On October 27, 2026, Anthropic published a progress report on Claude drug development. For several months the AI company has been working with six organizations that develop new medicines, and it now runs research programs of its own. The reported results range from shorter gene components for cancer therapies to time savings on regulatory paperwork. Almost all of it is still preclinical, meaning it has not been tested in humans.
We summarize what Anthropic actually reports, which numbers come from whom and where the limits of the claims lie.
Key Takeaways
- The partners are BigHat Biosciences, Bristol Myers Squibb, Divergence Labs, Enveda, Generate Biomedicines and The Column Group.
- Claude is used not only for molecule design but also for lab planning, data analysis, manufacturing documents and regulatory drafts.
- Anthropic has acquired Mithril Medicine, a one-person company working on treatments for two rare forms of muscular dystrophy.
- Most results are preclinical. Anthropic expects human trials from its own program to start no earlier than late 2028.
- All figures come from Anthropic and its partners, partly backed by technical reports and preprints, but without independent peer review.
Anthropic’s Strategy for Claude Drug Development
In its research post “Early progress on accelerating drug development with Claude”, Anthropic describes its approach. The bottlenecks in drug development are not in one place but spread across the whole process: finding a target, designing a molecule, optimizing it, manufacturing it reliably, writing regulatory filings, and designing and running clinical trials. Instead of limiting itself to the best-known application, proposing new molecules, Anthropic wants Claude involved in as many of these steps as possible.
To that end, the team works inside real drug programs. Anthropic says this serves three goals: understanding the strengths and limits of its models, providing case studies for the industry and, ideally, contributing directly to effective medicines, for example against food allergies, rare childhood diseases and several forms of cancer.
The Case Studies at a Glance
Generate Biomedicines wants to make CAR T-cell therapy, which has mainly helped against blood cancers, usable for solid tumors. To do that, part of the therapy’s genetic instructions needs to be shorter to make room for additional instructions. Six Generate scientists used Claude Code to design new variants. According to Anthropic, one design was about 70 percent shorter than the standard version and more effective in lab tests: the immune cells made roughly ten times as many receptors, killed tumor cells more efficiently and stayed active longer.
Divergence Labs is working on a new approach to food allergies. Instead of only neutralizing IgE circulating in the blood, its drugs aim to detach IgE from immune cells entirely. Claude turned the goals into a design plan and, largely autonomously, generated roughly 35,000 antibody designs in about a month. Early tests found three candidates that bind IgE and 20 that could dislodge it. Three of those 20 appeared to dislodge IgE more than the standard medication.
BigHat Biosciences combines AI with an automated lab. Claude was connected to BigHat’s lab data system and proposed changes to two antibody fragments. The lab made and tested about 1,010 variants in roughly ten weeks. Along the way, the team suspected that the standard heat-stability and clumping metrics were misleading. Claude and an Anthropic researcher reanalyzed the raw data and proposed different metrics. According to the report, the best variants bound more than ten times more tightly, one stayed in solution at a temperature 18°C higher, and the other clumped more than 100 times more slowly.
Enveda used Claude for manufacturing records and regulatory filings. Its chemists had spent up to 75 percent of their time on paperwork. In tests on 1,061 pages, Claude caught all eleven issues that human reviewers had found and flagged seven more they had missed. Human review time fell by 89 percent, from over 25 hours to under three. Claude wrote the first draft of one important regulatory document in four hours instead of about three months.
The Column Group, a biotech investor, is setting up Pleiades, a company meant to run far more drug programs with Claude than a small team normally could. Pleiades’ first Claude skill assesses whether a drug target is worth pursuing. It has been evaluated on 45 target-disease pairs.
Bristol Myers Squibb recently started using Claude to assess clinical trial plans for an early-stage drug and estimate their likelihood of success. The collaboration is set to expand in 2027.
In-House Research: The Mithril Medicine Acquisition
Alongside the partnerships, Anthropic is doing its own research. The company acquired Mithril Medicine, a one-person startup focused on two rare forms of childhood muscular dystrophy. According to Anthropic, founder Cody Tranbarger used Claude to coordinate contracts, logistics and experiments with eleven external labs. That took a possible treatment from idea to promising results in mice in less than 18 months. If the candidates keep holding up, a human clinical trial could start as soon as late 2028.
What the Results Do Not Show (Yet)
As impressive as some numbers sound, they should be read with care:
- Preclinical: Anthropic itself says the work is mostly preclinical. Lab successes and mouse data often fail in later stages of drug development.
- Self-reported: the results come from Anthropic and its partners, who have an interest in positive accounts. There are technical reports and preprints, but no peer review yet.
- Humans stay in the loop: in every case study, scientists work alongside Claude and run the lab experiments. This is not autonomous drug development.
- Baseline: claims such as “ten times tighter binding” compare against the respective starting molecule, not an approved drug.
Earlier announcements come to mind. When Anthropic reported a new CRISPR-like enzyme system in September, expert reactions were mixed. The debate about biological risks also remains relevant, as the UK’s AI biosecurity lab shows.
New Offerings for Researchers
Anthropic wants to make its models available to as many researchers as possible. It names Claude Science, an AI workbench for scientists, Claude credits for life scientists and the Life Sciences Verification Program. Through that program, trusted organizations can use the most powerful models with more permissive safeguards, because the standard filters can impede legitimate biological research. There is also an open call for guest investigators.
The relaxed safeguards program in particular is likely to raise questions. Anthropic itself justifies the safeguards as necessary to prevent misuse, and the post does not describe in detail how “trusted organizations” are vetted. There is also the business context: Anthropic is preparing to go public, as we covered in our article on the Anthropic IPO and its S-1 filing. Success stories from medicine fit this phase well.
Conclusion
The Claude drug development update is more concrete than many AI announcements of recent months. It names partners and numbers and points to technical reports. The unspectacular results, such as faster review of manufacturing documents, may deliver the biggest practical benefit in the short term. Whether new medicines actually come out of it will only be decided in clinical trials, and for most programs those are still years away. We will follow how independent experts assess the preprints.
