Ethical issues regarding the use of AI in psychotherapy

Abstract

Artificial intelligence (AI) is also increasingly establishing itself in psychiatric and psychotherapeutic practice. Against the backdrop of global care disparities, AI-based applications are being discussed or are already in use for screening and diagnostics, clinical decision support, interventions, monitoring, and administrative and educational support. This article analyses opportunities and risks of this development from the perspective of medical ethics. While there are potential benefits, such as improved access to care, destigmatization, clinical decision support, and personalization and training opportunities, there are also significant risks, including insufficient evidence of efficacy, inadequate crisis management, susceptibility to errors ranging up to hallucinations, erosion of the therapeutic alliance, data protection deficits, algorithmic biases, and a regulatory vacuum. Four fundamental ethical questions are discussed in depth: 1. the normative status of AI as a tool or agent, 2. the conditions for trust in conversational AI, 3. simulated empathy, and 4. unrealistic expectations and excessive dependence. Current regulatory frameworks, including the WHO’s guiding principles, the EU AI Act, the EU Medical Device Regulation, and recommendations from professional associations, are also contextualized. Based on this, action-oriented recommendations for psychotherapeutic practice are derived. These include taking a systematic AI-assessment, being transparent about technological limitations, preserving professional judgement and human control, strictly evaluating data protection and security, establishing robust crisis protocols and the development of AI competence. The authentic therapeutic relationship remains the indispensable agent of the therapeutic process. AI should be developed as a complementary tool, not a substitute for human professionals.