Anthropic Pushes Claude Into Drug Discovery, Starting With Neglected Diseases

As generative AI enters life sciences, its most valuable moment may not be producing polished answers, but whether it can connect sprawling experimental data, molecular models, and computational workflows into testable research paths. Anthropic recently launched Claude Science, an AI work platform for scientists, and said it will use the system for an internal drug discovery program, with an initial focus on neglected diseases that have long lacked commercial investment.

According to information Anthropic released on June 30, Claude Science is positioned as an “AI workbench for scientists,” integrating commonly used research tools, software packages, auditable analytical outputs, and access to computing resources. The platform is currently available in beta to Claude Pro, Max, Team, and Enterprise users, and supports macOS and Linux environments.

The tool’s life sciences capabilities cover a fairly broad range. Anthropic says the platform includes more than 60 curated skills and connectors that can be used for workflows in genomics, single-cell analysis, proteomics, structural biology, chemoinformatics, and other areas. Foreign media reports also noted that Claude Science can help present 3D protein structures, chemical molecular models, and genome browser tracks, allowing researchers to handle data analysis, code execution, and scientific database queries within the same workspace.

What makes this release go beyond a general research assistant tool is Anthropic’s simultaneous disclosure of its drug development ambitions. The Verge reported that Eric Kauderer-Abrams, Anthropic’s head of life sciences, said the company will focus on discovering treatments for neglected diseases; Times of India also reported that Anthropic is launching an internal preclinical drug discovery program and plans to use Claude Science to identify related therapeutic directions. However, it has not yet disclosed which diseases it will target, nor has it clearly explained how subsequent laboratory validation, animal studies, clinical trials, or manufacturing will be arranged.

In practical biomedical use, Claude Science is closer to an integrated platform that can operate research workflows than to a black box that directly “invents new drugs.” Anthropic says early testers have already used it for single-cell RNA sequencing analysis, CRISPR screen design, protein structure prediction, chemoinformatics, and other biomedical research tasks. These tasks are indeed at the core of modern drug discovery: from identifying disease-related targets to designing molecules, evaluating structures, and screening candidates, all could be accelerated by automation and better data connectivity.

But speed is not the only bottleneck in drug development. External drug discovery experts cautioned in related reports that AI can assist many early exploratory stages, but it cannot replace wet-lab experiments, toxicity assessments, and human clinical trials. For regulators and the research community, the key question is not only whether a model can propose candidate molecules, but also whether the data sources are reliable, whether the reasoning process can be traced, whether experimental results can be reproduced, and whether the safety evidence is sufficient to support moving to the next stage.

Anthropic says Claude Science runs on its existing model family and has passed dedicated biosecurity evaluations; this is especially sensitive as AI enters biological design settings. If the platform can lower the barrier for researchers to use computational tools, it may offer practical help in disease areas with insufficient resources; but without public target diseases, validation data, and a collaboration structure, the plan still looks more like a statement of direction than a drug pipeline whose therapeutic prospects can be evaluated.

Background Context

Recently, the role of generative AI in drug development has been gradually extending from literature search, programming assistance, and document organization to molecular design, experimental planning, and data interpretation. Anthropic’s new move pushes this trend toward an end that carries greater responsibility: if AI is to participate in discovering treatments for neglected diseases, it must confront not only algorithmic efficiency, but also who provides samples, who performs validation, who bears clinical risks, and whether the eventual therapies can truly reach the people who need them.