Anthropic Conducts Drug Development Collaborations with Six Organizations
Anthropic’s official blog disclosed that over the past few months its scientists have carried out in-depth research collaborations with six new drug development organizations: BigHat Biosciences, Bristol Myers Squibb, Divergence Labs, Enveda, Generate Biomedicines, and The Column Group, and has also launched its own drug discovery project, planning to substantially expand related work on this basis.
At Generate Biomedicines, six scientists used Claude Code to propose new designs for the signaling domain of CAR-T cell therapy, with one design about 70% shorter than the standard version. Laboratory tests showed that the design enabled immune cells to produce about 10 times as many receptors, killing tumor cells more efficiently, while the activity of the immune cells lasted longer. This move aims to shorten the genetic instructions to free up space, advancing CAR-T from blood cancers to the treatment of solid tumors.
The remaining collaborations cover antibody optimization, IgE targeting, review of production documents, and clinical trial planning, among other areas, and are currently mostly at the preclinical stage. Anthropic said the relevant results are still early, and it is continuously expanding Claude’s role across the entire drug development process.
Original sources
Anthropic official blog, Early progress on accelerating drug development with Claude (2026-10-27) https://www.anthropic.com/research/early-progress-drug-development
Generate Biomedicines technical report and blog, Less is More: Refactoring the CAR Signaling Domain https://generatebiomedicines.com/blog-posts/less-is-more-refactoring-the-car-signaling-domain/
ABAB AI Insight
This collaboration reflects a shift in AI models from assisting with literature retrieval and molecular design to directly participating in laboratory iteration and operational closed loops. In the Generate case, the signaling domain refactoring proposed by Claude Code and refined through multiple rounds of screening shortened the length while improving expression and function, showing the compression and optimization capabilities of generative models in exploring protein sequence space. On the capital and R&D side, big pharmaceutical companies (BMS) and specialized biotech firms are both being integrated in parallel, lowering the threshold for early validation; Anthropic’s own project (such as Mithril Medicine) embeds the model in one-person/small-team operations and coordination with external labs, shortening the cycle from concept to animal experiments. Current data remain preclinical, and actual drug development success rates and regulatory pathways still need to be validated by subsequent clinical data.
Cognitive laws
The model proposes designs → the laboratory validates them → iteration follows, shortening the molecular optimization cycle rather than replacing the final decision.
Shorter length in exchange for functional gains shows that in sequence space, “shorter instructions can carry higher expression.”
Parallel access by multiple organizations distributes a single model’s capabilities across different bottleneck points in drug development.