Fixing cracks in the artificial intelligence drug development pipeline

In April 2025, the US Food and Drug Administration (FDA) published a roadmap for leveraging new approach methodologies (NAMs), including in silico approaches such as artificial intelligence (AI), to reduce and replace animal testing in preclinical safety studies of drugs and biologics. Animal models on drug safety and efficacy not only present ethical concerns but are also inadequately representative of human response. More than 90% of animal-tested drugs do not reach the market, partly owing to the drugs failing safety and efficacy evaluations in humans.