Novo is piloting Claude Science on drug discovery problems its own scientists named

The two companies say they will work on drug discovery problems that Novo's scientists and computational teams have identified, and build targeted tools for specific scientific tasks and workflows that support biological reasoning ^[2]. The problem list comes from inside the building, and the tool gets built around it. The first step is a test of Claude Science, the workbench Anthropic launched in June, on specific research and development workflows ^[3][4].

Months earlier, Novo agreed to use OpenAI's AI across the company, according to Bloomberg ^[7]. Novo is therefore running a company-wide licence and a workflow-scoped pilot at the same time, and Mike Doustdar, Novo's chief executive, said the Anthropic partnership builds on the company's AI work "with other technology partners" ^[8].

What separates the two arrangements is how anyone can tell whether they worked. A pilot on specific workflows gives you a checkable answer: the scientists who named the bottleneck either keep using the tool at that step or drift back to what they had. A company-wide licence gives you provisioned seats. Novo will also use Anthropic's frontier models for AI-driven software development ^[5]. That is a second job with its own separate test.

Dario Amodei, Anthropic's chief executive, said "AI's increasing capability brings with it the potential to compress a century's worth of biological and medical breakthroughs into a decade" ^[11]. Novo's commercial pressure is nearer than that. Ozempic and Wegovy face generic competition in the US and Europe early in the next decade, Bloomberg reported ^[18], the shares are down 15% this year ^[19], and in the second quarter Novo held 38.8% of the obesity drug market against 60.9% for Eli Lilly, on CNBC's reading of a Lilly presentation based on IQVIA data ^[16]. That is a gap of 22.1 percentage points, with Lilly's share about 1.57 times Novo's ^[20].

Novo said the collaboration was designed with "robust data governance and human oversight" ^[12]. The phrase carries more weight here than it would in most vendor announcements. Earlier in September Anthropic detailed how Claude had been misused for surveillance and weapons ^[13], and in August its own risk report showed it had run contractor chats with its bioweapon filters off ^[14]. Novo did not publish what the oversight consists of.

For a team weighing the same purchase, the split that helps is between who wrote the problem list and what result would end the pilot. Novo has answered the first on the record: its own scientists and computational teams ^[2]. The second is still open, and the honest measure is whether those same scientists are still using Claude Science at that step of the workflow six months from now.