CT Open Launches AI-Powered Clinical Trial Prediction Platform with Automated Decontamination

CT Open introduces an open-access, live platform designed to host four annual challenges for predicting clinical trial outcomes, a domain where even experts struggle. The platform allows anyone to submit predictions, which are evaluated against trials whose results were not public at submission time but later revealed. A key innovation is a fully automated decontamination pipeline that uses iterative LLM-powered web searches to identify the earliest mention of trial outcomes, addressing the difficulty of determining when results become public. This approach tackles the lag in official registry postings, which can take years, and the potential for first mentions in obscure articles. The platform aims to test whether AI systems can predict real-world events more reliably than current methods, focusing on the high-stakes challenge of clinical trial outcome prediction. The initiative is documented in arXiv preprint 2604.16742v1, highlighting its novel methodology for ensuring uncontaminated evaluation. By running regular challenges, CT Open seeks to advance predictive accuracy in a critical field, leveraging open participation and automated verification to foster innovation.