Anthropic launched Claude Science in San Francisco on Tuesday, a research workbench that wires the Claude model into more than 60 scientific databases and lab tools including PubMed, Jupyter, and R. In the same breath, the company said it would start running its own internal preclinical drug programs, aimed at diseases the pharmaceutical industry has decided aren’t worth pursuing commercially.
The dual announcement is the point. Anthropic isn’t just selling a tool to drugmakers; it’s becoming a drugmaker, at least in the earliest stages, and using its own products to do the work its customers do.
Forbes described the software as a “harness” that stitches existing Claude capabilities to external sources rather than a new model. Pharmaceutical Technology reported it runs locally on Linux or macOS, or against a remote machine. The launch demo, per Forbes, planned and ran a search for a molecule to stabilize the enzyme behind phenylketonuria: 2,200 compounds screened across 80 GPUs, narrowed to 4 candidates, with 100 rare diseases triaged in parallel. MIT Technology Review noted that Claude Science is standalone, unlike the plug-in-based Claude for Life Sciences that Anthropic shipped in October. It’s now in beta for paid subscribers.
The internal-lab move is the more interesting story. “We’re doing this because we believe first and foremost that to build the right models, products and tools to accelerate the industry, we need to live it along with all of you,” said Eric Kauderer-Abrams, Anthropic’s head of life sciences. An Anthropic spokesperson pointed to the company’s public benefit corporation status: “we can choose programs on patient benefit, including work the commercial market overlooks.” Jonah Cool, the company’s head of life sciences partnerships, was also on hand. What Anthropic didn’t address is what happens if something works. Clinical trials cost hundreds of millions and sit well outside a software company’s operating model.
The competitive frame is unavoidable. Northeastern’s coverage positioned Claude Science as Anthropic’s answer to OpenAI’s GPT-Rosalind, released in April, while Google is shipping its own life-sciences tooling.
Researcher reactions gathered by Northeastern were warm but hedged. Michael Pollastri, who works on repurposing drugs for tropical diseases, called it something that “looks like it’s going to be an unbelievable tool” and said it could speed experimentation “by orders of magnitude.” Jared Auclair was more measured, noting that models “can hallucinate or miss nuance in regulatory guidance or assay design — errors that carry real consequences in drug development.” He framed the traditional 10-to-15 year drug timeline as potentially compressible to 2-to-5 years, then added the caveat: “it’s not a shortcut to discovery — it’s a co-pilot that requires a skilled pilot.”
There’s a specific historical rhyme here. In the late 1990s, IBM turned Deep Blue’s chess win into a life-sciences pivot, selling computational biology services while positioning the lab work as proof of seriousness. The tools sold; the drugs didn’t. Anthropic is running the same play with better software and a public-benefit charter attached.
Sources
- https://www.cnbc.com/2026/06/30/anthropic-launches-ai-drug-discovery-program-claude-science.html
- https://www.technologyreview.com/2026/06/30/1139987/claude-science-is-anthropics-newest-flagship-product/
- https://www.pharmaceutical-technology.com/news/anthropic-launches-claude-science-ai-tool-drug-discovery/
- https://news.northeastern.edu/2026/06/30/anthropic-claude-science-launch/
- https://www.forbes.com/sites/johndrake/2026/06/30/anthropics-new-ai-workbench-mapped-my-field-for-26-now-imagine-it-aimed-at-the-rest-of-science/