How Anthropic is using Claude to fast-track drug development
From cancer research to clinical trials, Anthropic’s Claude is entering drug development, helping biotech firms tackle research, rare diseases and regulatory paperwork.
Anthropic is testing how its AI assistant, Claude, can help develop real medicines, going beyond answering questions and writing code.
The company says its life sciences team is working with six organisations: Bristol Myers Squibb, BigHat Biosciences, Generate Biomedicines, Divergence Labs, Enveda, and The Column Group. Anthropic has also started its own drug discovery work.
The goal is not to replace scientists with a chatbot. Instead, Anthropic wants to see whether AI can speed up the difficult process of finding a promising target for a drug, designing molecules, preparing official documents for regulators, and checking records about how medicines are made.
Claude helps design drugs
Developing a drug takes more than finding a molecule that works. Researchers must improve its design, test it, make it consistently, and show that it is safe before it can reach patients.
Anthropic is testing Claude at different stages of this process. Some early results show where AI might help.
At Generate Biomedicines, six scientists used Claude Code to suggest new designs for part of the genetic instructions used in CAR T-cell therapy. This treatment changes immune cells so they can attack cancer.
Anthropic says one design was about 70% shorter than the standard version and worked better in laboratory tests. It helped immune cells produce about 10 times as many receptors.
At Divergence Labs, Claude helped create possible antibody treatments for food allergies. The idea is to separate IgE, a molecule involved in allergic reactions, from immune cells. Anthropic says Claude produced around 35,000 antibody designs in about a month.
Researchers then selected designs for testing. They found some that could attach to IgE and others that could push it off.
These results are still early, but they show how AI could help researchers explore many more ideas before choosing which ones to study more closely.
AI also handles paperwork
Laboratory work is not the only thing that can slow down drug development. Preparing documents, checking quality, and sending information to regulators can also take specialists many weeks.
At Enveda, Anthropic says Claude reviewed 1,061 pages of documents about manufacturing. It found all 11 problems identified by human reviewers and flagged seven more. The review took 89% less time, falling from more than 25 hours to under three hours.
Claude also prepared drafts of documents for regulators. One document reportedly took four hours to prepare with Claude, compared with about three months when done manually.
These results suggest that AI could do more than help discover new treatments. It could also reduce the time researchers spend checking documents, giving them more time to study scientific evidence.
From early experiments to real medicines
Anthropic is also working with Bristol Myers Squibb to plan clinical trials, which test treatments in people. Separately, it has bought Mithril Medicine, a small drug discovery project focused on rare forms of muscular dystrophy, a group of diseases that cause muscles to become weaker.
The project has shown promising results in mice. If the possible treatments continue to show good results, a trial in humans could begin as early as late 2028.
There is an important limitation: most of this work is still at the preclinical stage, which means the treatments have not yet reached testing in humans. Better designs and faster paperwork do not automatically lead to medicines that are safe and effective.
Claude may help researchers work faster, but the real test is whether these improvements continue to help throughout the long and difficult process of developing a drug. In medicine, a promising experiment is only the beginning, not the finish line.