Novo brings Anthropic’s Claude Science into drug discovery

Novo Nordisk and Anthropic have announced a collaboration to put Claude models, including Claude Science, into selected research and development workflows. The companies say they will focus first on drug-discovery problems identified by Novo scientists and computational teams, with targeted workflows intended to support biological reasoning. Novo will also use Anthropic frontier models for AI-driven software development across the company.

The initial programme is explicitly a testing phase. Novo says Claude Science will be evaluated on specific R&D workflows and scientific problems where the two companies expect the technology to have the greatest impact. Reuters reports that Novo already uses Claude for tasks including trial-report generation and patient-documentation work, so the new agreement moves the relationship closer to core discovery research rather than starting from a blank slate.

This is a significant deployment test for frontier models in pharmaceutical R&D, where useful gains depend on more than fluent text generation. Drug discovery requires models to work across specialised literature, biological data, computational tools and internal scientific workflows while producing outputs that researchers can inspect and validate. Claude Science was built around that kind of tool-connected research environment, including auditable artefacts and access to common scientific computing workflows.

If the partnership produces repeatable gains, the value could come from compressing parts of the research loop such as literature synthesis, hypothesis generation, analysis and scientific software work. It would not remove the slower experimental, clinical and regulatory stages of medicine development. Claims that the collaboration will materially shorten the full path to an approved drug therefore remain prospective.

Novo says the collaboration has been designed with data governance and human oversight, but the announcement does not specify what proprietary datasets Claude will access, how outputs will be validated, what retention controls apply, or whether any patient-level information is in scope. Those details matter in a regulated research environment and should be treated as open implementation questions until the companies publish more technical information.

The companies also provide no benchmark results, deployment milestones, financial terms or measured reduction in discovery timelines. Anthropic CEO Dario Amodei describes the broader potential of AI to compress biological and medical progress, but that is an ambition rather than evidence from this partnership. The confirmed fact today is the collaboration and its intended scope, not a demonstrated acceleration of drug development.

The useful next signals will be named production workflows, comparisons with existing scientific processes, independently reviewable or peer-reviewed results, and details about validation and compliance controls. Evidence that Claude Science improves a specific research task without degrading scientific reliability would make this partnership more consequential than the announcement alone.