On October 15, 2026, OpenAI announced the "Rosalind Academic Plan" for university research labs in the United States. Approved labs can use up to 30 seats of ChatGPT Business Standard free of charge for up to 12 months, and can also access GPT-Rosalind, OpenAI's model for life sciences research, after an access review. Even without a universitywide Enterprise agreement, a principal investigator can apply on behalf of the lab. However, there are conditions on who can use the free seats and on how research data is handled, so price alone does not tell a lab which tasks it can entrust to the plan.
Up to 30 seats for literature, data, and analysis code
The Rosalind Academic Plan lets labs use ChatGPT and Codex in a shared workspace. It can be used to search the literature, work with research data, and write analysis code. Members who have been approved and activated also get access to GPT-Rosalind.
GPT-Rosalind is a model that supports research work such as examining biological questions and combining the literature, databases, and scientific tools. OpenAI lists genome analysis, protein and sequence analysis, and experimental design among its uses. What this announcement widens is the route by which labs can start using the model.
Based on OpenAI's announcement and its usage guide, the scope of what is offered is as follows.
| Item | What is offered and conditions |
|---|---|
| Free seats | Up to 30 Business Standard seats. $0 per user per month, for up to 12 months |
| ChatGPT/Codex | Included in Standard seats. Each is subject to its own usage limits |
| GPT-Rosalind | Available to eligible members after approval through access review and activation |
| API | Neither access nor usage allowance is covered by this plan |
What is free is the designated seats, not unlimited use of the model. OpenAI itself states in the announcement that API access is not currently included. Researchers who use the model through ChatGPT or Codex and those who embed it in their own research applications via the API need different approvals.
The GPT-Rosalind guide also distinguishes between the model and "Rosalind Workbench." The latter is a research environment that bundles tools for working with biological structures and sequences, data displays, and work procedures. Being able to use Workbench does not in itself mean approval to use GPT-Rosalind, and simply installing the Life Sciences Research Plugin in Codex does not grant model permissions.
Application routes depend on the university's contract
Eligible applicants are labs at US universities, with the principal investigator (PI) or a designated lab lead as the point of contact. Researchers with no affiliation and commercial organizations are excluded. Japanese university labs are likewise not covered by this free plan.
If the university has no OpenAI Enterprise agreement, the PI or equivalent applies on the lab's behalf. For labs at universities that already have an agreement, access is sought through the university's administrators or OpenAI's account team. Whether a separate Business workspace may be set up must also be checked against the university's policies.
The review requires information about the institution and the lab, as well as the intended research uses and the affiliations of the users. The authority of the PI or others who supervise the research is also checked, and information on a biosafety contact designated by the institution is requested. This is not a program that anyone with a lab email address can simply use.
Identity verification uses Persona, but completing that procedure is separate from approval of the application. After approval, the lab accepts the terms of the offer, sets up the workspace, and then activates GPT-Rosalind use for its members. A payment card must also be registered when the free allotment is set up, and the process includes a step to confirm the seats, term, and pricing terms of the offer being accepted.
PIs and equivalents also handle management after use begins. The verified PI or lead is the sole workspace owner and cannot appoint additional owners or administrators. Simply joining as a member does not confer model access; the PI or equivalent must enable it. When circumstances change, such as a member leaving the lab, permissions should be reviewed and unneeded access removed.
A free allotment that can't be extended to collaborators at other universities
Even with 30 seats, a lab cannot freely invite collaborators from other universities. The usage guide requires all members to use verified institutional email addresses on the same email domain as the approved PI or lead. Visiting researchers and collaborators who use a different institution's domain cannot be added.
In other words, whether someone works on the same research project and whether they can join this workspace are separate conditions. The free allotment can be used for work with literature and analysis code inside a lab, but it will not necessarily provide the same access to every member of a team that spans institutions. If a lab is adopting the plan for collaborative research, participants' affiliations and email domains determine the scope of what can be shared.
The exclusion of the API also affects how work is organized. API use of GPT-Rosalind is granted to separately approved organizations and cannot be substituted by approval for the Academic Plan. Labs planning to call the model from their own analysis tools need to check the terms of use separately from work done in ChatGPT and Codex.
Data not used for training, but retention conditions apply
OpenAI explains that, by default, research data shared under the Rosalind Academic Plan is not used to train its models. However, separate conditions apply to data retention. Labs use a managed ChatGPT Business workspace, and by default the standard retention terms apply.
The usage guide states that settings that do not retain data are not available under this plan. Approval for GPT-Rosalind and the free seats do not change Business data-handling terms. Enterprise protections also do not automatically come with the Academic Plan, so labs need to confirm whether the research data they enter complies with university regulations.
This difference bears directly on decisions about handling unpublished research information. The conditions to check differ between a requirement to avoid use for training and a requirement to limit data retention. Labs also need to factor into their operations that, depending on workspace settings and applicable law, administrators may be able to access data, export it, or delete it.
For what happens after the free period ends, OpenAI advises confirming the end date and terms of the offer that was accepted. The 12-month maximum alone does not allow a lab to assume a uniform extension or post-expiry pricing. Plans for continuing research should include how things will be handled when the free allotment ends.
Newsquawk characterizes the announcement as a move that prioritizes spreading adoption among researchers and building relationships over short-term monetization. However, the outlet itself notes that verifiable details beyond the headline are scarce, so this is not an assessment of actual adoption effects.
For labs, the entry cost has dropped, but responsibility for managing users and data remains. If the members a PI can manage, the conditions on research data, and the handling after the free period all fit, the plan offers a way to try out everything from literature review to analysis code writing in a shared environment.