This paper presents a fully autonomous AI-native sponsor operating system for Physical AI oncology clinical trials, replacing traditional human-staffed pharmaceutical sponsor functions with a multi agent software architecture. The system comprises twelve functional agents organized into four layers: governance, study execution, site and robotics interface, and trust infrastructure. Each agent automates a distinct sponsor responsibility, from portfolio management and protocol design through safety monitoring, regulatory submissions, and robotic procedure authorization, while maintaining compliance with adapted regulatory frameworks including 21 CFR Part 312, 21 CFR Part 50, and ICH E6(R3). The architecture integrates with national MCP server infrastructure, federated learning networks, and Physical AI trial sites equipped with ten categories of robotic systems. This version includes automated code generation results: 108 Python scripts generated from the paper’s LaTeX instructions by Claude Code Opus 4.6, with successful execution of a 24-hour simulation producing 288 sponsor decisions across 155 patients and 75 text diagrams across three perspectives. A human sponsor-of-record retains legal accountability and override authority for safety-critical decisions, ensuring that autonomous operation occurs within established regulatory boundaries.
Kevin Kawchak (Sun,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: