Apply for Anthropic’s AI for Science rare disease research grants

Anthropic is narrowing the aperture on its AI for Science initiative, and the first target is rare disease research — a field where the science is fragmented and the patients, despite numbering an estimated 400 million worldwide across more than 7,000 conditions, are scattered too thin for any single hospital or lab to study at scale. Instead of another broad call for proposals, the company is running a themed grant cycle: up to $50,000 in Claude API credits over six months, split across two tracks, with applications open through August 2, 2026.

Track one is aimed at basic science — clinicians, patient groups, and data scientists trying to find shared mechanisms across diseases that are usually studied in total isolation. Anthropic's anchor partner here is the Monarch Initiative, the group behind the Mondo Disease Ontology, which tries to reconcile the mess of competing disease-naming systems (OMIM, Orphanet, ICD, and dozens more) into something computable. Monarch has also built something called DisMech, a library designed so an AI agent can chew through case reports, variant databases, and raw registry data and flag mechanistic overlaps between diseases that a human researcher would likely never connect. Grantees get direct access to contribute to and use these tools, with outputs published openly at monarchinitiative.org.

Track two goes after the slower, more bureaucratic side of the problem: turning a confirmed genetic diagnosis into an actual treatment, a process that today eats one to two years, mostly spent in manufacturing queues, running safety studies one after another instead of in parallel, and assembling regulatory paperwork by hand. Anthropic's pitch is that Claude can compress this — drafting IND sections and investigator brochures, checking whether a drug target is druggable across small molecules, antibodies, or gene therapies, and spotting shared mechanisms across ultra-rare conditions that might let several be folded into a single basket trial instead of separate applications for each patient.

The company points to existing grantees as a preview of what it wants more of: Every Cure sifting millions of drugs for repurposing candidates, the Centre for Population Genomics building a system to draft variant classifications for expert sign-off, and the Violet Research Institute — a small nonprofit working on diseases affecting fewer than 1 in 50,000 births — using Claude to navigate FDA guidance and run its own bioinformatics pipelines.

Anthropic is careful to hedge its own pitch, though. It admits Claude is useless where data is too sparse or badly organized to begin with, and it can't touch the parts of the so-called diagnostic odyssey that come down to insurance denials or lack of access to testing facilities — problems no amount of API credit will fix. The framing throughout is less about AI curing rare disease and more about it being one tool among several needed, alongside better data collection and public-private partnerships that actually reach patients.