Background: The integration of artificial intelligence (AI) in mental health services has experienced exponential growth, with global market projections from USD 1.13 billion (2023) to USD 5.08 billion (2030). This expansion raises specific ethical challenges that exceed traditional bioethical frameworks, particularly regarding algorithmic opacity, systematic biases, and preservation of therapeutic relationships.


Objective: To conduct a narrative review examining ethical aspects and informed consent challenges in AI applications for mental health, analyzing international regulatory frameworks, and proposing guidelines for responsible implementation.


Methods: Narrative review of scientific literature (2019-2025) across PsycINFO, Scopus, PubMed, and Web of Science. The inclusion criteria focused on studies with specific ethical components related to AI in mental health, encompassing controlled empirical studies, systematic reviews, international regulatory frameworks, and theoretical foundations.


Results: Controlled empirical studies demonstrate preliminary effectiveness in specific populations, although they reveal limitations in longitudinal validation and vulnerable populations. International regulatory frameworks converge on fundamental principles—transparency, human oversight, equity—but differ substantially in implementation mechanisms. Systematic reviews identify critical tensions between technological scalability and clinical singularity, and between automation and preservation of professional judgment.


Conclusions: AI should be conceptualized as a tool for clinical enhancement rather than therapeutic substitution. Responsible implementation requires multidimensional validation frameworks, gradual multi-stakeholder adoption processes, and preservation of the therapeutic relationship as an irreplaceable core. Robust longitudinal studies and international consensus on ethical standards are urgently needed.