Summary
- Anthropic named six research partners developing medicines.
- Most reported work is preclinical and has not established approved treatments.
- Claude is being used across molecular design, experiments, documentation and trial planning.
Six partnerships across the drug pipeline
Anthropic said it is working with BigHat Biosciences, Bristol Myers Squibb, Divergence Labs, Enveda, Generate Biomedicines and The Column Group on drug-development tasks. The projects place Claude in workflows ranging from early molecule design and laboratory analysis to manufacturing documents and clinical-trial planning. [1] [2]
The company emphasized that most of the work remains preclinical, meaning the potential medicines generally have not reached human testing. Reported improvements in a molecule, assay or planning workflow are therefore early research results, not evidence that a treatment is safe, effective or likely to receive regulatory approval. [1] [3]
Examples from antibodies and CAR-T research
In a BigHat collaboration, an Anthropic researcher and Claude proposed variants of two antibody fragments while BigHat's automated laboratory made and tested about 1,010 designs over roughly 10 weeks. Anthropic said selected versions improved binding, thermal stability and resistance to aggregation, with experimental scientists reviewing the work. [1] [2]
Generate Biomedicines reported an agent-guided effort to shorten an intracellular domain used in CAR-T cells. Its technical report describes a 45-amino-acid design that increased receptor expression and activity in laboratory tests. The results are scientific findings in experimental systems, not a completed clinical trial. [1] [3]
What comes next
Bristol Myers Squibb is using Claude to analyze proprietary data and assess early clinical-study plans, while Enveda connected manufacturing documents to the model to flag possible issues. Anthropic also said it has begun internal programs for rare childhood muscular dystrophies and wants to expand its life-sciences collaborations. [1]
The central question is whether early productivity gains survive the slow, highly regulated stages of validation, manufacturing and clinical testing. Anthropic's announcement identifies promising uses and limitations, but independent replication, peer review and eventual human data will determine whether the approach produces better medicines rather than faster paperwork or more candidates. [1] [2] [3]
Why human and experimental checks remain central
Drug programs fail for many reasons that are difficult to predict from sequence or documents alone, including toxicity, poor distribution in the body, manufacturing problems and weak effects in patients. Laboratory experiments, animal studies where appropriate, regulatory review and controlled clinical trials remain the evidence chain for evaluating any candidate. [1] [2] [3]
The partnerships are therefore best understood as tests of research productivity and scientific reasoning. If AI helps teams identify better experiments or interpret raw data sooner, it may reduce wasted cycles. It cannot turn an early result into a medical claim, and Anthropic explicitly describes the work as early and in need of continued model improvement. [1] [2] [3]
Reporting sources
- Anthropic · Early progress on accelerating drug development with Claude ↗2026-10-27T00:00:00Z
- BigHat Biosciences · BigHat and Anthropic collaboration report ↗2026-10-27T00:00:00Z
- Generate Biomedicines · Refactored intracellular domains enhance CAR expression and function ↗2026-10-27T00:00:00Z
This article follows our editorial standards. To report an error, see our corrections policy.