How AI Models Learn Bias from Clinical Trial Data

AI Ethics with Fexingo: Bias, Safety, and Responsible Artificial Intelligence · Jul 13, 2026 · 7m

In this episode of AI Ethics with Fexingo, Lucas and Luna examine a specific case: an AI model trained on breast cancer clinical trial data that performed poorly on Black women because the original trials enrolled only nine percent Black participants. They discuss how regulatory guidelines from the FDA and NIH have tried to fix trial diversity since the 1990s, but real enrollment still lags. The hosts trace the problem from trial design to model deployment, asking whether fixing the data pipeline is enough when the underlying healthcare system has biases built in. They also touch on a recent…