AI Development· Analyse
Claude in drug research: early results, no proven medicines
Anthropic describes six research partners and early applications of Claude. Laboratory results and faster documents are not yet clinical evidence.
Anthropic describes early experiences with Claude in drug research on 27 October 2026. The company is working with six organisations on research into targets, molecular design, production, documentation and planning of clinical trials. The announcement is a research update, not an announcement of an approved medicine [1].
Not one bottleneck, but an entire chain
The partners are BigHat Biosciences, Bristol Myers Squibb, Divergence Labs, Enveda, Generate Biomedicines and The Column Group. Anthropic states that acceleration need not only come from new molecules. Analysis of laboratory data, preparation of documents and operational tasks can also be part of the research [1].
At Generate Biomedicines, six scientists used Claude Code to propose designs for a component of CAR T-cell therapy. Anthropic describes positive laboratory results for one design. That is not the same as demonstrated efficacy in patients. Divergence Labs used Claude for designing antibodies, after which candidates were selected and tested. Here too, these are intermediate steps, not a completed treatment [1].
Documents and human oversight
Enveda linked production documents to Claude to flag potential problems. In the described evaluation, the system found the eleven problems that human reviewers had identified and flagged seven additional problems. Anthropic also reports drafts for regulatory documents. These figures come from the partner research that the supplier discusses; we have not repeated that evaluation ourselves [1].
According to the announcement, Bristol Myers Squibb uses Claude to assess plans for a drug in early clinical development. Anthropic says it wants to expand the collaboration in 2027. This is an intention, not evidence that the likelihood of successful studies has already been increased [1].
Significance for research teams
Our editorial interpretation: these examples are useful for identifying concrete work steps in which researchers can test support. A faster draft document or an interesting molecular variant is not yet a clinical outcome. Check methods, data quality, reproducibility and human assessment before applying results.
Read more
Also see Claude in AI for Science.
Sources
This article was produced with AI assistance from the original sources cited and checked against our editorial policy. Editorial policy