AI Goes Regulatory: The NRC’s New Reality

For years, AI application in nuclear has been a story about the future and potential. Smarter reactors and true digital twins while predictive maintenance creeps along. But last week, at the NRC’s Regulatory Information Conference (RIC), AI was no longer spoken of as a pipe dream. It was presented as a way to climb mountains of paperwork and fine-tune risk modeling.

And that… that’s a very big change in this industry.

The First (Blue) Wave: From Cool Tech to Nuclear Red Tape

When AI first dipped its virtual toe into nuclear, it was all about optimizing reactor performance and crunching big datasets. But now? It’s wading into the rapids of regulatory compliance—where the most brilliant nuclear engineers spend their days filling out monotonous forms instead of building the future.

Take 50.59 screenings, for instance—the process that determines whether a nuclear plant can make a change without triggering a full-blown NRC approval marathon. Normally, this means engineers plowing through reams of historical documentation, regulations, and technical specs. Now, thanks to Blue Wave AI Labs, this process is being automated by AI-based tools. Their system scours NRC licensing databases and past rulings, drafting 50.59 screenings in minutes instead of days.

This is regulatory intelligence on steroids. And it’s coming to the rescue at just the right time.

It raises a fascinating question: If AI can justify nuclear plant modifications with regulatory precision, how long before it starts making the rules instead of following them? (Probably a while.)

Westinghouse’s AI Wants to Write the NRC’s Homework

Of course, Westinghouse wasn’t about to let Blue Wave own the regulatory AI space. They rolled out Bertha, their nuclear-specific generative AI model, designed to crank out licensing documentation like a caffeinated intern. Bertha it writes first drafts of submittals, flags compliance gaps, and even suggests language that regulators are statistically more likely to approve.

If that last part doesn’t make you pause, it should. But don’t worry. Westinghouse is bending over backward to make Bertha explainable, incorruptible, and fully compliant with AI governance frameworks.

The NRC Is Already Using AI—And They’re Not Shy About It

So where does the NRC itself stand on all this? Turns out, they’re already using AI—but in a way that’s more… tentative than assigning Bertha to write wholesale licensing justifications. Their Data Warehouse project is quietly deploying machine learning and natural language processing to process regulatory filings, inspection reports, and event reports faster. One example? AI-enhanced Licensee Event Reports (LERs). Traditionally, these are human-written PDFs that describe operational incidents. Now, AI is parsing them into structured data, allowing regulators to spot trends across reactors faster than ever.

Fusion, Risk Models, and AI Creeping In From Every Angle

AI’s fingerprints were all over the conference—even in fusion energy regulation. While fusion still has one foot in the future, regulators are already exploring data-driven licensing approaches, which means AI will eventually play a role in risk assessments for commercial fusion reactors.

And speaking of risk, AI is starting to nudge its way into the NRC’s risk-informed decision-making (RIDM) processes. This is where regulators weigh probabilities and consequences instead of just following rigid rules. AI’s ability to analyze past regulatory rulings and risk profiles means it could soon be helping regulators determine what’s truly “safe enough.”

AI also showed up in nuclear plant site characterization. Geotechnical engineers—who assess the robustness of the ground beneath a nuclear plant—are looking at AI to improve seismic modeling. Right now, seismic site assessments rely on labor-intensive data collection and expert judgment. AI could speed up modeling, detect anomalies, and refine hazard predictions. In other words, it might soon be deciding whether your site assessment is on shaky ground.

The Endgame: AI vs. AI?

Here’s where this gets even more interesting.

• AI inside nuclear plants drafts a 50.59 screening, arguing a change is safe.
• AI at the NRC reviews the same request and disagrees.
• We’re no longer talking about AI vs. human regulators—we’re talking about AI battling AI over nuclear regulations.

Though it sounds ludicrous, it might not actually be too far-fetched. And if it does, who gets the final say? The engineers? The regulators? Or the AI that understands the rules better than either of them?

What’s Next for AI in Nuclear?

The NRC is using it, the industry is using it, and nobody wants to be the last one to figure out how to make it work. The only thing we don’t know yet? Whether AI will become the ultimate compliance enforcer—or the first regulatory loophole that writes itself.

Either way, this ride is just getting started.