Towards autonomous medical artificial intelligence agents

A large language model artificial intelligence agent demonstrated autonomous capabilities within a sandboxed electronic health record system on June 17, 2026, according to a publication in Nature. This AI agent successfully performed complex medical tasks including taking patient histories, ordering diagnostic tests, interpreting test results, diagnosing medical conditions, and proposing treatment plans. In its evaluations, the AI agent outperformed experienced human clinicians. Crucially, the agent also adhered to established safety standards and clinical guidelines throughout its operations. The research highlights the potential for AI agents to operate autonomously in healthcare settings, managing patient care pathways from initial assessment to treatment recommendation. This development signifies a significant step towards integrating advanced AI into clinical workflows, potentially improving efficiency and diagnostic accuracy while maintaining patient safety protocols.