The fusion rule is out. Luckily I have AI to read it.

One of the best projects of my NRC career was working on the fusion rulemaking. We were charting new territory for a class of facilities that had never been regulated at commercial scale. Should fusion be handled like a reactor or a particle accelerator? Did we need new waste rules? And, frankly, how do you write a licensing framework for technology that doesn't fully exist yet? It was intellectually stimulating and urgent—my favorite kind of regulation.

I left NRC almost a year ago, while all this was still cooking. So when the proposed rule dropped today, I was seeing the final product for the first time. But now I have AI! I asked Gordian, to read the proposed rule, draft guidance, environmental assessment, and regulatory analysis and tell me what it found. What came back genuinely surprised me. More on that in a moment.

What the rule actually does — and why it matters

The short version: the NRC is proposing to regulate fusion machines as byproduct material facilities under 10 CFR Part 30, the same framework used for cancer treatment and industrial irradiators. We all knew this, but it’s still good to see it in writing. Regulating fusion like a fission reactor doesn’t make sense, and doing it this way is in line with both the hazards of fusion and the regulatory expertise of the National Materials Program.

The path to get here was not straightforward. In January 2023, NRC staff submitted a paper to the Commission recommending a hybrid approach: start with Part 30, but build in criteria to escalate future designs to the reactor framework if warranted. The Commission rejected that recommendation unanimously and directed staff to implement the pure Part 30 approach.

Congress then codified that direction in the ADVANCE Act of 2024, which added a statutory definition of "fusion machine" and explicitly brought fusion-generated radioactive material within the byproduct material definition.

The rule itself has a lot of sub-parts. The core additions are new license application requirements, a new license issuance standard, and a new waste disposal pathway for fusion machine waste. It's a framework designed for "near-term" fusion machines—the ones that can't sustain a reaction without active engineered systems, can't go critical, and don't require active cooling after shutdown. The NRC is explicit that if a future design presents hazards beyond those characteristics, they'll revisit. This was a key point for us when drafting the rule.

Four things that might catch attention

First: the rule is officially characterized as a "deregulatory action" under the current executive orders on regulatory reform. I understand the legal logic—the NRC argues it's clarifying existing requirements rather than imposing new ones—but calling a new licensing framework for an entirely new class of nuclear technology "deregulatory" is a framing choice that could draw comment. It's not wrong, exactly. It's just... a choice.

Second: the regulatory analysis shows that the guidance-only alternative (no rule) would have produced a higher monetized net benefit ($2.03 million) than the rulemaking alternative the NRC actually recommends ($1.38 million). The NRC recommends the rulemaking anyway, based on non-monetized qualitative benefits. I’m a firm believer that rulemaking is the right answer to provide clarity and consistency across the NRC and 40 Agreement states, but this is a case where many-significant-figure cost benefit analyses in rules just aren’t very useful.

Third: the waste classification tables under 10 CFR 61.55 are not being updated in this rule, even though the Advisory Committee on Reactor Safety recommended in 2022 that they should be. The NRC's reason is straightforward: they don't yet know which radionuclides will drive the risk significance of fusion machine waste, because the machines don't exist yet at commercial scale. The environmental assessment acknowledges directly that fusion machines could create "significant inventories of activation products that are not included in these waste classification tables." So we're establishing a licensing framework while acknowledging the waste classification system may not be adequate for the waste that framework will produce. That could cause some tension.

Fourth: the rule is explicitly separate from—and may be revised by—the broader NRC reform effort underway under E.O. 14300. The NRC says so right in the preamble. I guess we’ll all keep our eyes on those other rules (see my prior blog for details). Luckily, this rule has a 90-day comment period (due May 27) unlike the 30 we’re expecting for other rules.

What Gordian found that I couldn't have found quickly on my own

Here’s my REAL surprise: I've spent 21 years in this business, over three explicitly leading rulemaking, and I've never seen an analytical tool do what Gordian did with this package.

The four documents run to hundreds of pages. Gordian researched the history of the rule, read the whole package, cross-referenced the four documents against each other, and surfaced things that would have taken a team of reviewers days to find. It did it while I was running to change planes in Denver, finished its report fast after I gave it directions in the air, and fact-checked this blog post while I ate breakfast.

Here are some nits it picked that my old red pen would be proud of:

• The NUREG's foreword contains "[TBD]" placeholders — boilerplate from other NUREG volumes that was never filled in — and states that the regulations "were amended" when they haven't been yet (the rule is still in comment, obviously).
• The environmental assessment cites its own ADAMS accession number (ML25168A335) when referring to the regulatory analysis — which is actually ML25168A339.
• The byproduct material definition is numbered differently in §§ 20.1003 and 30.4, creating a cross-reference inconsistency in the compatibility table.

None of these are fatal flaws (and Gordian spent more time telling me about substantive comment areas). The NRC staff who worked on this rule did serious, careful work over years, and I have enormous respect for what they produced. But version control errors happen, especially in large multi-document packages under deadline pressure. I know. I've been there.

Now Everstar has the kind of tools I never had at the NRC. And that means we can help the entities who need to comment on this rule — fusion developers, Agreement States, waste disposal operators, public interest groups — can do it with the kind of analytical depth that actually moves the needle.

Get ready to comment

Keep May 27, 2026, on your calendar as the due date. If you're in the fusion space, or you work with Agreement States, or you care about how the U.S. regulates the next generation of nuclear technology, this is your moment. The NRC is explicitly asking for input on tritium reporting thresholds, waste disposal approaches, and more. Don’t miss your chance.

Everstar is happy to help:

• A plain-language cheat sheet summarizing the key issues and comment opportunities is available FREE if you comment “FUSION” on my LinkedIn post about this blog (find it at https://www.linkedin.com/in/theresavclark/).
• Gordian subscribers can also have the full history report and package analysis that the tool pumped out tonight—just email me at theresa@everstar.ai and I’ll save you the step of generating it.
• If you aren’t a subscriber and this makes you realize you need to be ... well, you know where to find me.