By Dongchan Kim
September 19, 2026 · 3 min readNovo Nordisk, maker of the weight-loss drug Wegovy, teamed up with Anthropic on September 16, 2026, to point the Claude AI at drug discovery, the companies announced, aiming to develop new medicines faster.
The deal shows AI companies moving from summarizing science to doing it, with models now designing proteins, tuning research code, and running physical experiments.
The Danish drugmaker will first test a tool called Claude Science inside its research and development workflows, focusing on the problems where the partners expect AI to help most.
Novo will also use Anthropic's models to build its own software faster, part of its stated ambition to lead the healthcare industry in AI use.
Reuters reported that Novo already uses Claude to automate clinical trial reports and to shrink patient-document preparation from months to minutes.
Dario Amodei, Anthropic's co-founder and chief executive, said the technology could "compress a century's worth of biological and medical breakthroughs into a decade."
The tool at the center of the deal launched less than three months before the announcement.
Anthropic released Claude Science on June 30, 2026, as a workbench that ties literature search, coding notebooks, biological databases, and remote computing power into one workspace.
Every output carries a record of how the system produced it, Anthropic says, so other researchers can check and reproduce the work.
Stephen Francis, an epidemiologist at the University of California, San Francisco, says the tool cut a brain-tumor analysis to one-tenth of the time it once took.
Anthropic's own experiments suggest the models can do far more than organize research.
In an August 2026 test, Claude designed binders, small proteins that latch onto a disease target, and succeeded against 14 of the 15 targets chosen.
Outside firms Adaptyv Bio and Twist Bioscience built and tested the designs in real laboratories, though Anthropic notes the results come from its own experiment.
The designs worked at rates up to 35.1 percent, well above the 10 to 15 percent range Anthropic calls typical for protein design campaigns.
In a separate September 2026 report, Claude rewrote the code behind more than 30 biology prediction models and sped them up about four times on average.
The job took just under four weeks with two staff members supervising, work that normally demands weeks from a team of engineers for each model, the company said.
Results like those explain why a drugmaker would aim the same models at its own pipeline of medicines.
Still, a prediction on a screen does not equal proof in a living system, and Anthropic acts on that gap.
Two days after the Novo deal, Reuters revealed that Anthropic quietly built a wet lab, a facility for physical biology experiments, in California's Bay Area.
Eric Kauderer-Abrams, Anthropic's head of life sciences, told Reuters that in biology "the final test is still and will be for a while in real lab work."
The company wants Claude to direct robots through experiments with limited human help, though a company spokesperson called human oversight essential for safety.
Anthropic says it will stop short of clinical trials, the human studies every new drug must pass, to avoid competing with customers such as Novo.
The pairing of software deals and physical labs marks a shift from AI that reads biology to AI that tests it.
Neither announcement named a first disease target or a timeline, and Reuters notes that most drug candidates still fail once human trials begin.
