US health agencies turn to AI to cut animal testing and rethink drug safety

On September 21, 2026, the US Department of Health and Human Services (HHS) announced a new set of actions to move federally funded biomedical research away from heavy reliance on animal testing and toward more “human-based” methods. Behind this shift is rapid progress in artificial intelligence and computational modeling, which now make it possible to predict drug safety and optimal dosing using cell models and human data with growing accuracy.

ARPA-H, the advanced research agency within HHS often described as a “DARPA for health,” is expanding investments in AI and simulation through its CATALYST program. The goal is to use machine learning models to flag toxicity risks earlier, narrow down which drug candidates proceed to clinical trials, and in the process both cut development costs and reduce the need for animal experiments.

The announcement is likely to accelerate AI adoption across the biomedical pipeline, while putting a spotlight on how regulators will vet the reliability and bias of these models. For Japanese pharmaceutical firms and research institutes involved in global trials, changing expectations at US agencies could reshape approval strategies and data requirements, making this a trend worth close attention.