Leaked ChatGPT for Science Plan Targets Research Labs

OpenAI is testing a dedicated ChatGPT for Science subscription, a leaked web build suggests, signaling a move to package ChatGPT for research-heavy users rather than only general chat or office work.

References to the new subscription and experience were spotted in ChatGPT’s web build, according to BleepingComputer. The product name appearing in those references is "ChatGPT for Science", but OpenAI has not announced pricing, eligibility, launch timing, or the exact feature set.

OpenAI tests ChatGPT for Science as a separate research-facing plan

The leak points to a product aimed at science-related workflows, not a cosmetic rename of the standard consumer ChatGPT plan. That matters because OpenAI already splits ChatGPT access across personal, team, and enterprise-style offerings, each with different user assumptions and access rules.

BleepingComputer reports that OpenAI currently offers ChatGPT for personal use, Teams, and business or enterprise use. Personal ChatGPT works broadly for individual users. Teams requires a company domain and at least three users. ChatGPT business is restricted to legal entities.

A science plan would fit that pattern if OpenAI limits access to verified labs, universities, institutes, or other research organizations. That is still not confirmed.

The strongest reading of the leak is simple: OpenAI is preparing a science-branded ChatGPT experience and testing it in public-facing infrastructure before a formal reveal. The weaker reading is that this is an internal experiment that may change or never launch.

Both are plausible. The leak confirms testing references. It does not confirm a finished commercial product.

ChatGPT for Science would push OpenAI deeper into expert workflows

Science is already a visible target inside OpenAI’s model strategy. BleepingComputer ties the leaked plan to GPT-Rosalind, a specialized model built on OpenAI’s GPT-5.5 architecture for enterprise-scale life sciences research.

The key detail is access. GPT-Rosalind is not described as a normal ChatGPT plan. It sits behind what OpenAI calls a "trusted-access deployment structure", with eligibility limited to organizations conducting legitimate, public-benefit scientific research.

GPT-Rosalind is locked behind what OpenAI calls a "trusted-access deployment structure."

BleepingComputer says eligible organizations include major pharmaceutical companies such as Novo Nordisk or verified research institutions. The same report says GPT-Rosalind requires enterprise-grade security and strong safety governance, matching or exceeding ChatGPT Enterprise requirements.

That gives the ChatGPT for Science leak more weight. OpenAI already has a science-specific model track. A subscription could be the packaging layer that brings some of those capabilities to a broader set of institutions.

The likely product question is not whether researchers can ask ChatGPT scientific questions today. They can. The question is whether OpenAI wants a distinct plan with scientific grounding, access controls, and features tuned for research work.

BleepingComputer frames the possible science plan as having “special grounding in discoveries and research around scientific topics” compared with a regular subscription. That phrase is important because scientific users need more than fluent answers. They need outputs that can be checked, challenged, and tied back to source material.

Related XOOMAR coverage of adjacent AI workflow pressure points, including AI Grant Writing Tools Can Save Proposals or Sink Them and ML APIs Break Past Demos in Ray Serve Deployment Guide, shows the same operational tension that would face any research-focused AI product: moving from impressive demos into work where errors are costly.

Access rules and safeguards are the fault line for ChatGPT for Science

The biggest unknown is who gets in. BleepingComputer says it is unclear whether ChatGPT for Science would be available to everyone regardless of background.

That access question is not cosmetic. If the plan inherits restrictions similar to Teams, business, enterprise, or GPT-Rosalind, OpenAI could limit it to verified institutions. If it opens the plan more broadly, OpenAI would need to explain how science-specific capabilities are governed outside formal research settings.

The launch checklist is still empty in public. OpenAI has not confirmed:

• Pricing: No subscription cost has been announced.
• Eligibility: No rules have been published for individuals, students, labs, universities, or companies.
• Model access: It is unknown whether the plan uses a separate science model, a tuned variant, or existing frontier models with a new interface.
• Research tools: Citation features, database integrations, collaboration tools, and file workflows have not been confirmed.
• Data terms: The leak does not say how research data would be stored, used, or protected.

BleepingComputer reports that the feature is being actively tested on the web and says an announcement is likely weeks away. Treat that as a reported expectation, not a launch date.

The practical takeaway is narrower but meaningful: OpenAI appears to be carving out science as its next premium ChatGPT category. Until the company confirms details, ChatGPT for Science is a signal of intent, not a product researchers can evaluate on terms, safety, or cost.

The next thing to watch is the access model. If OpenAI makes it institution-only, ChatGPT for Science becomes a controlled research product. If it opens the door wider, the company will have to prove that a science-branded chatbot can handle expert work without overselling what its models can verify.

The Bottom Line

• OpenAI may be preparing a dedicated ChatGPT tier for researchers and science-heavy workflows.
• A separate science plan could bring new access rules for universities, labs, institutes, or research organizations.
• No pricing, launch timing, eligibility, or feature details have been officially confirmed.