Beyond the chatbot: empathy, safety boundaries and continuous clinical supervision – the relational design of Anna AI

Where does artificial intelligence belong in mental health care - what can it do with empathy, where are its safety boundaries, and under what clinical supervision can it be admitted? The discussion makes the question concrete through Anna AI: a free chat-based support tool running since 2023 for cancer patients, their families and health professionals, in some fifteen languages, around the clock. Its practice is clearly bounded: it declines medical questions outside psycho-oncological competence and refers to the clinic, a clinical psychologist reviews the anonymised conversations, and its declared design goal is not to imitate psychotherapy. The international debate turns on the same points: what the system should refuse, who is responsible for recognising and escalating a crisis, and in what form supervision is carried out. The panel opens with a live demo and sets the developers' and the clinicians' perspectives against each other.