Fusion: Safeguards and Security for a Technology That Is Not Fission
Why the verification system built for uranium and plutonium does not simply transfer to fusion, what a fusion machine could be misused for, why its fuel supply chain reaches back into fission, what it actually produces as waste, and how regulators are choosing to treat it.
Learning objectives
- Explain why fusion is outside IAEA safeguards today and what that does and does not mean.
- Describe the three misuse pathways that make fusion a nonproliferation question at all: breeding, tritium, and dual-use components.
- Trace where fusion fuel comes from and why the supply chain is not independent of fission.
- Describe what a fusion plant produces as radioactive waste and how the United States has chosen to regulate it.
- State the design-in argument and why the window for it is now.
Adapted from a lecture by the author and Jackson Stanley for a fusion safety and regulation course organized by the Clean Air Task Force and the University of Tennessee in July 2026. The regulatory status of fusion changes month to month; the current state is on the situation page. This module keeps the structure of the problem.
1. Two dates
Fusion power has been “fifty years away” for seventy years. That joke hides a fact worth starting with: humanity first released fusion energy on 1 November 1952, at Enewetak Atoll, in a test called Ivy Mike — the first full-scale thermonuclear explosion, on the order of ten megatons. Our first fusion “reactor” was a weapon. Whatever else is said about fusion’s proliferation risk, it cannot be said that the two subjects have nothing to do with each other, and it is the reason safeguards for fusion cannot be an afterthought.
The second date is now. Private fusion companies have raised billions of dollars; several are building machines they describe as pilot plants; the United States has rewritten its regulatory framework for them; and the International Atomic Energy Agency has begun scoping what safeguards for fusion might look like without yet adopting anything. Commercial fusion is arriving faster than its safeguards. This module is about the gap.
2. The system we have was built for fission
The Nuclear Non-Proliferation Treaty and the safeguards system that verifies it were designed around one idea: fissile material. Enriched uranium and plutonium are what the treaty controls, what a comprehensive safeguards agreement lets the IAEA account for, and what every inspection technique — item counting, material balance, seals, cameras, environmental sampling — is built to track (Module 2, section 6). Decades of experience with reactors, enrichment plants and reprocessing plants stand behind it.
A fusion machine, in normal operation, runs on deuterium and tritium — two isotopes of hydrogen — and produces helium and neutrons. It uses no fissile material and makes none. So as the rules are written today, a pure deuterium–tritium fusion plant is not subject to IAEA safeguards at all: there is no “nuclear material,” in the treaty’s sense, to safeguard. That is not an oversight anyone has decided to fix; it is simply what the definitions say.
Two features of the system matter for what follows. Safeguards bind states, not facilities. The obligation flows from a state’s treaty commitment; the state declares its facilities and accepts verification on its territory; a developer does not make its own deal with the Agency. So any workable approach has to let each state show its compliance credibly, and industry has to help shape the rules in every country that will build fusion, not just one. And the baseline agreement sees only what is declared. Detecting undeclared activity — which is fusion’s real concern, as the next section explains — is the job of the Additional Protocol, which is voluntary and not in force everywhere (Module 1, section 4).
3. Why the misuse potential is real anyway
If fusion makes no fissile material, why is this a nonproliferation subject at all? Three reasons.
Neutrons breed. A deuterium–tritium plant is an intense neutron source — that is what makes it work. Neutrons that land on uranium-238 make plutonium-239; neutrons that land on thorium-232 make uranium-233 (Foundations 4). A fusion machine deliberately fitted with a blanket of fertile material would be a production reactor of a new kind, and analyses of realistic tokamak designs have found that the quantities are not trivial. Nothing about normal operation does this; the concern is a purpose-built or covertly modified facility — an undeclared activity, exactly the case the baseline safeguards agreement is weakest at finding.
Tritium is a weapons material. It is used to boost the yield of fission weapons, it is scarce, it decays with a 12-year half-life so that stockpiles have to be continuously renewed, and a commercial fusion industry would handle it by the kilogram. Deuterium is a heavy-water moderator and a boosting material too, and lithium-6 is the feedstock from which tritium is bred. Each moves through a commercial supply chain, and a state could skim material from that chain. None of the three is “nuclear material” under the treaty, and accountancy for them is nowhere near as developed as it is for uranium and plutonium.
Components are dual-use. Much of what goes into a fusion machine — tritium-handling systems, lithium isotope separation, high-power lasers or magnets, neutron diagnostics — has uses in a weapons program. Export controls already cover some of it; the rest is a question that trade at scale will force.
Put together: the proliferation risk from fusion in normal operation is low, lower than fission — but not zero, and lower than fission only if safeguards are designed in. That is the finding of the most-cited academic treatment of the question, Goldston and Glaser’s 2012 paper, and the authors of this module concur. The design-in window is now, while the first commercial machines are still on paper.
4. Where the tritium comes from
Fusion is often described as freeing the energy system from the fission fuel cycle. Its fuel supply says otherwise, at least for now.
Tritium is rare in nature and decays quickly, so a fusion plant has to be supplied with it. Today’s supply comes from fission: mainly from CANDU heavy-water reactors, which make tritium as a by-product in their heavy water, and in the United States from tritium-producing rods irradiated in a commercial power reactor for the weapons program. The world’s civil inventory is a few tens of kilograms. A fleet of fusion plants would exhaust it quickly.
The long-term answer is for each plant to breed its own tritium from lithium-6 in a blanket around the machine. That in turn calls for lithium enriched in the lighter isotope — a separation step with its own dual-use character. Either way, the fusion fuel cycle reaches back into fission reactors and into isotope-enrichment technology. Anyone drawing the boundary of “what needs safeguarding” around a fusion industry has to draw it wider than the fusion machine.
5. What fusion actually produces
Set misuse aside and ask what a fusion plant leaves behind. Two things.
Activation products. The neutrons that carry the energy out of the plasma make the surrounding structure radioactive. Activated steels and other components become intermediate-level waste — bulky, but with no transuranic elements and no fission products, which is the difference from fission that matters most for disposal. With sensible material choices, most activated steel can be cleared or recycled after about a century.
Tritium. Radioactive hydrogen contaminates everything it touches, permeates metals, and is environmentally mobile as tritiated water. It drives much of the waste-handling burden of a fusion plant. Its saving grace is its half-life: after roughly thirty years most tritium-bearing waste has decayed below low-level-waste limits, so interim storage is a real strategy in a way it is not for fission waste.
The result is a waste stream that is manageable rather than a show-stopper: no high-level waste at disposal after cooldown, no repository of the kind the fission back end requires. That is the tractable half of the problem, and it should not be oversold or undersold.
6. How regulators are choosing to treat it
The United States redrew the map in 2024. The ADVANCE Act defined material produced by fusion as byproduct material, and the Nuclear Regulatory Commission is therefore regulating fusion machines under its materials framework — 10 CFR Part 30 — rather than the power-reactor rules written for fission. The proposed rule appeared in February 2026; guidance scales oversight with the size of the tritium and activation inventories; and security requirements are to be decided case by case rather than imposed as a class. It is a lighter, risk-informed path than fission licensing, and it settles one question: fusion will not be licensed like a fission reactor, at least in the United States. Whether and when the rule becomes final is tracked on the situation page.
The United Kingdom and the European Union (through ITER and Euratom) have gone facility by facility, with detailed activation and tritium waste protocols and a distinctive Euratom legal framework. Other fusion nations are taking their own paths. There is convergence on the waste question and none yet on verification.
7. Where things stand
Converging: fusion will not be licensed like fission; activation and tritium waste are manageable.
Unresolved: there is no international safeguards approach designed for fusion; how tritium export controls apply as trade scales; and who writes the verification playbook, and when. The IAEA has begun scoping work and published successive editions of a fusion outlook; nothing has been adopted.
The argument this module makes is that industry has to lead — bring developers, regulators and the IAEA to the same table, shape workable protocols in every fusion nation rather than waiting to react, and turn practical experience into verifiable standards — while the IAEA designs a fusion-tailored approach and states universalize the Additional Protocol. The alternative is to repeat the fission experience of the 1960s: build first, verify later, and spend the next half-century catching up.
Explore. (a) A country with a comprehensive safeguards agreement but no Additional Protocol builds a fusion pilot plant. What can the IAEA verify there, and what can it not? (b) List the three fusion fuel-chain materials that are weapons-relevant and, for each, say where a state would most plausibly acquire it. (c) Why does the byproduct-material framework fit fusion’s waste well and its proliferation questions poorly? (d) “Fusion has no fuel cycle.” Argue against that sentence using section 4.
Sources
- A. Glaser and R. J. Goldston, “Proliferation risks of magnetic fusion energy: clandestine production, covert production and breakout,” Nuclear Fusion 52, 043004 (2012).
- Analyses of fissile breeding in ARC-class fusion plants, Nuclear Fusion (2025); Hua et al., arXiv:2207.14348 (2022).
- International Atomic Energy Agency, World Fusion Outlook (2023, 2024, 2025 editions).
- IAEA safeguards status as of 31 December 2025 (comprehensive safeguards agreements in 183 states; additional protocols in force in 144 states plus Euratom), as compiled for the source lecture from IAEA and Arms Control Association data.
- ADVANCE Act of 2024 (P.L. 118-67), fusion provisions; U.S. Nuclear Regulatory Commission, “Regulatory Framework for Fusion Machines,” proposed rule, Federal Register, 26 February 2026; NUREG-1556, Vol. 22.
- O. Sauter, Max Planck Institute for Comparative Public Law and International Law (2023); Sauter and Murphy (2024); Chew and Ngo, RSIS Commentary (2026) — legal and policy treatments.
- H. L. Hall and J. Stanley, “International Regulatory Considerations for Fusion Safeguards and Byproduct Material Management,” lecture, Fusion Safety and Regulation Course (Clean Air Task Force / University of Tennessee), Knoxville, 29–31 July 2026, with its annotated Fusion Safeguards — Reference Reading List (Institute for Nuclear Security, 2026).
This material is based in part upon work supported by the Department of Energy National Nuclear Security Administration through Defense Nuclear Nonproliferation’s Enabling Capabilities in Technology Consortium under Award Number DE-NA0004197. The views and opinions expressed are the author’s own and do not necessarily state or reflect those of the United States Government or any agency thereof.