Anthropic has established a biological wet lab in the San Francisco Bay Area, expanding its artificial intelligence work in life sciences from literature review, data analysis, and computational modeling into hands-on physical experimentation. The company intends for Claude to help guide laboratory robots, creating a closed loop in drug discovery and rare disease research that covers experiment design, execution, and result interpretation.
From computational research to physical experiments
A wet lab is designed to handle tangible materials such as biological samples, chemical reagents, and cells. With this new facility in place, Anthropic can take research hypotheses, candidate drug targets, or experimental protocols generated by Claude and pass them to laboratory equipment for execution, then adjust subsequent steps based on actual outcomes. Earlier, Anthropic launched Claude Science, a research platform offering tools across genomics, single-cell analysis, proteomics, structural biology, and cheminformatics. The platform can search specialized databases, analyze protein structures, run computational workflows, and retain records of code, data sources, and modifications.
The physical lab fills the validation gap. Candidate molecules or biological hypotheses identified through computational models still need to be tested against real samples and experimental data. Anthropic has not disclosed the lab's size, staffing levels, equipment investment, or specific research projects.
Claude coordinates robotic arms and liquid handling systems
In August, Anthropic released a research preview of its "Model Hardware Standard," an effort to create a unified interface for AI-controlled physical devices. The standard can connect to microscopes, liquid handlers, robotic arms, and microplate readers, enabling Claude to recognize instrument status, sequence operations, monitor experimental results, and adjust parameters. In a published proof-of-concept, Claude coordinated a liquid handler, robotic arm, and detection equipment through the unified interface to complete a protein concentration assay. When tip pickup failures and liquid detection errors occurred during the run, the system managed some anomalies and recalibrated operational parameters based on the results.
At this stage, the model does not yet fully grasp the physical, chemical, and biological constraints required for lab work. For instance, sample foaming can affect pipetting accuracy and optical readings, and the AI sometimes misinterprets physical issues as software faults. As a result, Claude's role in the laboratory remains focused on assisting researchers and coordinating equipment, while critical experiment design, result interpretation, and safety oversight require expert supervision.
Life sciences business expands in parallel
Anthropic has launched a life sciences validation program that grants approved research institutions access to more capable models with fewer restrictions on specialized biology tasks. Eligible applications cover drug discovery, fundamental research, clinical development, and pharmaceutical manufacturing. Applicant institutions must undergo review of research qualifications, safety standards, and ethical oversight. Higher-risk projects require separate permission and more stringent usage monitoring. This framework aims to broaden Claude's applicability in drug development while containing the risks associated with AI handling sensitive biological tasks.
The physical lab, Claude Science, and the equipment control standard together form Anthropic's life sciences technology stack. The company has not yet disclosed specific drug candidates it is advancing, development timelines, or related revenue. Whether the lab can produce commercially viable drug development outcomes remains to be seen.