Anthropic is moving from selling AI to drug companies to becoming one. At an event this week titled "The Briefing: AI for Science," the company unveiled Claude Science, an AI workbench that pulls fragmented research tools and datasets into a single environment and generates figures and visuals. Head of life sciences Eric Kauderer-Abrams then said Anthropic itself will develop drugs, starting with neglected diseases.
It is one of the most direct public attempts by a frontier AI lab to become a drugmaker rather than a supplier to drugmakers. Anthropic already counts a long list of biotech and pharma companies as customers of Claude, and it will now be selling software to firms it plans to compete with in the clinic.
The company gave almost no specifics. Kauderer-Abrams did not say which diseases Anthropic will target first, whether it will partner for lab work, animal testing, clinical trials, or manufacturing, or what it will do if Claude surfaces a promising candidate. Anthropic did not respond to requests for further detail.
Key facts
- 01Anthropic announced Claude Science, described as an AI workbench for scientists, at 'The Briefing: AI for Science' event.
- 02Head of life sciences Eric Kauderer-Abrams said Anthropic will develop its own drugs, focused on neglected diseases.
- 03The move puts Anthropic in direct competition with pharma customers already using Claude, and with Isomorphic Labs and Insilico.
- 04Anthropic has spent the last year hiring biologists, building wet labs, and recruiting from Big Pharma and academia.
- 05No AI-designed drug has yet cleared clinical trials and FDA approval — any payoff for Anthropic is likely close to a decade away.
The move drops Anthropic into a crowded field. Google DeepMind spinout Isomorphic Labs, AI-first drug company Insilico, and a wave of biotech startups are already racing to design molecules with machine learning. Big Pharma incumbents including AstraZeneca, Novo Nordisk, and GSK are building or buying AI tools of their own, and rival AI firms OpenAI, Amazon, and Google each have life sciences platforms in market.
The trouble is that "AI drug discovery" is a term that stretches across everything from suggesting new molecules to cleaning up clinical trial data. Namshik Han, a professor at the University of Cambridge and cofounder of AI biotech CardiaTec, called it "a really broad term" and noted that every major drug company is now using AI in some form. Matthew Todd, a professor of drug discovery at University College London, described it as a "catchall phrase."
“AI is applied at "every single stage of drug discovery."”— Namshik Han, Professor at the University of Cambridge and cofounder of CardiaTec
Given Anthropic's frontier models, the presumed play is generative search across chemical and biological space — proposing new molecules against known disease targets, or finding new uses for existing drugs. That is genuinely useful for speeding up early research, and Han pointed to AstraZeneca, Novo Nordisk, and GSK as examples of how AI is already reshaping the front end of the pipeline.
None of that shortens the back end. Frank von Delft, a professor at the University of Oxford and head of protein crystallography at the Oxford Centre for Medicines Discovery, said AI models are advancing quickly but the field still runs on physical experiments. Candidates have to be tested for efficacy, toxicity, and whether they can actually be manufactured, stored, and dosed safely. If Anthropic wants to ship a drug, von Delft said, it "is going to have to spend a lot on experiments."
“haven't yet come close to making experiments unnecessary”— Frank von Delft, Professor of structural chemical biology at the University of Oxford
The hiring pattern suggests Anthropic knows this. Over the last year the company has been building wet labs and recruiting biologists, and it has several live openings for life sciences roles. Han said Anthropic has been actively recruiting from Big Pharma and prestigious academic institutions and has already hired several candidates his colleagues were also approached by.
Payoff, if it comes, is a long way out. Todd said the field is "a long way off" from an AI-designed drug being approved for human use, and that human oversight is required throughout the process. No AI-designed drug has yet cleared clinical trials and FDA approval. Some AI-derived candidates are in trials, but it is often unclear where AI actually contributed or whether those candidates outperform conventionally designed drugs.
There are also data limits. Todd and Han both flagged the shortage of publicly available, high-quality experimental data on how chemicals behave in the body — a gap that constrains even the best models. Trials themselves add years: "It takes time to show experimentally that something's safe," Todd said. A realistic timeline for a first Anthropic drug reaching patients is closer to the back end of the decade than the front end.
For Anthropic, the strategic logic is straightforward even if the science is not. The API business is a race to the bottom on price, coding tools are being commoditized by every lab in the market, and Alibaba just moved to block its staff from using Claude Code. Owning a drug asset — even a single approved neglected-disease treatment a decade from now — is a different kind of moat than a model release cycle. It also gives investors a story that does not depend on next quarter's token pricing. The risk is that Anthropic ends up doing what pharma has always done, at pharma speeds and pharma failure rates, while its software customers watch closely and decide whether to keep paying a competitor.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
