Anthropic’s Claude Is Now Designing Drugs — Here’s What It Found

Anthropic has released an early progress report on its life sciences initiative, detailing results from research partnerships with six pharmaceutical and biotech organizations: BigHat Biosciences, Bristol Myers Squibb, Divergence Labs, Enveda, Generate Biomedicines and The Column Group. The company is also running in-house drug discovery programs of its own. The announcement marks one of the most detailed, numbers-heavy disclosures any major AI lab has made about applying a general-purpose model across the full arc of drug development.

The core argument Anthropic is making is architectural: rather than building a narrowly specialized system for, say, protein folding or molecule generation, the company is betting that Claude — connected to lab data systems, documentation pipelines and scientific literature — can compress timelines at every stage of drug development simultaneously. That includes target identification, molecule design, lab optimization, manufacturing records, regulatory filings and clinical trial planning.

What the Partnerships Actually Produced

The case studies vary significantly in scope, but several share a common thread: speed gains that would be difficult to achieve with human labor alone.

At Generate Biomedicines, six scientists used Claude Code to redesign a component of CAR T-cell therapy — a cancer treatment that has historically worked best against blood cancers. The goal was to engineer a version that could also attack solid tumors. One Claude-proposed design was roughly 70% shorter than the standard version; in lab tests, it enabled immune cells to produce approximately 10 times as many cancer-targeting receptors, killed tumor cells more efficiently, and kept the immune cells active longer.

At Divergence Labs, which is developing a novel treatment for food allergies, Claude generated approximately 35,000 antibody designs in about a month. The model was working toward a two-armed drug: one arm to bind to IgE (a molecule that triggers allergic reactions) and another to detach IgE from immune cells entirely — a more aggressive intervention than existing medications. Early tests identified three designs that could bind IgE and 20 that could dislodge it from immune cells, with three of those 20 outperforming the current standard medication in preliminary assays.

At BigHat Biosciences, Claude worked alongside a BigHat–Anthropic research team to improve two antibody fragments — one targeting cancer cells, one designed to extend a drug’s residence time in the body. After testing about 1,010 variants over roughly 10 weeks, the best candidates bound their target more than 10 times more tightly than the originals; one remained stable at temperatures 18°C higher, and the other clumped more than 100 times more slowly. Notably, midway through the project, Claude flagged that standard measurement metrics were potentially misleading, prompting a protocol revision that the researchers ultimately validated.

Perhaps the most immediately practical results came from Enveda, a company focused on plant-derived medicines. Manufacturing chemists there were spending up to 75% of their time on documentation required for regulatory submissions. After connecting its records to Claude, Enveda ran the system over 1,061 pages of documents. Claude caught all 11 issues that human reviewers had already identified — plus seven they had missed. Human review time dropped by 89%, from more than 25 hours to under three. A separate regulatory document that normally takes roughly three months to draft was completed in four hours, with only one minor error caught during review.

Anthropic’s Internal Drug Program — and Its Own Company Acquisition

Beyond the partnerships, Anthropic disclosed that it acquired Mithril Medicine, a one-person company working on two rare childhood forms of muscular dystrophy. The founder used Claude to manage contracts and coordinate work with 11 external laboratories — essentially running what would typically require a larger team. That program moved from initial concept to promising results in mouse models in under 18 months. If the science holds, Anthropic is targeting the start of human clinical trials as early as late 2028 for one condition and 2029 for the other.

Where Anthropic Stands in a Crowded Field

Anthropic is entering a space where several well-capitalized competitors have a significant head start. Google DeepMind’s AlphaFold transformed protein structure prediction, and its spinout Isomorphic Labs — which raised a $2.1 billion Series B in May 2026 and signed deals worth billions with Eli Lilly and Novartis — is targeting its first human clinical trials in late 2026. Insilico Medicine already has an AI-discovered drug in human trials. Anthropic’s most optimistic clinical timeline, late 2028, puts it behind both.

But the strategic bet is different. Where Isomorphic Labs and Insilico have built specialized platforms focused primarily on molecule design and protein structure, Anthropic is positioning Claude as a horizontal agent capable of handling scientific reasoning, operational logistics, regulatory writing and lab data analysis within a single system. The Mithril Medicine example — one person, 11 labs, 18 months — is the clearest illustration of that thesis.

What This Means If You’re Studying Biology, Chemistry or CS

The most instructive detail in Anthropic’s report for students may be the Mithril Medicine case. A single researcher, equipped with AI tools, ran a multi-lab rare disease program from scratch in a year and a half. That is a meaningful signal about what early-career scientists who are fluent in AI workflows can accomplish at small biotech startups — and why AI drug discovery sits at the intersection of machine learning, biology, chemistry, and regulatory affairs rather than belonging cleanly to any one discipline.

Demand for professionals who can navigate that intersection consistently outpaces supply. Entry-level salaries in AI drug discovery in the United States typically range from $90,000 to $120,000, with senior roles commanding considerably more.

For students who want to get hands-on without enterprise access, Anthropic announced Claude credits for life scientists and an open guest investigator call as part of this initiative — concrete pathways worth exploring now rather than after graduation.

Source: Anthropic