OpenAI Launches Rosalind Academic Plan for US University Labs – Unite.AI

OpenAI introduced the Rosalind Academic Plan on October 15, 2026, offering eligible academic labs at US universities up to 30 free ChatGPT Business Standard seats for up to 12 months, with GPT-Rosalind access contingent on Trusted Access approval.

The plan bundles seats, priced at $0 per user per month during the free period, with use of every commercially available frontier OpenAI model within the plan’s usage limits. OpenAI says it doesn’t use research data shared through the plan to train its models by default.

What the free places cover

According to OpenAI’s help center guidance for the program, a Standard seat includes ChatGPT and Codex subject to existing usage limits, and the free offering does not remove those limits. A payment card is required when setting up the offer even if the seats do not incur any charges for the 12 months. Offer does not include API credits or API usage, and plan approval does not guarantee access to GPT-Rosalind via the API. Thirty seats is the maximum in a single workspace, and labs that need more are directed to contact OpenAI support rather than create a separate paid workspace, which does not extend the offering or transfer GPT-Rosalind approval.

Participating labs operate in a ChatGPT Business managed workspace where standard data retention applies by default and the offering does not support Zero Data Retention. The principal investigator or verified laboratory manager is the sole owner of the workspace and cannot appoint additional owners or administrators. Workspace participation alone does not grant a member GPT-Rosalind access, which the PI or Lab Manager must individually enable in the workspace administration settings.

GPT-Rosalind and its target research areas

OpenAI describes GPT-Rosalind as its dedicated life sciences model, pairing frontier reasoning with Rosalind Workbench for orchestrating specialized tools in medicinal chemistry, genomics, wet lab support, and other scientific applications. The company’s Rosalind product page describes the Workbench as an orchestrated research preview workspace, offering an Explore mode for general scientific questions and a Research mode that applies GPT-Rosalind to more complex biological questions and advanced research workflows. The model can also be used within the Codex.

Through trusted access, OpenAI says, approved researchers can investigate sensitive scientific questions under safeguards designed for legitimate research. The research areas listed are drug and protein design; disease detection and monitoring; vaccines and immune responses; genetic and RNA therapies; agriculture and animal health; ecology and conservation; and antimicrobial resistance. In the announcement, OpenAI framed the plan around the role of academic research in previous advances, citing the Human Genome Project, the curiosity-driven work on bacterial immune systems that led to CRISPR-Cas9, and decades of academic research on the mRNA behind COVID-19 vaccines.

Eligibility and application process

The plan is currently open only to academic laboratories at US universities. Principal investigators and lab leaders at universities without an OpenAI Enterprise agreement apply on behalf of their labs through the company’s Trusted Access process; University labs with an existing agreement must instead apply for GPT-Rosalind through their university’s Enterprise access process.

Applicants log in with a ChatGPT account linked to an institutional.edu email address, verify that email, and submit their institution, department, lab name, and official website; the institutional profile, role and research objective of the PI or laboratory manager; the intended use of GPT-Rosalind; people who need access with their institutional email and affiliation; any existing workspace, intended access locations and research data management requirements; and the name, title and contact details of a trained biosafety contact at the institution. The help center directs applicants to its office of environmental health and safety, research safety, or biosafety to identify the contact, who must be an institution-designated official with authority over the proposed work; a PI or laboratory manager may fill the role if they meet the requirements and their institution designates them.

The review uses Persona to verify the identity of the university and the PI or lab manager. Completing Persona verification does not mean an application is approved, and OpenAI notes that an existing university email address or subscription does not guarantee approval. Once approved, private investigators and lab managers can activate free access for up to 30 nominated members. Each member must use a verified institutional email address on the same domain as the PI or approved lab manager, so it is not possible to add visiting researchers or collaborators at other institutions. Unaffiliated researchers and commercial organizations are not eligible.

Verification, guarantees and comparison of published accesses

OpenAI states that before activating access it verifies the legitimacy of the institution and the identity of the PI or laboratory manager; reviews the manager’s professional background, institutional appointment, and authority to supervise access to the laboratory; evaluates the laboratory’s affiliation, research activities and intended use; and establishes the responsibilities of the laboratory manager and a designated institutional security contact. Monitoring and enforcement continues after activation, and the company says its controls for high-sensitivity registered use cases take the approved research context into account; when questions arise, OpenAI can contact researchers to clarify the work and confirm permission. Access may be revoked if safety monitoring and responsible use requirements are no longer met.

The announcement also publishes a prompt-based entry example that requires a 20-variant sequence and presentation matrix for a second-generation Nipah antiviral lipopeptide campaign, as well as modeling targets, paired testing, and artifact checks. In the example, GPT-Rosalind produced a candidate comparison matrix, modeling plan, and study plans coupled with controls and interpretive guidance, while a model labeled Mainline Astra withheld sequence-level designs and implementation-level experimental plans, responding that such details would provide actionable care involving high-consequence pathogens and instead provided high-level modeling context and quality artifact controls. OpenAI notes that the published assessments address textual fulfillment of required deliverables and do not verify scientific correctness, feasibility, or whether a method works.

Academic early adopters

The announcement mentions several academic labs already applying GPT-Rosalind in their workflows, in work it describes as ranging from protein design to understanding neurodegeneration. At UCSF, the Zhao Lab is using the model for quantitative drug development, gathering evidence about drug properties to model how drugs move through the body and produce their effects. “GPT-Rosalind helped me collect physicochemical parameters of drugs faster,” the lab said in comments published by OpenAI. “PubChem and NCBI tools helped me work on ten drugs while refining the distinction between predicted and experimentally measured values.”

Dr Mathieu Bourdenx of the Bourdenx Lab said he uses GPT models in his research, from exploring the literature to refining experiments and computational approaches. Dr Brian Hie of Stanford University and the Arc Institute said: “Rosalind will not only help us improve our current discovery and design tasks, but will also apply them at larger scales than previously possible, while providing useful insights to biological researchers.”

OpenAI said it is starting with academic labs as part of a broader effort to put useful artificial intelligence tools in the hands of more scientists, and that by reducing the cost of entry, it hopes to give more researchers space to pursue ideas whose benefits could reach beyond their labs.