Anthropic is expanding its work on drug discovery, saying it has teamed up with six biotech and pharmaceutical organizations and started research programs of its own to test how its Claude AI can contribute across the development of medicines. The company says its next step is to scale up those efforts, while acknowledging that most of the work remains at an early, preclinical stage.
The partnerships span tasks from designing potential therapies to reviewing manufacturing records and planning clinical trials. Anthropic argues that AI could help across the entire drug-development process, rather than only proposing molecules: researchers must also identify promising biological targets, improve candidate drugs, prepare regulatory paperwork and organize trials.
Now loading...
Among the collaborations, Generate Biomedicines is exploring ways to adapt CAR T-cell therapy, which has mainly been used against blood cancers, for solid tumors. Anthropic says six Generate scientists used Claude Code to redesign part of the therapy’s genetic instructions. In laboratory tests, one proposed design was about 70 percent shorter than the standard version and enabled immune cells to produce roughly 10 times as many cancer-targeting receptors. The cells also killed tumor cells more effectively and stayed active longer, according to the companies. Generate has published a technical report on the work.
Divergence Labs is pursuing a different approach to food allergies. Rather than simply neutralizing IgE, an immune molecule involved in allergic reactions, the company wants to remove it from immune cells. Anthropic says Claude helped translate the project’s goals into a design plan and generated about 35,000 antibody designs in roughly a month. Initial tests identified three designs that bound to IgE and 20 that could detach it from immune cells. Three of the latter appeared to outperform a standard medication on that measure. The next challenge is combining the two functions into a single drug candidate.
At BigHat Biosciences, Claude worked with scientists to improve two antibody-fragment candidates, one aimed at cancer cells and another intended to help a drug remain in the body longer. Connected to BigHat’s Reccy data system, the AI suggested molecular changes that the company tested in about 1,010 variants over 10 weeks. After the team questioned whether some routine measures were giving a misleading picture, Claude and an Anthropic researcher reanalyzed the underlying data and proposed alternative metrics and a revised test protocol. BigHat says the best candidates bound to their targets more than 10 times as tightly as the originals; one remained dissolved at temperatures 18 degrees Celsius higher, while another clumped more than 100 times more slowly.
Enveda is applying Claude to the paperwork involved in manufacturing drugs and preparing regulatory submissions. The company linked its manufacturing records to a system that lets Claude flag possible problems. In a review of 1,061 pages, the AI identified all 11 issues previously caught by human reviewers and raised seven additional ones, Enveda says. The process cut human review time by 89 percent, from more than 25 hours to under three. Claude also drafted two regulatory documents: one reportedly took four hours to produce, compared with about three months of manual work, while a human review of the other found one minor error.
Anthropic is also working with investment firm The Column Group, which is establishing a new company called Pleiades. The proposed model is to use Claude to run more drug programs, and more complex ones, than a small biotech team could typically manage. Pleiades is developing tools and workflows for the system; its first Claude skill, designed to assess whether a drug target merits further work, has been evaluated across 45 target-and-disease pairings.
Bristol Myers Squibb is using Claude to assess plans for an early-stage clinical drug program and estimate the likelihood that trials will succeed. The companies say they intend to broaden the work in 2027. Michael Pourdehnad, a senior vice president at BMS, said Claude produced an initial analysis from a single prompt that helped the company compare development options in less time than the work would usually take, while keeping sensitive data secure.
Anthropic says it has also acquired Mithril Medicine, a one-person company founded by Cody Tranbarger that was developing treatments for two rare forms of muscular dystrophy. Tranbarger used Claude to coordinate contracts, logistics and experiments with 11 external laboratories, among other operational tasks. One potential treatment moved from an initial idea to promising results in mice in less than 18 months, according to Anthropic. The company says it hopes the candidate could be ready for human trials as soon as late 2028 if further work is successful.
These projects are not yet evidence that AI-developed medicines will work in people. Anthropic says most remain preclinical, and that it plans to extend Claude’s role into clinical development as programs advance. Its broader effort includes research tools, support for life-sciences researchers and a program that lets selected organizations use powerful models with safeguards adjusted for legitimate biological research.
The company says the partnerships are intended both to enhance Claude and to illustrate how AI might fit into drug development. Anthropic has not said that any of the reported candidates has been proven safe or effective in humans. For now, the results offer a series of early tests of whether AI can help researchers move faster through some of the scientific and administrative work that stands between a promising idea and a medicine.
