Anthropic reports early results from Claude drug-development work

Anthropic's life sciences team says that over the past several months it has run deep research partnerships with six organizations developing new medicines: BigHat Biosciences, Bristol Myers Squibb, Divergence Labs, Enveda, Generate Biomedicines and The Column Group. It has also begun drug discovery efforts of its own. The stated strategy is to put AI into many stages of drug development, not only the commonly discussed task of proposing new molecule designs, because the bottlenecks are spread across the whole process: target identification, molecule design, manufacturing, regulatory filings, and the design and running of trials. Anthropic says the work serves three aims: learning where its models are strong and weak (it says it has already prioritized specific model improvements as a result), producing case studies for the wider life sciences community, and, if the positive trajectories continue, having a direct impact through new medicines for food allergies, rare childhood diseases and several forms of cancer.

Generate Biomedicines is trying to make CAR T-cell therapy, which has historically helped treat blood cell cancers, work against solid tumors. That means shortening one part of the therapy's genetic instructions to make room for additional instructions. Six Generate scientists used Claude Code to propose new designs for that part. One design was about 70% shorter than the standard version. In lab tests it let immune cells make roughly 10 times as many receptors, killed tumor cells more efficiently, and kept the cells active for longer.

Divergence Labs is working on a food allergy treatment that would detach IgE from immune cells entirely, rather than only neutralizing IgE circulating in the blood as standard medications do. That needs a two-armed design: one arm binding tightly to IgE, another helping prise it from the immune cell. Claude turned Divergence's goals into a detailed design plan and then, largely autonomously, generated roughly 35,000 antibody designs in about a month, to be narrowed down and tested. Early tests found hits for both arms: three designs that could bind to IgE and 20 that could dislodge it. Three of those 20 appeared to dislodge IgE more than the standard medication. The next step is to build a drug that combines the two arms.

BigHat Biosciences pairs AI with an automated lab. BigHat scientists and an Anthropic researcher worked with Claude to improve binding, stability and resistance to clumping in two antibody-fragment leads, one aimed at cancer cells and one meant to keep a drug in the body longer. Claude, connected to BigHat's lab data system Reccy, proposed changes, and the lab made and tested about 1,010 variants in about 10 weeks. Partway through, the team suspected the standard heat-stability and clumping metrics were misleading; Claude and the researcher reanalyzed the raw data and proposed different metrics and a protocol change. BigHat's scientists refined and ran it, and the main predictions held. The best versions bound more than 10x more tightly than the originals, the cancer-targeting lead stayed in solution at a temperature 18°C higher, and the longer-lasting-drug lead clumped more than 100x more slowly.

At Enveda, manufacturing chemists spent up to 75% of their time on paperwork. Enveda built a system that connects all its manufacturing documents to Claude, which flags possible issues. In tests on 1,061 pages of documentation, Claude caught all 11 issues human reviewers had spotted and flagged seven they had missed, and human review time fell by 89%, from over 25 hours to less than three. Separately, Claude wrote first drafts of two important regulatory documents. One took four hours, against about three months by hand. For the other, Enveda's standard human review found one minor error.

The Column Group is establishing Pleiades, a new company meant to use Claude to run many more, and more complex, drug programs than a small team usually could. The aim is to cut scientific and administrative overhead so investors can back more ambitious ideas, and Anthropic says that if it works the model could be replicated across the industry. Pleiades is still building tools and workflows; its first Claude skill, which helps assess whether drug targets are worth pursuing, has been evaluated on 45 target-disease pairs.

More recently, Bristol Myers Squibb began using Claude to assess the plans of an early-stage drug in clinical trials and to estimate the likelihood those trials succeed. The two plan to expand the work in 2027. Michael Pourdehnad, MD, Senior Vice President, Head of Early Clinical Development, Hematology, Oncology and Cell Therapy, is quoted saying that from a single prompt Claude produced a robust initial analysis that would otherwise take days.

Anthropic is also running research programs itself. It has acquired Mithril Medicine, a one-person company developing treatments for two rare forms of muscular dystrophy. Founder Cody Tranbarger used Claude to organize contracts, logistics and experiments with 11 external labs and delegated many other operations to it. Mithril took a possible treatment for one form of muscular dystrophy from an idea to promising results in mice in less than 18 months. Tranbarger is now at Anthropic, continuing the research and planning several new programs. If his drug candidates keep holding up, Anthropic expects a drug to be ready to start human clinical trials as soon as late 2028.

Anthropic stresses that all of this is early and mostly preclinical, meaning for the most part not yet at the stage of testing in humans. It plans to extend Claude into clinical trials as programs advance and to partner with organizations already at the clinical stage. Beyond the partnerships, it has rolled out Claude Science, an AI workbench for researchers, is providing Claude credits for life scientists, created the Life Sciences Verification Program (which lets trusted organizations use its most powerful models with more permissive versions of its safeguards), and has opened a call for guest investigators in the life sciences. Drug discovery organizations that want to collaborate are invited to get in touch.

Key facts

• Anthropic reports several months of research partnerships with six organizations: BigHat Biosciences, Bristol Myers Squibb, Divergence Labs, Enveda, Generate Biomedicines and The Column Group.
• Claude generated roughly 35,000 antibody designs in about a month for Divergence Labs; early tests found three designs that bind IgE and 20 that could dislodge it, and three of those 20 appeared to beat the standard medication.
• At Enveda, a Claude review system cut human review of manufacturing documents by 89%, from over 25 hours to less than three, and caught all 11 issues humans had found plus seven they missed.
• Anthropic has acquired Mithril Medicine, a one-person company working on two rare forms of muscular dystrophy, and expects a drug could be ready for human trials as soon as late 2028 if candidates keep holding up.
• Anthropic says the work is still early and mostly preclinical, with no human trial results reported.

Why it matters

This is Anthropic's own account of moving beyond the usual AI-for-drugs pitch of proposing molecule designs. The post argues that bottlenecks sit throughout drug development, so it describes Claude working on antibody and CAR T-cell design, lead optimization, manufacturing paperwork, regulatory drafting, company creation and clinical trial planning. It also shows Anthropic becoming a drug developer itself, through the Mithril Medicine acquisition and its muscular dystrophy programs, where it says Claude handles much of the scientific and operational work. Anthropic says it plans to scale this work up substantially.

Who it affects

Biotech and pharma teams are the direct audience: Anthropic says that to raise the rate of new cures it needs to work with many more firms and integrate Claude more deeply into their processes. The partners named are BigHat Biosciences, Bristol Myers Squibb, Divergence Labs, Enveda, Generate Biomedicines and The Column Group. Investors are in view too, since Pleiades is meant to cut the cost and time of discovery so backers can support riskier ideas. Researchers in the life sciences can use the tools and programs Anthropic lists.

How to use it

Anthropic points researchers to Claude Science, its AI workbench for researchers, and says it is providing Claude credits for life scientists. The Life Sciences Verification Program lets trusted organizations use its most powerful models with more permissive versions of its safeguards. There is also an open call for guest investigators: research groups can propose projects using Claude to investigate key questions in drug discovery and other areas. Organizations focused on drug discovery can contact Anthropic about collaborating. The post gives no pricing or credit amounts.

How solid is it

This is a first-party Anthropic post, and the numbers come from partner case studies, with technical reports and preprints linked for methods and early results. No peer review of the results is mentioned. The figures are specific and described as lab or test results: a design about 70% shorter with roughly 10 times as many receptors, 1,010 variants made and tested in about 10 weeks, and 1,061 pages of documents reviewed at Enveda. Most of the claims about the wider impact are qualitative or conditional, such as 'if this works' for Pleiades and 'as soon as late 2028' for a Mithril-derived drug.

Risks and caveats

Anthropic itself says all of this research is at an early stage and mostly preclinical, so no human clinical trial results are reported. Hits at Divergence Labs are early-test results, and the next step, building a drug that combines the two arms, is still ahead. The source does not say which Divergence designs or BigHat variants are headed to clinical use, and it gives no outcomes for the Pleiades skill's evaluation on 45 target-disease pairs. Anthropic's optimism that some projects may yield effective medicines is explicitly conditional on positive trajectories continuing. No Claude model version is named for most case studies, and no acquisition price or partnership financial terms are given. Anthropic also notes that its safeguards are needed to prevent misuse but can impede legitimate biological research.

“From a single prompt, Claude generated a robust initial analysis that helped inform our assessment of different development options, which would otherwise take days to complete.”

— Michael Pourdehnad, MD, Senior Vice President, Head of Early Clinical Development, Hematology, Oncology and Cell Therapy