Deepcell and Gilead are co-developing an AI foundation model for CHO cell line development, training on Deepcell’s REM-I single-cell morphology data using self-supervised learning and planning an open-weight initial model. In pathology AI, PolyU’s TRUECAM framework focuses on when AI predictions are uncertain versus reliable, supporting a human-in-the-loop workflow. Separately, an FDA TEMPO pilot approach is testing how regulators evaluate digital health devices.