OpenAI launches Prism, a new AI-powered workspace for scientific research

OpenAI on Tuesday launched Prism, a new scientific workspace programme available for free to anyone with a ChatGPT account, positioning the tool as an AI-enhanced word processor and research environment designed to support scientific writing and analysis.

Prism is deeply integrated with GPT-5.2 and can be used to assess scientific claims, revise academic prose and search for prior research. OpenAI said the tool is not designed to conduct research independently and requires human guidance, with the company emphasising that Prism is intended to accelerate the work of scientists rather than replace them. Executives compared the product to AI-powered coding interfaces such as Cursor and Windsurf, highlighting its role as a productivity layer built around advanced models.

Kevin Weil, vice-president for science at OpenAI, said during a press call announcing the launch that 2026 is expected to mark a turning point for AI in science, in a similar way to how 2025 shaped AI adoption in software engineering.

The web-based software arrives as OpenAI sees a sharp increase in scientific and technical queries across its consumer products. The company said ChatGPT receives an average of 8.4 million messages each week on advanced topics in the hard sciences, although it remains unclear how many of those queries originate from professional researchers.

AI-assisted research is becoming increasingly common in academic settings. In mathematics, AI models have been used to solve several long-standing Erdős problems through a combination of large-scale literature review and novel applications of existing techniques. While the importance of some of these results continues to be debated, they have been cited as early validation for AI-driven reasoning and formal verification systems.

In December, a statistics paper used GPT-5.2 Pro to establish new proofs related to a central axiom of statistical theory, with human researchers focused on prompting and verifying the model’s output. OpenAI highlighted the work in a blog post, presenting it as an example of future human–AI collaboration in research. The company said that in domains with axiomatic theoretical foundations, frontier models can help explore proofs, test hypotheses and identify connections that would otherwise require significant human effort.

Much of Prism’s value comes from product integration rather than entirely new capabilities. As reported by TechCrunch, the workspace integrates with LaTeX, the open-source typesetting system widely used for scientific papers, while extending beyond the functionality offered by most existing LaTeX tools. Prism also uses GPT-5.2’s visual capabilities to allow researchers to convert online whiteboard sketches into structured diagrams, addressing a common pain point in scientific workflows.

Another core feature is Prism’s approach to context management. When users open a ChatGPT window within Prism, the model can access the full context of the research project, enabling more relevant and precise responses. While similar outcomes are possible for advanced users of GPT-5.2, OpenAI is betting that a more streamlined interface will help attract researchers more quickly.

Weil said the impact mirrors the trajectory seen in software engineering, where progress was driven not only by stronger AI models but also by deeper integration into everyday workflows.