Nuclear power is being pulled back into the centre of energy policy at the very moment the rules that kept it peaceful are being renegotiated. This paper argues that the question of whether nuclear cooperation raises proliferation risk is badly framed, because the risk does not sit in reactors but in a narrow band of the fuel cycle (enrichment, reprocessing and, increasingly, high-assay fuels), and in the verification terms attached to cooperation.

Drawing on the IEA’s energy-security data, the IAEA’s assurance-of-supply mechanisms, and four live cases (Iran, Saudi Arabia, China and India), it concludes that cooperation can strengthen energy security without raising proliferation risk, but only when verification is treated as the price of entry rather than a negotiable extra.

Background

Energy security is having its second great shock in five years. The IEA’s 2026 investment review describes the current Middle East disruption as the second major global energy crisis since Russia’s invasion of Ukraine, and its chief has called it the most significant energy security crisis in decades. Governments have responded the way they did in the 1970s, by looking for fuels that do not cross contested sea lanes. Nuclear power fits that logic. Global capacity stood at about 420 GW at the end of 2025, investment is expected to exceed $80 billion a year, with nearly 80 GW under construction across 15 countries, and the IEA notes that support for expansion now exists in more than 40 countries. Yet every reactor programme carries an older inheritance. The bargain struck in the Atoms for Peace era, later written into Article IV of the Non-Proliferation Treaty, promised states access to peaceful nuclear technology in exchange for forswearing weapons and accepting inspection. The question for 2026 is whether that bargain can stretch to cover a nuclear revival that is larger, more commercial and more geopolitically driven than anything its drafters imagined.

The first step is to separate what is dangerous from what is merely nuclear. A power reactor is not a bomb factory; the material that matters is enriched uranium and separated plutonium.

The distinction is sharpest at the enrichment stage, because, as an NTI analysis of fuel-assurance schemes stresses, the same enrichment facility can, with relatively small modifications, produce either low-enriched uranium for power plants or highly enriched uranium for weapons. Advocates of nuclear exports make exactly this point to dismiss alarm about new reactors, arguing that proliferation risk comes from enriching fuel, not from owning power plants. Iran offers a grim confirmation of the logic. Inspectors have been able to visit facilities such as the Bushehr power plant, the civilian reactor, while access to enrichment sites has been blocked.

The IAEA estimates that Iran held 440.9 kg of uranium enriched to 60 percent, enough for roughly ten weapons if enriched further, and in October Director General Rafael Grossi told the Security Council that the agency cannot confirm that none of Iran’s safeguarded material has been diverted. The lesson is not that nuclear energy is inherently dangerous. It is that the reactor was never the problem, and that a country can have a perfectly legitimate power programme alongside a fuel cycle nobody can see into.

If the risk sits in the fuel cycle, then energy security and non-proliferation collide precisely there, because the fuel cycle is also where energy security is weakest. The IEA reports that more than 99 percent of enrichment capacity sits in four supplier countries, with Russia alone holding 40 percent. Western utilities have spent years trying to reduce that exposure, and the IEA’s own officials have flagged that France is now dependent on Russian uranium. The natural policy response is to build enrichment capacity elsewhere, and that is happening: the United States and China are both seeking to expand enrichment as they scale up reactor fleets.

A 2026 paper for the EU Non-Proliferation and Disarmament Consortium captures the resulting tension well, arguing that fuel-cycle diversification emerges from the interaction of energy security, industrial strategy and non-proliferation governance, and that these three goals are not automatically aligned. Diversification is good for supply security and, mechanically, bad for the goal of keeping sensitive technology in few hands. Anyone who says the two objectives are simply complementary has not looked at where the centrifuges have to go.

The tension deepens with the technology that is driving much of the current enthusiasm. Many small modular and advanced reactors need high-assay low-enriched uranium (HALEU), enriched between five and twenty percent uranium-235, a range that raises efficiency but also proliferation concern. Critics at the Bulletin of the Atomic Scientists note that HALEU is currently subject to far less stringent safeguards than highly enriched uranium and point out that Russia has already deployed barge-mounted SMRs running on it, with further exports planned. John Carlson of the Vienna Center for Disarmament and Non-Proliferation argues that current IAEA approaches may need adapting to HALEU’s higher risk. The optimistic school answers with design. The US National Nuclear Security Administration says it is helping industry build proliferation resistance and safeguards into reactor and fuel-facility designs from the outset, noting that retrofitting something as mundane as an inspection camera port into a finished facility can cost more than $100,000.

Research in Nuclear Technology suggests that graphite-matrix TRISO fuel in high-temperature reactors can lower proliferation risk relative to other fuels and reactor types. Both camps have a point. But safeguards built around large light-water plants may be ill-suited to SMRs, and the regulatory framework for commercial HALEU production, transport and use is still being developed. The IEA’s projection that SMRs could reach 80 GW by 2040, about a tenth of global capacity, means this window for getting the rules right is short.

The institutional answer to the fuel-cycle problem has long been multilateralism: take the sensitive steps out of national hands, and guarantee supply so that no one feels compelled to build their own. The clearest embodiment is the IAEA’s low-enriched uranium bank in Kazakhstan, designed as a supply of last resort. It is owned and controlled by the IAEA but operated by Kazakhstan, and it was funded through contributions from the EU, Kuwait, Norway, the UAE, the United States and others. A state buying from it must meet IAEA safeguards, and in the original design was expected to refrain from operating its own enrichment or reprocessing facilities. Russia’s reserve at Angarsk operates alongside it. The weakness of these schemes is instructive. The bank’s volume is sized to cover two to three reloads for a 1,000 MW reactor, and member states insisted that it would not limit rights under Article IV. In other words, it solves the reliability problem (what if my supplier cuts me off?) but not the sovereignty problem (why should I depend on anyone?). States that want enrichment often want it for status and hedging as much as for fuel, and an insurance policy does not satisfy that motive.

That unsolved motive is exactly what the new US-Saudi civil nuclear agreement tests. Signed on 22 July 2026, the agreement requires a two-year feasibility study on commercial enrichment in the kingdom, and the United States has never before supplied enrichment technology under a cooperation agreement. It also does not require Saudi Arabia to accept the IAEA’s Additional Protocol, the enhanced-access arrangement that, in Grossi’s words, gives the agency more access. That departs from the 2009 US-UAE agreement, which included the Additional Protocol and is widely regarded as the gold standard. The administration’s case is largely commercial, and it has substituted a bilateral safeguards agreement for the Protocol.

Analysts at Brookings take a middle view: since Saudi leaders have vowed to acquire weapons if Iran does, there are inherent risks in supporting enrichment there, and the agreement’s provisions tying Riyadh to US cooperation reduce but do not eliminate them. Brookings also expects it to take effect as early as December. My own reading is that the deeper danger is precedent. If a close partner can obtain a path to enrichment without the strongest verification, other states will ask for the same terms, and a standard that once functioned as a norm becomes a bargaining position.

Iran shows what the absence of verification costs once trust has already gone. In September the Quad of France, Germany, the UK and the US told the IAEA Board that the agency had lost all visibility over Iran’s highly enriched stockpile for over a year, and the IAEA itself warned that the loss of continuity of knowledge needs to be addressed with the utmost urgency. A June interim deal between Washington and Tehran, requiring the stockpile to be diluted in exchange for sanctions relief, depends entirely on inspectors being allowed in, and as of late October Grossi was still urging Iran to grant access. Observers fear material could be moved to undeclared locations. The Iranian case is an extreme one, but it clarifies the general principle: the value of safeguards lies in continuity of knowledge, and once that is lost, even a good deal rests on faith.

Contrast this with the posture of the world’s largest nuclear builder. China operates 60 reactors with 36 more under construction and says it has placed 30 of its nuclear facilities under IAEA safeguards, though as a recognised nuclear-weapon state its safeguards are voluntary and far narrower than those applied to non-nuclear-weapon states.

India supplies a final test case, and an unusually relevant one for a reader in Pune. The SHANTI Act replaced the Atomic Energy Act of 1962 and the 2010 liability law, ending the state monopoly and opening nuclear generation to private and foreign participation in service of a target of 100 GW by 2047, from under 9 GW today. Supporters note that the reform leaves critical safeguards and strategic functions under sovereign oversight, and India’s civil facilities already operate under a safeguards agreement in force since 2009 and an Additional Protocol since 2014. Critics focus on a different risk: that operator liability caps of ₹100 to ₹3,000 crore leave the public purse exposed in a severe accident. The India case matters here because it shows a large expansion proceeding with verification embedded in the legal architecture rather than negotiated away. It also shows that the proliferation debate is only one part of nuclear governance; liability and safety are separate tests of whether cooperation is being designed responsibly.

Taken together, these cases support a more conditional answer than either optimists or alarmists usually give. Scientific and commercial cooperation can strengthen energy security, because reactors, shared engineering, and diversified supply chains reduce exposure to the concentrated suppliers the IEA warns about. And cooperation does not in itself raise proliferation risk.

What raises it is cooperation that transfers sensitive capability while discounting verification. Four design principles follow.

First, enhanced verification, meaning the Additional Protocol or its equivalent, should be the price of entry for any deal touching enrichment or reprocessing, as the UAE precedent showed was achievable.

Second, the world should share reactors and services widely while keeping the most sensitive steps under multilateral or tightly supervised custody, and the fuel bank model should be expanded to cover the HALEU that advanced designs will need.

Third, safeguards-by-design should become an export standard rather than a marketing point, and the IAEA’s approach to HALEU should be revisited before deployment scales.

Fourth, supplier diversification should be coordinated with non-proliferation policy, so that new enrichment plants are located and governed with both goals in view.

There are limits to this argument. SMR deployment is still early, so claims about how those designs will perform under safeguards remain largely prospective. The US-Saudi agreement has not yet entered into force, and Iran’s situation is moving week to week. The sources also differ in their incentives: industry-linked analysts emphasise how hard diversion is, while arms-control institutions emphasise how weak the rules can become, and a careful reader should weigh both. But the direction of the evidence is clear enough.

The atom’s bargain has always been an exchange, and the exchange only holds if both halves are enforced. Nuclear cooperation can make the world’s energy supply more secure and its politics no more dangerous, provided the countries doing the cooperating are willing to keep verification at the centre of the deal even when commercial and strategic pressures say otherwise.