How generative AI reconfigure clinician–AI and clinician–patient relationships

How generative AI reconfigure clinician-AI and clinician-patient relationshipsThe rapid uptake of generative AI (GAI) systems in clinical settings-particularly large language models (LLMs) such as DeepSeek-marks a transformative moment for healthcare and necessitates timely regulatory responses to protect safety and equity.In current practice, LLMbased GAI is entering clinical practice through a distinctive 'dual-interface' configuration.On one side, models locally deployed or fine-tuned by healthcare institutions are embedded into hospital information systems and patient-facing portals to streamline care processes, support clinical decision-making and facilitate personal health management. 1 On the other, general-purpose models developed by companies, such as OpenAI, Google and DeepSeek, are reaching clinicians and patients through a growing ecosystem of chatbot applications.2,3 These channels make GAI increasingly accessible to both sides of the medical encounter.Clinicians use GAI to retrieve medical knowledge, generate analytic reasoning and obtain suggestions for diagnostic and therapeutic decisions (DTD).4 Simultaneously, patients consult these models about symptoms, use them to interpret complex clinical rationales and review clinicians' recommendations.5 This dual-interface setting is shifting healthcare away from two separate dyads-'clinicianpatient' and 'clinician-AI tool' (Figure 1A)-toward a configuration where multiple human-AI relations coexist (Figure 1A-D).This Letter examines two linked axes of change.First, when GAI operates as an interactive cognitive collaborator, it enters the clinician-led chain-of-thought (CoT).In configuration (B), a previously clear 'clinician-tool' relation becomes a clinician-AI collaboration requiring a redefinition of roles.Second, when systems capable of generating clinical CoT interact directly with patients and act as decision-making agents, configurations (C)-(D) recast the clinician-patient relationship-its modes of explanation, patient involvement and trust-into a tripartite arrangement that explicitly includes AI.