Anthropic Guest Investigators Program Gives Academics Internal Claude Access and Full Data Rights
Anthropic has started a new program called Guest Investigators. It gives university researchers something they cannot get through a normal paid AI service: the chance to work directly with Anthropic's research team, use AI models that have never been released to the public, and keep the legal rights to their research data and any future research based on it. These rights apply no matter what researchers discover while working with Anthropic.
Anthropic announced its open call for investigators on Wednesday. It invites lead researchers, called principal investigators, and their teams from universities, nonprofit organizations, and government laboratories to apply. Selected teams will work closely with Anthropic for 8 to 12 weeks. Private businesses and startups cannot apply. Anthropic will pay all the costs of running its AI models. Researchers must bring a specific scientific problem they want to solve. They will keep full ownership of their data and can publish their findings or continue related research on their own.
Applications will be accepted for about 14 days starting October 22, 2026. Anthropic says it will try to tell selected applicants about 14 days after the application period ends.
How This Program Is Different From Normal AI Access Grants
This is not a normal grant program. Researchers will not simply receive access codes for Anthropic's AI services and a budget for using them. Instead, Anthropic's own scientists will work directly with outside researchers. Anthropic staff will operate the company's AI models and special research tools to help solve problems brought by the visiting researchers. These tools can search for information, analyze it, and create useful research materials. They are not available in Anthropic's commercial products.
This difference is important because of how the technology works. The Guest Investigators program gives researchers access to what Anthropic calls "models and harnesses." A model is the AI itself. A harness is the collection of tools and steps that helps the AI carry out scientific work. These tools can search scientific databases for useful information, produce organized research results, and work through complicated experiments or analyses in several steps. Much of an advanced large language model's usefulness in a laboratory comes from these extra tools. They are not included in any commercial version.
The program is therefore more like having a visiting scientist join a research team than buying a software license. Outside researchers will work on scientific problems with Anthropic's team, using systems that even well-funded university laboratories could not recreate simply by paying for commercial AI access.
Research Areas and Scientific Goals
The program accepts research in five areas: causal human biology, structural and computational molecular biology, medicinal chemistry, drug discovery, and clinical science.
AI has already shown different levels of success in these fields. The program seems deliberately designed to include both problems where AI already works well and problems that remain very difficult.
In structural biology and drug discovery, AI has already produced results that scientists can measure and check. The important AlphaFold protein structure paper helped change protein research around 2020 and 2021. Computer-based methods for predicting ADMET, which stands for Absorption, Distribution, Metabolism, Excretion, and Toxicity, have also helped several major drug companies shorten the early stages of finding new medicines. ADMET describes how a drug enters the body, moves through it, gets broken down, leaves the body, and may cause harm. Claude can help examine files showing protein structures, reason about how strongly molecules attach to one another, explore ways to make chemical compounds, and search through possible compounds in medicinal chemistry. Scientists have many real measurements in these fields, including crystal structures and laboratory test results. Those measurements help them check whether an AI answer is correct.
Causal human biology is much harder. Scientists need to find out what actually causes a human disease, not just what happens alongside it. To do this, they must rule out other factors that could explain the results. These factors are common in studies that observe people, and testing them directly can be expensive. Scientists have developed methods for studying causes, including Mendelian randomization, instrumental variables, and do-calculus, which was developed by Judea Pearl. These methods have led to important discoveries. However, they depend on having reliable genetic information and on assumptions that must be true for the conclusions to work. AI cannot check all those assumptions by itself. In this field, Claude is most useful for helping plan studies, identify possible hidden factors, and bring together findings from published research. It cannot replace real experiments needed to prove that something causes a disease. The program's 8-to-12-week schedule is especially short for projects that need new experimental data.
Applicants must show two things. First, they need to explain why working directly with Anthropic's scientists would help their research more than simply getting access to an AI model. Second, they must show that their project could produce something useful during the program. This could include early findings, datasets shared openly with others, computer tools for analyzing information, or resources available to the public. Some projects may be allowed to continue beyond 12 weeks.
Applications should contain only information that is already publicly available. Anthropic is not asking researchers to share private experimental data when they first apply.
Why the Rules About Research Data Are Unusual
The program's rules about data ownership and use are unusually clear compared with many other industry programs. Researchers keep the rights to their data throughout the project. They also keep the rights to any separate research they carry out afterward based on the work. Anthropic promises to make related code and data openly available and to publish research results together with the participating scientists.
These rules address a major reason many university laboratories avoid working with commercial AI companies. Researchers may worry that sharing valuable experimental data with a business could lead to arguments about who owns the information or the discoveries made from it. Commercial licensing agreements could also limit what a laboratory is allowed to do with its own research later.
Universities and businesses have dealt with disagreements about intellectual property rights for many years. Intellectual property, or IP, includes inventions and other valuable work protected by law. The Bayh-Dole Act of 1980 gave universities the right to patent inventions made with federal research funding. But ownership of data in AI partnerships raises different legal questions that are not fully settled. For example, who owns the data when an AI model trained or adjusted during a partnership produces a scientific discovery? Anthropic's announcement does not answer every possible legal question. Still, its clear promise that researchers keep their data and can continue independent research goes further than the statements made by many other industry programs.
The rule allowing only publicly available information in applications follows the same idea. Researchers can apply without having to reveal unpublished work before they know whether they have been selected.
How Anthropic Has Been Building Its Life Sciences Programs
The Guest Investigators program offers the deepest level of involvement in a group of life sciences tools and programs that Anthropic has been building since late 2025.
The first major step was the Claude for Life Sciences initiative, launched in October 2025. On June 30, 2026, Anthropic introduced the Claude Science research workbench. This is an AI research system connected to more than 60 scientific databases. It can display three-dimensional protein structures, views of genetic information, and chemical drawings directly. Claude Science is designed to create research materials that other scientists can check. It also helps researchers examine published studies, carry out research involving several steps, improve scientific figures, and prepare materials suitable for publication. The Allen Institute and the Howard Hughes Medical Institute, or HHMI, were announced as founding institutional partners in Anthropic's life sciences effort in February 2026. Both were still working with Anthropic when Claude Science launched.
On August 27, 2026, Anthropic made 10,000 free scientist seats available. It also expanded its AI for Science credit program beyond biology to include researchers in fields that require large amounts of computing power.
In September 2026, Anthropic introduced the Life Sciences Verification Program, or LSVP. This program gives verified life sciences professionals access to Claude Mythos, Opus, and Sonnet models with fewer restrictions on certain biology-related tasks. It is meant for professionals working in drug discovery, clinical development, and biological manufacturing. The program exists because Fable models block some biology questions due to concerns about dual-use technology. Dual-use means the same AI abilities that could help scientists create medicines might also help someone develop harmful biological agents. Under the LSVP, professionals must prove that they work at an eligible institution before receiving access to these less restricted abilities.
The Guest Investigators program goes further than all these earlier programs. The LSVP gives verified professionals access to AI models, and AI for Science provides computing credits. Guest Investigators lets outside researchers work directly inside Anthropic's scientific research process.
Anthropic has also made moves showing that it wants to do more than provide research tools. Earlier in 2026, it bought Coefficient Bio, an AI-based biology company, in a deal reported to be worth about $400 million. Anthropic has also confirmed that it has a drug discovery program working on medicines for neglected diseases before they reach human testing. Together, these efforts suggest that Anthropic has larger plans in life sciences. The Guest Investigators program may be one way to build relationships with the scientific community as part of those plans. The idea is that AI companies that can show real, trustworthy contributions to open academic research may be in a stronger position as the field grows and more people closely examine claims about AI's scientific abilities.
How Anthropic Plans to Work With Outside Scientists
The Guest Investigators program shows how Anthropic has chosen to work with scientists who specialize in fields outside AI.
Most commercial AI programs for science give researchers tools to use on their own. Examples include Google's TPU Research Cloud and OpenAI's academic access program. Anthropic is offering a different approach. Its scientists and outside researchers will investigate the same problems together, share their results, and publish findings as a team. The requirement to share code and publish results openly does more than help researchers. It also gives independent scientists a way to test and confirm what Claude can actually do in real scientific work.
This approach is similar to how drug companies pay university researchers to study their medicines. When independent university scientists carry out a study, their results may be more convincing to regulators and other scientists than results produced only by the company selling the medicine. For Anthropic, working closely with respected life sciences researchers can produce something that records of AI usage and internal performance tests cannot. It can create published research, checked by other experts, showing what Anthropic's models can really do when specialists use them. These studies can show where the models work well and where they fail. Such evidence could help Anthropic earn trust from businesses considering AI for drug development, regulators reviewing the use of AI in medicine development, and government organizations deciding whether to fund AI research.
The clear rules about data rights also help both sides work toward the same goals. By promising that university researchers keep their data and can publish freely, Anthropic creates conditions where findings can be shared even if they reveal problems or weaknesses in its AI models. This gives the research a kind of credibility that high scores on AI performance tests cannot provide.
What Research Teams Should Know Before Applying
For lead researchers thinking about applying, the most important practical questions are what problem they want to solve and how much work the project will involve.
Eight to 12 weeks is not much time for most biology research. The program is meant for projects that can produce early results, useful tools, or datasets rather than final answers to major scientific questions. Teams are especially well suited if they already know exactly where their computer-based research is getting stuck. For example, they might have a data analysis process that becomes too slow with large amounts of information, a problem involving more published research than their current tools can handle, or a protein-design question that requires examining many structures. Strong applicants will also have the laboratory equipment and research methods needed to understand and check Claude's answers. Teams hoping AI will replace experiments needed to prove what causes a disease are unlikely to get that result in such a short period.
Research fields where AI has already helped scientists work faster will probably attract many strong applicants. These include protein structure analysis, cheminformatics, which uses computers to study chemical information, interpretation of CRISPR screening experiments, and single-cell transcriptomics, which studies gene activity in individual cells. Teams studying harder problems, such as what causes diseases or how to find entirely new targets for medicines, may find the program involves more exploration and uncertainty. The program appears designed to allow that kind of work too.
Researchers only need to include publicly available information in their applications. This protects teams that cannot share unpublished data before being accepted. The later research partnership, along with its data-rights rules, covers the stage when researchers might actually share private or valuable information.
Researchers can apply through the Guest Investigators application form, which Anthropic links to from its announcement. Applications will remain open for about 14 days starting October 22, 2026.
Frequently Asked Questions
Who can apply for Anthropic's Guest Investigators program?
The program is open to lead researchers and their teams at universities, nonprofit organizations, and government laboratories. Private companies and startups cannot apply under the current rules. The program focuses on five areas of life sciences: causal human biology, structural and computational molecular biology, medicinal chemistry, drug discovery, and clinical science. Applicants must explain why working directly with Anthropic's researchers would help more than simply using its AI models. They must also show that their project can produce useful results during the 8-to-12-week partnership. The official announcement explains the full eligibility rules and application requirements.
How is this different from other programs that give universities access to AI?
Most academic AI programs, including Google's TPU Research Cloud and OpenAI's academic program, give researchers credits or computing resources that they use independently. Anthropic's program brings outside researchers directly into its research work. Anthropic's scientists work alongside them, using internal AI models and special research tools that are not available in commercial products. The protections for research data are also unusual. Researchers keep full ownership of their data and the rights to any independent research they carry out afterward. Many other industry programs do not explain these rights as clearly.
Does Anthropic gain anything from publishing the results, or is open science only meant to help researchers?
Both sides benefit, and it is important to understand why. Researchers get openly shared code, data, and other results that they can use for future work and publish freely. Anthropic gets scientific studies reviewed by other experts that show what its internal AI models can actually accomplish when experienced scientists use them. These studies can also show where the models do not work well. Evidence from independent university researchers working under open-science rules may be more convincing than Anthropic's own tests. It could help Anthropic gain trust from scientists, businesses considering AI for drug development, and regulators. Requiring open publication is therefore a way for Anthropic to build credibility, not simply a benefit it offers to universities. Researchers should understand this arrangement. It does not make the program less useful for academic research, but it helps explain why the rules protecting data rights are so important.
How does this program fit into Anthropic's larger life sciences plans?
The Guest Investigators program is the most involved level of a group of programs Anthropic has built since late 2025. At the simplest level, the AI for Science credit program gives researchers computing credits to use independently. The Life Sciences Verification Program gives verified professionals access to AI abilities for drug discovery and clinical work that are normally more restricted. Claude Science provides a connected research system with more than 60 scientific databases for partner institutions. The new Guest Investigators program goes further by letting researchers work directly with Anthropic's internal AI models, special research tools, and scientific staff. It also includes joint publication and clear protections for data rights. Anthropic's purchase of Coefficient Bio for about $400 million and its own early-stage drug discovery programs suggest that the company also wants to develop drug candidates itself. This makes Guest Investigators both a way to support scientific research and an investment in relationships with university scientists who may eventually evaluate those drug candidates. More information is available on Anthropic's life sciences initiative page.