From DNA to space telescopes, Claude enters 20 scientific institutions

Anthropic expands AI for Science to 20 scientific partners, from genomics labs to NASA, to support research with Claude.

Artificial intelligence is gaining ground in laboratories, reaching into human genome analysis, observation of the Universe, and cancer research. Twenty foundations, institutions, and scientific teams are joining the expansion of the AI for Science program, led by Anthropic. The initiative provides access to free credits to use Claude in research projects, with an ambition that now goes beyond individual support for scientists: integrating AI into the infrastructure, instruments, and workflows of major research organizations.

Created in May 2025, AI for Science has already supported several hundred projects across biology, medicine, physics, chemistry, climatology, and mathematics. Researchers notably use Claude to formulate hypotheses, design experiments, analyze complex data, and communicate their findings. This new phase relies on three categories of partners: funding organizations, scientific institutions, and teams engaged in specific projects. From NASA telescopes to genomics labs

Among the new partners is the Space Telescope Science Institute, which oversees scientific operations for NASA's Hubble, James Webb, and Roman space telescopes. The institute also manages vast astronomical archives utilized by the international scientific community. Its integration into the program will allow researchers to access Claude's resources, although Anthropic has not yet detailed the specific concrete applications envisioned.

The biomedical sector is particularly well represented. The Wellcome Sanger Institute, specializing in large-scale genetic sequencing, stands alongside the Francis Crick Institute, which brings together over 100 teams studying the biological mechanisms of diseases. The European Molecular Biology Laboratory (EMBL), Imperial College London, and the Gladstone Institutes are also participating in this expansion. These organizations produce and analyze massive volumes of data, particularly concerning cells, genetic mutations, and living mechanisms.

Six funding organizations and scientific societies are also joining the initiative, including the Alfred P. Sloan Foundation, the Simons Foundation, Schmidt Sciences, and the U.S. National Science Foundation. Their role will be to redistribute Claude credits to researchers supporting selected scientific projects. Understanding the brain, exploring the climate, and detecting rare diseases

Several initiatives offer a clearer picture of the intended uses. At Princeton, the FlyWire project produced the first complete map of neuronal connections in an adult fruit fly's brain. Its open data is used to study how neural circuits generate behaviors and perceptions. Access to Claude contributes to the continuation of this neuroscience work.

At Columbia University, the LEAP center combines artificial intelligence and physics to improve climate models, particularly their representation of clouds, oceans, and land surfaces. In another domain, AstroPath, developed at Johns Hopkins University, borrows from astronomical calibration techniques to analyze tumors cell by cell and compare observations across laboratories.

Genomics constitutes another field of application. The gnomAD database, developed at the Broad Institute, aggregates genetic data from over 800,000 individuals. This open resource helps researchers and clinical laboratories interpret DNA variations, particularly in the study and diagnosis of rare diseases.

Mathematics is not left out of the program. The collaborative Tau Ceti project is building a library of mathematical knowledge formalized in Lean, a language that allows a computer to rigorously verify proofs. When AI agents begin operating instruments

Beyond data analysis, Anthropic is also working on connecting AI models with scientific equipment. In collaboration with the HHMI Janelia Research Campus, the company is developing the Model Hardware Standard, a specification designed to enable agents to interact with various laboratory instruments through a common interface.

This standard, still in the experimental phase, notably applies to microscopes, robotic arms, and automated liquid handling systems. It allows an agent to coordinate multiple devices, monitor measurements, and adjust certain parameters during an experiment. In initial trials presented by Anthropic, teams at Carnegie Mellon conducted dilution and dose-response measurement experiments approximately three times faster than before.

The Allen Institute is exploring another approach by assembling teams of agents tasked with analyzing its vast datasets on the brain, cells, and the immune system. According to Anthropic, certain operations that once required several months of manual analysis can now be completed in a matter of hours. These results pertain to specific projects and do not yet allow for the generalization of such gains across all scientific research. A scientific collaboration yet to be measured

The expansion of AI for Science does not rely on direct funding for laboratories, but primarily on providing access credits for Claude. The program also includes a discounted Team plan for scientists, while certain collaborations focus on developing shared tools and standards.

Anthropic emphasizes that several partnerships are still in their early stages. At this point, no overall assessment makes it possible to measure their contribution to new discoveries. Nevertheless, the company hopes this work will feed into resources accessible to the entire scientific community, including open databases, open-source software, technical standards, and shared libraries.

The evolution of the program thus illustrates a shift in scale in the scientific uses of AI. Following individual assistants capable of aiding in writing or data analysis, experiments are now focusing on systems integrated into the very operation of laboratories. The results of upcoming collaborations will help determine the extent to which this integration truly transforms research practices.