Anthropic has expanded its AI for Science program by providing Claude credits to 20 scientific foundations, institutions and research projects. The effort broadens access to the company’s models for teams working in fields including biology, medicine, physics, chemistry, climate science and mathematics, while moving the program beyond its earlier focus on individual researchers.
For research organizations, model credits can remove part of the cost barrier to experimenting with AI-assisted hypothesis generation, experiment design, data analysis and scientific communication. Anthropic said participating scientists are already using Claude for those tasks. The company did not disclose the volume of credits, the duration of access or the governance terms that apply to each partner.
The program began in May 2025, according to Anthropic. The new institutional partners include the Alfred P. Sloan Foundation, Simons Foundation, Schmidt Sciences, the U.S. National Science Foundation and CAS, a division of the American Chemical Society. Anthropic said those groups will distribute the resources to researchers pursuing high-impact scientific work.
Other participants include the Space Telescope Science Institute, Wellcome Sanger Institute, Francis Crick Institute, Imperial College London’s Department of Life Sciences, the European Molecular Biology Laboratory and Gladstone Institutes. The list also includes specific projects such as FlyWire at Princeton University, the Learning the Earth with Artificial Intelligence and Physics center at Columbia University, AstroPath at Johns Hopkins University, gnomAD at the Broad Institute and the Tau Ceti mathematics project.
Anthropic said many of the collaborations may produce public assets such as open datasets, open-source software, open standards and shared libraries. That focus is notable for IT teams supporting research computing because reusable data and software components can affect how a research organization manages access, provenance, integration and long-term maintenance. The company did not state which outputs will be open or provide delivery dates for individual projects.
The expansion also builds on work with the Howard Hughes Medical Institute’s Janelia Research Campus on the Model Hardware Standard, which Anthropic describes as a shared specification intended to let AI agents operate laboratory instruments safely. The standard remains in research preview. The company said it also supports the Allen Institute, where teams are building AI agents to analyze large datasets used in brain, cell and immune-system research.
AI-assisted research introduces governance questions that differ from a simple chat interface. Research teams may be handling unpublished findings, sensitive health data, specialized instruments or valuable datasets. Model access therefore needs to be paired with controls over data use, permissions, validation and reproducibility. Anthropic’s post describes research goals and partners, but it does not specify how those controls are implemented across the individual collaborations.
The program is an example of an LLM vendor extending a model platform through institutional access rather than a standalone product release. Its immediate benefit for participating teams is access to Claude capacity for defined research work. The longer-term value will depend on whether individual projects can turn that access into validated methods, shared infrastructure or research outputs that other organizations can use.
Anthropic said it plans to keep expanding the program and to share the impact of the collaborations. For enterprise and research-IT leaders, the initiative underscores that deployment planning for advanced models increasingly includes specialized scientific workflows, not only general business productivity or software-development use cases.