Michael JohnsonResearch institutes and foundations can now distribute Claude credits among researchers through Anthropic’s AI for Science program. In an update published on October 28, the company named 20 new partners and described projects in genomics, neuroscience, climate science, astronomy, and mathematics. For scientists, this changes how they can access computing resources: support may come through an organization or research initiative, rather than only as an individual program participant.
What changed
Anthropic launched AI for Science in May 2025 as a program offering free Claude credits to scientists. In this new phase, the company is adding collaboration with foundations, scientific societies, institutes, and specific research projects to its individual support. Some partners will be able to distribute credits to researchers, while others will use them within their organization or allocate them to a particular initiative, according to the program expansion announcement.
This is an organizational change, not a separate scientific tool or a new Claude feature available to everyone. Anthropic has not published the credit amount for each partner, the distribution timelines, or common researcher selection rules. The specific route to participation therefore depends on the organization, and details should be confirmed with it.
Who joined and what the resources will support
Named institutional partners include the Wellcome Sanger Institute, the Francis Crick Institute, the European Molecular Biology Laboratory (EMBL), the Space Telescope Science Institute, and the Gladstone Institutes. The listed scientific foundations and societies include, for example, the Alfred P. Sloan Foundation, the Simons Foundation, and the Open Source for Science Fund. Anthropic also named research initiatives that will receive credits for specific projects.
In genomics, the gnomAD project at the Broad Institute supports an open database of human genetic variation; Anthropic says it brings together data from more than 800,000 people. FlyWire at Princeton University develops open data and tools for studying neural connections in the fruit fly brain. Columbia University’s LEAP project uses machine learning alongside physical models to improve climate models of clouds, oceans, and land. These examples show that the program spans work in which large datasets and software tools are part of a sustained scientific process.
Anthropic also described a collaboration between HHMI Janelia Research Campus and the company on the Model Hardware Standard, a specification for safely controlling laboratory instruments with AI agents. The standard is currently in research preview. This example connects computing support to a broader infrastructure challenge: research systems need to interact with physical equipment according to agreed rules.
What the research community could gain
The expansion’s central idea is to move from supporting individual scientists to providing resources that entire teams and organizations can use. Anthropic expects some collaborations to produce publicly accessible outputs, including datasets, open-source software, standards, and shared libraries. Other research teams may benefit from these materials even after a specific project ends, by using the open tools and data in their own work.
The company illustrates this direction with the Allen Institute, which is building teams of AI agents to analyze scientific data and, according to Anthropic, aims to reduce months of manual analysis to hours. This claim describes the institute’s work as presented by Anthropic; the announcement does not provide an independent assessment of the time required for individual tasks or research outcomes. Readers should therefore distinguish the stated goal from an already measured impact.
The most useful practical step for prospective participants is to check opportunities through their laboratory, institute, or scientific society. The announcement lists partners and areas of work, but does not provide a single application form for all researchers. When contacting an organization, researchers should ask who is eligible for credits, which tasks they cover, and whether the project plans to release open data or code.
What is known about the program’s scale so far
The update marks an expansion of the program, but Anthropic describes many projects as just starting or being at an early stage. The company has not disclosed the total amount of credits allocated, the number of researchers who may use them, or a schedule for future results. It says its partners span 20 scientific foundations, institutes, and research groups; that figure alone does not show how many individual laboratories will gain access or how support will be distributed.
For research teams, the change provides a concrete first step: find out whether their institute or a relevant initiative is part of the program, then ask the partner about access rules. For the wider research community, a significant outcome would be projects making tools and materials available to other scientists. For now, the announcement sets out the partner network and areas of work, rather than completed scientific discoveries.
