Anthropic has expanded its AI for Science program from individual researchers to 20 named foundations, research institutions and research projects, giving scientists free Claude credits to pursue high-impact work. The expansion, announced on October 28, shifts the program from reviewing single applications to partnering with organizations that can distribute credits or use them across many research groups. Anthropic says many of the collaborations are still in early stages, and it has disclosed no dollar figure for the new partnerships.
Three routes to researchers
The 20 partners fall into three groups. Six funders and scientific societies will pass Claude credits on to researchers they already support, six research institutions will use credits inside their own labs, and six named projects will receive credits directly. Two deeper institutional partnerships sit alongside them: HHMI Janelia Research Campus and the Allen Institute.
The funder group includes the Alfred P. Sloan Foundation, the Simons Foundation, Schmidt Sciences, the Open Source for Science Fund at Renaissance Philanthropy, the US National Science Foundation and CAS, a division of the American Chemical Society. Anthropic notes the NSF alone funds about a quarter of basic research at American colleges and universities.
The institutions using credits internally span astronomy and the life sciences. The Space Telescope Science Institute runs science operations for NASA’s Hubble, James Webb and Roman telescopes, while the Wellcome Sanger Institute, the Francis Crick Institute, Imperial College London’s life sciences department, the European Molecular Biology Laboratory and the Gladstone Institutes cover genomics, biomedicine and disease research.
The six funded projects show the range Anthropic wants to reach. FlyWire at Princeton produced the first wiring map of a fruit fly brain, Columbia’s LEAP center blends machine learning with physics for climate models, Johns Hopkins’ AstroPath adapts astronomy methods to tumor imaging, and the Broad Institute’s gnomAD, Tau Ceti and the Centre for Population Genomics complete the set.
The deepest technical work
The deepest technical work sits with the two institutional partners. With HHMI Janelia, Anthropic co-developed the Model Hardware Standard, a specification that lets AI agents operate lab instruments such as microscopes and liquid handlers through a standard driver. The standard is in research preview, and Anthropic says it plans to open-source it once safety evaluations are further along.
The Allen Institute, the second institutional partner, is building teams of AI agents to analyze its open datasets on the brain, cells and the immune system, aiming to cut months of manual analysis down to hours. Anthropic offers no dataset, paper or timeline for that work beyond the general description, so the claim remains an ambition for now.
Why It Matters
The shift from individual grants to institutions matters because it changes who decides where AI help goes. A funder redistributing credits can reach researchers Anthropic would never review directly, while an institution using credits internally can spread them across dozens of groups at once. The risk is concentration: famous names could crowd out smaller labs.
Anthropic’s science ambitions are not new. The company recently bought a one-person biotech firm to pursue new cures with Claude, showing how far it wants AI to reach into experimental biology. The AI for Science expansion is broader and less speculative: it funds working scientists rather than betting on a single acquisition.
The program also tests a familiar bargain in AI-funded science. Free compute is genuinely useful to cash-strapped labs, but it ties their workflows to one company’s models and terms. Anthropic says resulting work will include public goods such as open datasets, software and standards, which would soften that concern if the outputs really are shared.
Background
AI for Science began in May 2025, offering free Claude credits to individual researchers whose projects promised high impact, judged on scientific merit and how much AI could accelerate the work. Anthropic says it has since supported hundreds of projects across biology, medicine, physics, chemistry, climate science and mathematics.
The wider field is moving fast, with lab automation companies raising money to put robots and AI agents to work on routine experiments. Anthropic’s bet is different in kind: rather than owning the robots, it supplies the intelligence that plans and interprets the work. Whether that division holds depends on standards like the Model Hardware Standard maturing beyond preview.
Source: Anthropic’s AI for Science post
0 Comments