In brief
- OpenAI launched Prism, a free LaTeX-based research platform with GPT-5.2 integrated into scientific workflows.
- The launch follows OpenAI statements signaling future outcome-based pricing in research and drug discovery.
- Experts warn of privacy, hallucination, and intellectual property concerns.
OpenAI is expanding into the scientific pipeline with Prism, a new workspace launched on Tuesday in a sign of the company’s clearest bid yet to make its models part of high-value research.
The tool is a web-based application that integrates ChatGPT (5.2) directly into scientific writing, enabling in-place drafting, revision, and collaboration, according to a statement on Tuesday.
“Over the past year, we’ve begun to see AI accelerate scientific work across domains,” OpenAI wrote. “Advanced reasoning systems like GPT-5 are helping push the frontiers of mathematics, accelerating the analysis of human immune-cell experiments, and speeding up experimental iteration in molecular biology.”
In a town hall on Tuesday, OpenAI CEO Sam Altman said the company is already hearing meaningful feedback from scientists about “nontrivial” research progress using its latest model.
“With 5.2, a special version we use internally, we’re now for the first time hearing from scientists that the scientific progress of these models is no longer super trivial,” Altman said. “I can’t believe that a model that can come up with new scientific insights is not also capable, with a different harness and trained a little bit differently, of coming up with new insights about products to build.”
Prism is based on Crixet, a San Francisco-based “LaTeX platform” that OpenAI acquired earlier this month. A LaTeX platform is a specialized writing environment that lets researchers write, format, and typeset scientific papers using code-based commands, making it easier to handle complex equations, citations, and technical layouts consistently.
Privacy, Ownership, and the Limits of AI
For Jonathan Schaeffer, a distinguished university professor emeritus of artificial intelligence at the University of Alberta and co-founder of AI developer Synsira, there are both promising and concerning factors in the use of AI in research.
“There are two issues with writing papers,” Schaeffer told Decrypt in an interview. “One of which is composing the text, and the other is doing the research or making the inferences or the insights that you’re going to add to your paper.”
He said Prism appears to excel at the former in that it helps researchers with writing, proofreading, and citations, which he said is great for literature search as opposed to actually aiding in the research process, which he called “a completely different can of worms.”
In August, research published in Science found that 22% of computer science papers showed signs of artificial intelligence as researchers increasingly turned to the technology.
More troubling, Schaeffer noted, are the intellectual property implications, saying that “the devil is in the details.”
“Standard protocol is, if I’m writing a paper, all I am doing is documenting my scientific research, and it’s my intellectual property, and I own it,” Schaeffer said. “Now, if you’re going to use ChatGPT to write these papers, then you’re actually exposing your intellectual property to a multinational company,” he said, noting additional privacy concerns or whether OpenAI would have any legal right to claim researchers’ intellectual property.
When questioned about the continued issue of AI hallucinations, Schaeffer predicted that “hallucinations will not go away. It will never get down to zero.”
He advocates thinking of AI as “augmented intelligence” rather than artificial intelligence, calling AI models “impressive but fallible.”
“Think of Prism or any of these large language models for research or writing or whatever you’re doing as being your graduate student or intern,” he said. “They can be used to suggest things to you, perhaps a paragraph of text, or perhaps they’re going to spout out a conclusion. They’re going to suggest things to you, but it’s your paper. You have to take responsibility.”
Despite the continued risk of hallucinations, the Prism launch coincides with a strategic pivot by OpenAI’s leadership and a focus on “outcome-based pricing.”
Last week, OpenAI CFO Sarah Friar published a blog post outlining an evolving business model for AI developers beyond subscriptions and API fees.
In the post, Friar wrote that as AI moves into “scientific research, drug discovery, energy systems, and financial modeling, new economic models will emerge.”
“Licensing, IP-based agreements, and outcome-based pricing will share in the value created,” Friar wrote. “That is how the internet evolved. Intelligence will follow the same path.”
While Prism is currently free for personal users, the company’s recent focus on fields like drug discovery suggests a long-term strategy of sharing in the economic value created by the breakthroughs researchers achieve using its tools.
During the town hall, Altman cautioned that, despite recent advances, today’s models still fall short of operating independently in scientific research.
“I think it’s still a long or reasonably long way away from the models doing truly completely closed loop autonomous research in most areas,” Altman said.
OpenAI did not immediately respond to Decrypt’s request for comment.
