OpenAI launched Prism, a new AI-enhanced scientific workspace, available free to anyone with a ChatGPT account. The tool, integrated with GPT-5.2, allows researchers to assess claims, revise prose, search prior work, and create diagrams, but is not designed to conduct independent research.
Prism incorporates LaTeX and visual capabilities, enabling users to assemble diagrams from online whiteboards while maintaining full context for research projects. OpenAI executives compare Prism to AI coding tools like Cursor and Windsurf, aiming to accelerate human-led research across disciplines. Kevin Weill, VP for Science at OpenAI, said Prism could make 2026 a landmark year for AI in scientific research.
AI-assisted research is already gaining traction. In mathematics, GPT models have helped prove long-standing Erdos problems, and a December statistics paper used GPT-5.2 Pro to establish new proofs with minimal human prompting. OpenAI positions Prism as a model for human-AI collaboration, allowing frontier models to explore proofs, test hypotheses, and uncover connections more efficiently than traditional methods.
At the same time, some researchers warn of risks from low-credibility sources: OpenAI’s ChatGPT has recently cited Elon Musk-backed Grokipedia across multiple topics, raising concerns that AI-generated outputs, even in scientific contexts, could be influenced by unreliable or misleading information.
