DOE Test Uses Everstar's Gordian AI to Draft NRC License Application Sections in One Day

The U.S. Department of Energy, working with Idaho National Laboratory, Argonne National Laboratory, Microsoft and Everstar, used AI to convert a reactor safety analysis into sections equivalent to a U.S. Nuclear Regulatory Commission license application. The resulting 208-page document took one day to generate, compared with the four to six weeks a team of people typically needs, DOE's Office of Nuclear Energy said in March 2026. "Now is the time to move boldly on AI-accelerated nuclear energy deployment," said Rian Bahran, Deputy Assistant Secretary for Nuclear Reactors. Advanced reactors demonstrated under DOE's own authorization pathway must later be licensed by the NRC for commercial deployment, and today's licensing process involves multiple rounds of manual document review that can take years. The test case used Everstar's Gordian AI, built on Microsoft Azure, to map the Preliminary Documented Safety Analysis for the National Reactor Innovation Center's Generic High Temperature Gas Reactor into NRC licensing format. Gordian is designed for nuclear-grade technical work: it uses physics and engineering tools and semantic ontology mapping so that outputs are computed and verified rather than inferred. It also flagged missing or incomplete information that would be needed to complete an NRC application. The team follows the principle that experts design, AI accelerates and experts validate; an expert reviewed Gordian's output for accuracy, gaps, consistency, grammar and structure and judged it rigorous, including in how it identified its own data gaps. Next, the team plans a reviewing agent that checks AI-generated documents against NRC guidance before submittal and a benchmarking rubric that gives Gordian's output a confidence grade. An NRIC study estimates AI could cut document development time and regulatory review cycles by as much as 50%. The work falls under DOE's Genesis Mission.