OpenAI Launches Prism to Empower Scientists with Dedicated AI Workspace

On Tuesday, OpenAI introduced Prism, a new scientific workspace program now freely available to all ChatGPT account holders. Designed as an AI-enhanced word processor and research assistant for scientific papers, Prism is deeply integrated with GPT-5.2. This integration allows researchers to evaluate claims, refine writing, and search for prior studies.

Prism is not intended to conduct independent research without human oversight. Company executives anticipate it will speed up the workflow of human scientists, drawing a parallel to AI-powered coding interfaces such as Cursor and Windsurf.

"We believe 2026 will be for AI and science what 2025 was for AI and software engineering," said Kevin Weill, OpenAI's Vice President for Science, during a press call announcing the tool.

This new web-based application arrives as OpenAI notes a surge in complex scientific queries within its consumer products like ChatGPT. According to the company, ChatGPT receives an average of 8.4 million weekly messages on advanced topics in the hard sciences. However, determining what portion originates from professional researchers remains challenging.

AI-assisted research is also gaining traction in academia. In mathematics, AI models have contributed to solving several longstanding Erdos problems by reviewing literature and applying existing techniques in novel ways. Although the importance of these proofs is still debated, they represent an early success for advocates of AI and formal verification systems.

A December statistics paper utilized GPT 5.2 Pro to establish new proofs for a core axiom of statistical theory, where human researchers primarily prompted and verified the model's output. OpenAI highlighted this accomplishment in a blog post, framing it as a template for future human-AI collaboration in research.

“In domains with well-established theoretical foundations,” the post stated, “cutting-edge models can aid in exploring proofs, testing hypotheses, and uncovering connections that might require considerable human effort to discover.”

Disrupt 2026 Tickets: One-time offer

Tickets are now on sale! Save up to $680 at these limited-time rates, and the first 500 registrants will receive 50% off a +1 pass. TechCrunch Disrupt will gather top executives from Google Cloud, Netflix, Microsoft, Box, a16z, Hugging Face, and many others for over 250 sessions crafted to accelerate growth and hone your competitive advantage. Network with hundreds of pioneering startups and participate in curated events designed to spark deals, insights, and inspiration.

Disrupt 2026 Tickets: One-time offer

Tickets are now on sale! Save up to $680 at these limited-time rates, and the first 500 registrants will receive 50% off a +1 pass. TechCrunch Disrupt will gather top executives from Google Cloud, Netflix, Microsoft, Box, a16z, Hugging Face, and many others for over 250 sessions crafted to accelerate growth and hone your competitive advantage. Network with hundreds of pioneering startups and participate in curated events designed to spark deals, insights, and inspiration.

A significant part of Prism's value stems from its thoughtful application of existing standards. It integrates with LaTeX, the open-source system for formatting and typesetting scientific papers, but offers functionality that surpasses most available LaTeX tools. The program also harnesses GPT 5.2's visual capabilities, enabling researchers to create diagrams from online whiteboard sketches—addressing a common frustration with current tools.

Perhaps its most potent feature merges standard AI capabilities with robust context management. When users open a ChatGPT window within Prism, the model can access the full context of the research project, leading to responses that are more relevant and insightful. 

While a proficient GPT-5.2 user could achieve similar results, OpenAI aims to attract scientific researchers faster with a more streamlined interface. Weill likened this strategy to the factors that made AI tools transformative in software engineering.

“Software engineering accelerated partly due to remarkable models,” he explained to reporters, “and partly due to deep integration into workflows.”