Anthropic's Push Into Drug Development Faces a Hard Reality: Experts Warn FDA Approval Could Take at Least a Decade

Since Anthropic's June 30 announcement that it would launch its own preclinical drug-discovery programs, the company has added no new specifics about which diseases it will pursue or how far it intends to take any candidates it finds.

Eric Kauderer-Abrams, Anthropic's head of life sciences, told attendees at the San Francisco event that programs would concentrate on neglected and rare diseases. Jonah Cool, Anthropic's head of life sciences partnerships and deployment, told STAT that the logic is partly structural: rare diseases often trace back to a single damaged gene, making the biology more tractable, even if the commercial economics for traditional drug companies look unattractive.

"These are areas that normal drug development economics don't incentivize or favor," Cool said.

Kauderer-Abrams framed the effort as a feedback loop. "We believe in the power of tight feedback loops, and there's no substitute for having our own experiences alongside you all in the trenches trying to develop drugs," he said, according to CNBC.

The reasoning is coherent. A company whose tools are supposed to accelerate drug discovery has a credibility interest in demonstrating those tools can actually find something worth testing.

What Anthropic Has NOT Said

According to The Verge, Anthropic did not respond to requests for comment on what diseases it will target first, whether it would partner for animal testing, clinical trials, or manufacturing, or what it would do with a promising drug candidate. ZME Science confirmed the same gap: as of the June 30 announcement, the company had described preclinical ambitions with no detail on commercialization or partnering strategy.

These questions are significant for a company that just told the industry it plans to compete with its own customers.

The Expert Verdict: Years, Not Quarters

Multiple researchers have warned that the distance between an AI-generated molecular target and a drug on pharmacy shelves is not a software sprint. Professor Frank von Delft of Oxford University said AI models "haven't yet come close to making experiments unnecessary." Even a correctly predicted novel molecular structure still has to survive years of lab validation, animal studies where required, and phased human clinical trials before FDA review — a point when many promising drug candidates fail. Von Delft said that if Anthropic wants to develop a drug, it is "going to have to spend a lot on experiments."

Experts cited by Kucoin warned that AI-developed drugs still face at least a decade of clinical trials before they can receive FDA approval and reach patients.

Professor Matthew Todd of University College London added that although AI has great potential for rapidly testing new drug ideas, chemical reactions within living organisms are extremely complex, and significant knowledge gaps remain in many areas of medicine that algorithms alone cannot fill. The drug discovery process would not run autonomously, with human input and supervision required throughout.

Namshik Han, a professor at the University of Cambridge and cofounder of AI biotech startup CardiaTec, told The Verge that "AI drug discovery" covers an enormous range of activities, from finding initial compounds to supporting manufacturing. Every major pharmaceutical company already uses AI in some form. Anthropic entering the preclinical space does not automatically mean it is doing something no one else is.

The Competitive Position Is Genuinely Awkward

Anthropic’s plan puts it in an unusual spot. It is selling Claude Science, an AI workbench for researchers and drugmakers, to the same pharmaceutical companies it may compete with in specific disease areas. The Verge and ZME Science both noted this tension, and it is real: a biotech customer licensing Claude for its own rare-disease pipeline now has to weigh whether Anthropic might be working on the same target.

The strongest counter-argument is that Anthropic explicitly said it would pursue areas its pharmaceutical customers are NOT already chasing. If that discipline holds, the competitive overlap could be minimal. But "we won't compete with our customers" is a promise that is easy to make at a launch event and harder to enforce as a drug pipeline evolves.

The China Access Problem Running in Parallel

This drug-development announcement is landing while Anthropic is simultaneously trying to close a different kind of gap. According to the Financial Times, as reported by ZeroHedge, Chinese firms including Ant Financial (roughly one-third owned by Alibaba, which the U.S. government has blacklisted over military ties) and ByteDance have used overseas subsidiaries, VPNs, and corporate accounts routed through Singapore-linked networks to access Claude Code. These workarounds apparently do not violate U.S. or Chinese law, but they do breach Anthropic's terms of service.

Anthropic’s terms of service ban Chinese companies and Beijing-controlled foreign entities from using its models. Anthropic said it prohibits access from unsupported regions including China and continuously updates enforcement systems to detect evasion. The company has specifically targeted "transfer station" services that relay requests from mainland China through overseas-registered accounts.

A company now planning to conduct proprietary preclinical drug research using its own AI models has a sharper interest than ever in keeping those models' outputs out of competitors' hands, particularly state-linked ones.

The Unresolved Question

Anthropic’s has set a high bar for itself: run actual drug-discovery programs, compete in a field where most attempts fail, and do it while selling tools to the companies it may eventually face in the same disease areas. The credibility test will come when the company names a target disease and describes what happens to any candidate it discovers. Until then, preclinical ambitions without a disclosed path to clinical trials or a commercialization partner leave the announcement somewhere between a genuine strategic pivot and a very expensive proof-of-concept.