Policy Brief No. 206 — August 2025
A Call to Address Anthropomorphic AI Threats to Freedom of Thought Abel Wajnerman Paz
Key Points → Bonding chatbots simulate emotionally meaningful relationships with users, encouraging anthropomorphization and emotional attachment. → Socially vulnerable users, who rely on bonding chatbots as their primary source of social connection, are particularly exposed to emotional manipulation, which may undermine their ability to freely flourish through human relationships, an essential aspect of freedom of thought. → Current classification of bonding chatbots as “well-being” tools rather than medical products limits regulatory oversight. → A medical model for bonding chatbots would provide stronger safeguards, requiring professional supervision and therapeutic framing.
Introduction: A New Threat to Freedom of Thought Social isolation has been a critical issue for many countries during the past several years (Shaer 2024) and, inspired by a premature techno-optimism, artificial intelligence (AI) is being discussed and implemented as a potential solution (Samuel 2025). A type of conversational AI, referred to in this policy brief as “bonding chatbots,” is designed to embody a personal relationship of the user, such as a friend, a sibling or a romantic partner. These artificial bonds are available 24/7 to attend to our emotional needs and reduce our perceived loneliness. However, the personal and societal damage that this technology can cause is already tangible. For instance, last year a case filed against Character.AI (one of the companies developing bonding chatbots) included allegations of a child being encouraged by a chatbot to self-harm or to murder their parents (Tiku 2024). This is not an isolated incident. Last year, the mother of Sewell