Anthropic has published early drug-development case studies from partnerships with six organisations, alongside research programmes of its own. The report describes Claude supporting scientific and operational tasks, while stressing that most of the work remains preclinical.
Different bottlenecks need different evidence
The partners include BigHat Biosciences, Bristol Myers Squibb, Divergence Labs, Enveda, Generate Biomedicines and The Column Group. Their projects span molecule design, optimisation, documentation and clinical-trial planning.
Those tasks should not be evaluated as if they were one result. A useful document draft, a promising laboratory measurement and an effective treatment are different outcomes. Anthropic presents early progress towards drug-development goals rather than evidence that Claude has delivered new approved medicines.
Laboratory work and paperwork examples
At BigHat, the report says scientists and an Anthropic researcher worked with Claude while a laboratory made and tested about 1,010 variants over roughly ten weeks. The work included revisiting potentially misleading measurement criteria, with scientists refining and running a changed protocol.
At Enveda, Anthropic reports tests on 1,061 pages of manufacturing documentation. Claude flagged all eleven issues previously identified by human reviewers and seven additional issues, while human review time fell by 89% in that evaluation. These are results reported by the participants, not an independent audit or a general guarantee for every manufacturing record.
Human testing remains a separate milestone
Anthropic describes much of the research as not yet involving drugs tested in humans. Its own muscular-dystrophy work includes promising mouse results and a possible human-trial timetable conditional on candidates continuing to hold up.
The practical reading is to retain the stage and endpoint of each result. Teams assessing similar work can begin with a defined task and existing reference material, then measure errors and review effort. Laboratory or documentation gains may support further development, but clinical benefit and safety still require their own evidence.