, they also pose significant risks, ranging from emotional dependency and manipulation to safety concerns and adverse effects on children’s development. Regulation is therefore essential, with transparency playing a central role. This paper compares end-user transparency obligations under EU and New York law, situating each within its broader AI regulatory framework. The EU’s AI Act adopts a broad, risk-based approach that covers both dedicated and GPAIS, but its lack of companion-specific safeguards and the “obviousness exemption” from certain end user transparency obligations under Art. 50(1) risk weakening user protection. By contrast, New York’s General Business Law §§ 1700–1704 reflects greater contextual awareness – particularly through safeguards against self-harm – yet remains narrower in scope and ambiguous in its treatment of GPAIS. Both frameworks reveal complementary strengths and limitations. To address the unique relational risks of AI companionship, targeted standards are required. Proposed state-level initiatives in the United States, such as North Carolina’s SB-624, California’s SB243, and Utah’s HB-452, provide useful points of reference for future regulatory development.