[Generative AI in medicine: current clinical applications and prospects for next-generation technologies]

Since the release of ChatGPT in late 2022, generative artificial intelligence (AI) has rapidly diffused into clinical practice. Large language models achieve approximately 87% accuracy on the United States Medical Licensing Examination, exceeding 90% with medically tailored prompting strategies, and produce differential diagnoses comparable to specialists in complex cases. In this review, current applications of generative AI in clinical documentation, patient education, differential diagnosis, and image-based decision support are summarized, with a focus on dermatology. Next-generation technologies are surveyed, including multimodal models, autonomous AI agents, on-device AI, and medical-specialized systems such as ChatGPT Health, MedGemma, Dragon Copilot, Cancer Copilot, and the model for facial pigmented lesions used at the author's institution. Regulatory and ethical challenges are discussed, including cross-border transfer restrictions under Article 28 of the Japanese Act on the Protection of Personal Information, hallucination risk, and accountability for AI-assisted decisions. Although generative AI promises major gains in efficiency and diagnostic safety, ultimate responsibility for medical decisions remains with the physician. Clinicians must cultivate critical appraisal skills and a patient-centered orientation so that AI functions as a collaborator rather than a replacement. Sustained literacy and flexible adaptation will define the dermatologist's role in the coming AI era.