Artificial intelligence has already trimmed years from some Nuclear Regulatory Commission licensing reviews. One category that once stretched four years now wraps in nine months, according to agency data chief Basia Sall.
Sall, who serves as Chief Data Officer and Deputy Chief AI Officer, shared the update June 25 at the ATARC AI for Impact Summit in Virginia. She tied the gains to a May 2025 executive order that set an 18-month target for certain reviews. AI tools pushed results even faster.
“I’m happy to report we’ve already reduced the amount of time it takes for licensing,” Sall said. “For example, one type of licensing would take four years. We said we’re going to get it down to 18 months. We just finished that first round of that licensing in nine months.”
She added that agency AI specialists saw room for more compression. “I think some of our AI gurus at our agency were like, ‘Oh, yeah, we can do it better.’”
The NRC pairs the technology with process changes. Staff now use AI to draft documents while checking precedents from prior decisions. Industry partners curate public NRC data into stronger applications, which arrive with fewer questions and clearer structure. Sall noted the agency receives submissions that move through acceptance and into review with less back-and-forth.
Internal development produced SimplifAI, built on Azure OpenAI. The tool supports regulatory documents and has reached version 2.0. NRC staff also train it for speech writing. The agency’s AI use-case inventory describes the system as boosting efficiency and consistency across licensing, oversight, and other activities.
Broader federal offerings speed adoption. NRC tests Anthropic’s Claude, Azure OpenAI, and Google Gemini through the General Services Administration’s OneGov program, which offers discounted access. Early results on public data look promising. The agency is also exploring the USAi testing platform.
Parallel work outside the regulator points to similar compression. In March 2026 the Department of Energy, Idaho National Laboratory, Argonne National Laboratory, Microsoft, and Everstar ran a test case. Everstar’s Gordian AI solution converted a 208-page safety analysis document from DOE format into NRC-equivalent sections in one day. The same task normally requires a team four to six weeks, according to the Energy Department.
The tool identified missing or incomplete information needed for a full NRC application. Experts then reviewed output for accuracy, consistency, grammar, and structure. Quality held up. “Experts design, AI accelerates, and experts validate,” the DOE summary stated.
A National Reactor Innovation Center study cited in coverage projected up to 50 percent cuts in both document preparation and regulatory review cycles, with gains in accuracy, consistency, and traceability. INL and Microsoft have run related pilots on engineering and safety analysis reports.
Policy support continues. The March 2026 finalization of 10 CFR Part 53 created a technology-neutral framework aimed at faster, safety-focused reviews. Executive orders and the ADVANCE Act reinforce timeline discipline. NRC staff explore human-in-the-loop workflows that could halve review times for new designs.
Industry tools multiply options. Atomic Canyon’s Neutron searches the NRC’s ADAMS database. Nuclearn automates workflows. Blue Wave AI Labs targets performance forecasts. Inductive positions itself as “TurboTax for nuclear licensing.” Argonne National Laboratory works on agent-based systems that simulate applicant-regulator exchanges.
Sall emphasized measured rollout. The agency maintains an inventory and governance structure to track use. Early internal pilots focused on low-risk tasks before expanding. Collaboration with other agencies and vendors supplies lessons without overreach.
Results appear in concrete cases. One licensing round finished in nine months instead of the prior four years. Applications arrive cleaner. Draft documents surface faster. Precedent checks run automatically. None of this replaces human judgment on safety questions.
The pattern shows incremental gains compounding. Regulatory reforms set the floor. AI tools lift performance above it. External demonstrations prove the approach scales beyond the agency. Staff confidence grows with each validated output.
Further compression remains possible. Sall’s team continues testing models and refining SimplifAI. DOE partners plan reviewing agents and confidence rubrics. National labs develop specialized training on nuclear terminology and simulations.
Timeline pressure comes from rising demand. Data centers and grid needs push developers toward advanced reactors. Shorter, more predictable reviews reduce financing risk and construction delays. The NRC’s experience shows technology can deliver without compromising standards.
Public data and vendor partnerships accelerate the shift. Companies train on agency records to submit stronger packages. The regulator spends less time on clarification and more on core analysis. Both sides report fewer rounds of questions.
Internal pride shows in tool development. SimplifAI moved to version 2.0 after the first model aged out. Staff experiment with speech assistance alongside document work. The agency treats the collection of available models as a menu rather than a single solution.
Broader federal infrastructure supports the effort. OneGov orders exceed 120 across agencies. USAi testing involves more than 25 organizations already. Discounts and shared platforms lower barriers for smaller teams.
Safety remains the fixed point. Every AI output faces expert review. Gaps in data knowledge get flagged. Consistency checks run against guidance documents. The approach mirrors long-standing regulatory practice of verification layered on assistance.
Early numbers tell the story. Four years to nine months on one track. Weeks to a day on document conversion. Projections of 50 percent cycle reductions. These figures come from agency statements and DOE test results, not projections alone.
The NRC’s measured path—governance board, inventory, pilot limits—keeps adoption aligned with mission. Sall’s update captures the current state: progress achieved, tools expanding, targets still in sight.
WebProNews is an iEntry Publication