# Stanford AI system analyzes drug trials with 37,075 agents

_Published Monday, October 5, 2026 at 3:18 AM EDT · Science, AI · Latest · Tier 2 — Notable_

![AI drug discovery research shown on a laptop in a genetics laboratory with scientists, test tubes, pipettes, and a microscope. — Primary](https://www.earth.com/img/16x9/1920/media-2026/2026/09/04165040/AI-drug-discovery-research.jpg)

A Stanford research team deployed 37,075 AI agents simultaneously to analyze drug-trial records in about six hours, Earth.com reported. The system examined trial registries, papers and press releases, at a median Anthropic tool cost of 23 cents per trial. The study appeared in Science, and its code is public on GitHub.

Human checks of 100 randomly selected trials agreed with the agents' assessments of whether trials met their main goals about 88% of the time. The analysis linked targets active in fewer cell types to better trial outcomes and fewer side effects. Those findings are associations from past trials; the agents conducted no laboratory experiments.

## Sources

- [Earth.com](https://www.earth.com/science/virtual-biotech-staffed-by-37000-ai-agents-found-a-marker-for-more-successful-medicines/)

---
Canonical: https://techandbusiness.org/newswire/JU686telzMFnuMBSvW9XiK
Published: 2026-10-05T07:18:53.777Z
Story chronology: 2026-10-05T06:02:00.000Z
Retrieved: 2026-10-05T09:09:44.865Z
Publisher: Tech & Business (techandbusiness.org)