Nuclearization and de-democratization: security,secrecy,and the French pursuit of nuclear weapons (1945–1974)

Abstract

Do states change when they acquire nuclear weapons? This article looks at the consequences of nuclear acquisition on democratic states. It argues that nuclear acquisition is best understood as a process of political change through which state actors adapt existing institutions to the new, and unprecedented, challenges created by nuclear weapons. One form of this process of “nuclearization” is the development of nuclear secrecy regimes, which results from actors’ desire to maintain control over information they perceive as potentially having major security implications. Actors may not know what, exactly, must be concealed. The ultimate stakes of nuclear policy, however, are so high that they have incentives not to take their chances. Secrecy has implications for democratic governance: it can exclude actors from decision-making, distort information made available to the public, and be abused by actors in search of autonomy from democratic control. As a result, nuclear secrecy affects the overall level of public consultation inside a state, causing a democratic recoil. To borrow Charles Tilly’s concept, during this process of nuclearization, states also de-democratize. To make this case, the article examines the French nuclear secrecy regime from 1945 to 1974. Drawing on primary sources, it traces the origins of nuclear secrecy in France back to security concerns and shows how this development ultimately reduced the level of public consultation in France and caused a form of de-democratization.

The invention of nuclear weapons has had profound and long-lasting effects on international politics. Yet the impact of nuclear weapons acquisition on states’ domestic politics remains a blind spot (Gartzke and Kroenig, 2016: 409). Prominent scholars (Sagan, 1996) have examined how domestic politics influences the acquisition of nuclear weapons or, more generally, the role of domestic politics in the making of nuclear weapons politics. ¹ However, no systematic study has yet examined the impact of the latter on the former. In this article, I make the case for looking not only at what states do with nuclear weapons, but also at what nuclear weapons do to states. Does a state change when it becomes nuclear?

This is an important question. At a general level, we need a full account of the effects of nuclear weapons. Though we know a lot about their impact on International Relations, their impact at the state level has been studied remarkably little. We rely on the assumption that nuclear acquisition is essentially a change in a state’s foreign policy standing, or material capabilities, and not a change in its political configuration. Contrary to that, I argue that nuclear acquisition is also a process of domestic political change Second, we need a more rigorous and empirical framework to study the impact of nuclear acquisition on democratic states specifically. It is generally accepted that nuclear weapon sit uneasily with liberal democracy, despite five out of the nine nuclear-armed states being qualified as democracies (Born et al., 2010; Dahl, 1985; Scarry, 2014; van Buuren, in press). It is not, however, entirely established why it is so. Some have asserted that nuclear weapons created “a variety of structural necessities that contradict the spirit and substance of democratic governance” such as “secrecy, lack of accountability, permanent emergency, concentration of authority, peacetime militarism, extensive apparatus of state intelligence and police” (Lifton and Falk, 1982: 262). But what exactly is the causal relationship between nuclear weapons and these outcomes? There is a need for more empirically informed studies on this specific relationship.

Thus, I ask the following question: How does the acquisition of nuclear weapons affect the institutional configuration of liberal democratic states? I answer this question in two steps. First, I argue that nuclear acquisition creates constraints for state actors in the form of new security challenges. Actors have to develop new regulations, practices and institutions to face these changes, thus engaging in institutional change. I focus specifically on the emergence of information control regime related to nuclear politics—nuclear secrecy. I argue that this emergence is an inherent product of nuclear programs. A state whose security relies on nuclear weapons can never acquire the certainty that it is secure. Its security relies on a small number of technological variables who could potentially be thwarted by an adversary (Buzan, 1989: 216–217). Failure would be intolerable. Actors have high incentives to develop information control regime designed to conceal knowledge they interpret as having security implications while retaining the possibility of strategically communicating other. The logic of nuclear secrecy is not that there is specific knowledge that needs to be concealed. It is a solution to the problem of uncertainty over transparency costs in a high-cost/low-reward situation.

Second, I argue that this process of “nuclearization” deteriorates the ability of a state to act democratically. To make this case, I rely on the concept of “de-democratization.” Charles Tilly argues that political regimes can be classified along different axes, notably the breadth and equality of political participation, the level of public consultation and protection against arbitrary state actions, all determining the overall level of democracy (Tilly, 2003). De-democratization refers to a process of change which causes a state to recoil along one or more of those axes. The emergence of nuclear secrecy regimes causes a recoil in the level of public consultation, “the degree to which polity members exercise binding collective control over governmental agents, resources, and activities” (Tilly, 2000: 5). Secrecy excludes unauthorized actors from knowing about the state’s nuclear activities, and distort the information made available about those. In practice, this means that the public, its representatives, and even sometimes elected officials, cannot effectively control nuclear policy since they rely on imperfect information. More than this, they are vulnerable to abuse by actors seeking to acquire more autonomy. As Colaresi (2014: 86) put it, the “potential for abuse does not exist separately from the uses of secrecy that justified having the capacity to keep information from public view in the first place.” Nuclear secrecy has the effect of flawing deliberation process, limiting oversight, and preventing accountability—in other words, nuclearization deteriorates the quality of public consultation.

Of course, security has classically implied democratic trade-offs, and secrecy is a common expression of these. It is generally agreed that it is democratically tolerable that certain topics could be dealt with secretly (Thompson, 1999). However, nuclear secrecy is different in two senses. First, nuclear secrecy differs from other forms of state secrecy. It introduced a supplementary level of secrecy, “the highest one,” in already existing state secrecy regimes (Laurent, 2024: 120). Second, the content of nuclear secrecy has unique implications. It concerns itself with nuclear policy, a domain which cuts to the very heart of the state’s fundamental purpose, to wit, to control and exert violence on behalf of its constituents. Limiting control over this domain has heavier consequences for public participation than limiting control over most domains of state actions. For this reason, the de-democratizing consequences of nuclearization are far from trivial.

Empirically, I use the case of the French nuclear secrecy regime’s development from 1945 to 1974 to illustrate my argument. I rely on primary sources, many of them untapped, from public and private archives. Using private correspondence, minutes, policy memos and other archival documents, I trace the process through which nuclear secrecy developed in France and affected democratic governance. With this, I intend to make three contributions. The first is a contribution to nuclear security studies, which has not tackled the question of nuclear weapons’ effect at the domestic level and relied on the assumption that such effects were either absent or irrelevant. I show that they are neither. Most specifically, the article contributes to the literature interested in the relation between liberal democracies and nuclear weapons by providing an empirically informed answer to the question of nuclear weapons’ impact on democratic governance. The second is more theoretical. I propose a framework that allows analysts to trace the effects of nuclear acquisition on states, and therefore to better understand how and why states nuclearize. Tracing specific political outcomes to the causal implications of nuclear armament, I show how the specific properties of technology combine with state actors’ quest for security and lead to institutional change. Finally, I also make an empirical contribution. Based on primary sources and untapped archives, the original case study contributes to the historiography of French nuclear policy, which has so far received limited attention in anglophone IR and international history (Hymans, 2021: 2), ² and to the historiography of nuclear secrecy which has so far not looked at the French case (Fraise, 2022b: 179).

The article proceeds in three sections. In the first, I provide a critical assessment of the existing literature and argue that the question of nuclear weapons’ effects at the domestic level has so far escaped analysis. In the second section, I develop the framework outlined above, a framework designed to allow for the empirical investigation of the effects of nuclear weapons acquisition on domestic politics. Finally, to probe the plausibility of this framework, I use the example of nuclear secrecy regimes, illustrated by the French example.

How do nuclear weapons affect domestic politics? Examining the existing literature

The problem of nuclear weapons’ effects on domestic politics has not been properly studied so far. In nuclear security studies, the question has been the object of little attention (Gartzke and Kroenig, 2016: 409). Domestic politics is invoked essentially as a variable to explain different outcomes in nuclear policies (Braut-Hegghammer, 2016; Hymans, 2012; Koch, 2023; Saunders, 2019). In International Relations more generally, the literature investigating the relationship between technology, the state, and the international system (for a recent overview, McCarthy, 2018) has not investigated this relation either. For example, Geoffrey Herrera (2006: 130) argues that the development of the atomic bomb led to the birth of “a set of institutions and practices” inside the US state, but focused only on their “profound impact on Cold War International Relations” and not on the state.

In (macro)historical approaches to state development, in spite of an interest for how technologies of violence shaped domestic structures (Finer, 1975; Tilly, 1975), nuclear weapons are rarely mentioned in classic studies of the genre (McNeill, 1984: 382; Mann, 2012: 33–36, 128, 308–309; Tilly, 1990: 202, 197). A recent exception is Warren Chin (2023: 67), who argues that nuclear weapons, if anything, reinforced “the connection between war society and the government.” What effects they had beyond war-making, however, is not investigated. In International Relations theory, authors associated with the tradition of “nuclear oneworldism” have argued that the invention of nuclear weapons fundamentally affected the nature of states. “Cut[ing] to the heart of the state as a war-making entity capable of securing itself,” nuclear weapons a certain mode of protection (the “real-state”) unviable (Deudney, 1995: 209, 2007: 246). John Herz (1959: 22) argued early on that nuclear weapons caused “the most radical change in the nature of power and the characteristics of power units since the beginning of the modern state system.” However, this argument does not shed much light on the domestic effects of nuclear acquisition as it is not primarily geared toward explaining to explain changes inside states. Finally, in critical security studies, few works have confronted upfront the relation between nuclear weapons and domestic-level institutions in the way it has been done for other emerging technologies, such as drones (Leander, 2013; Walters, 2014). Rather, not unlike nuclear security studies, critical scholars locate nuclear weapons’ politics primarily at the international level (Biswas, 2014; Peoples, 2019; van Munster and Sylvest, 2016, 2023).

This likely stems from the tendency of International Relations scholars to view the state “from ‘the outside’” (Adler-Nissen, 2013: 179), which limits their ability to recognize potential feedback effects occurring at the domestic level. However, it is notable that the relation between nuclear energy and the state has been, to an extent, studied by political scientists and sociologists interested in public policy (Arnhold, 2024; Jasper, 1990; Jungk, 1979; Topçu, 2013), but similar frameworks have not been applied to nuclear weapons.

Exceptions can be noted. Studies on military adaptation have looked at how nuclear weapons acquisition could transform a state’s military (Bacevich, 1986; Boureille, 2008; Moody, 2020), showing how nuclear weapons produce intra-institutional change. The question of nuclear weapons effects on polities has been the object of attention in anthropology (Masco, 2006). However, this work focused mainly on the United States and did not seek generalizable conclusions. Similarly, works on civil defense have outlined how the threat of nuclear war affected state development (Geist, 2019; Grant, 2010; Grossman, 2001), but civil defense is not specific to nuclear-armed states, and not specific to the nuclear threat either. Similarly, a number of studies have looked specifically at the effects of nuclear weapon acquisition on liberal democracy. Gary Wills (2010) has argued that nuclear weapons paved the way for the rise of the US “imperial presidency,” but does not provide a generalizable theory. Similarly, the edited volume by Born et al. (2010) on the domestic governance of nuclear weapons empirically establish a “democratic deficit” in that domain, without providing a theoretical explanation, however. At the other end, theoretical arguments about nuclear weapons anti-democratic effects have not sought an empirical validation of their claims about nuclear weapons causal effects (Dahl, 1985; Falk, 2019; Scarry, 2014; van Buuren, in press).

Investigating nuclearization: historical security materialism, the state, and nuclear weapons

Nuclear acquisition has, so far, not been studied as a process of domestic institutional change, for neither democratic nor non-democratic states. In this section, I propose a framework to study those effects, before specifying its application to democratic states. Based on a historical security materialist (HSM) argument, I argue that the intrinsic properties of nuclear weapons—an enormous destructivity against which no modes of defense are viable—directly interact with the state’s security goal and create incentives for actors to adapt their institutional arrangements. I make my case using nuclear secrecy regimes as an example of a domestic political nuclearization outcome. I do not argue that the secrecy regime is the only (bitter) fruit nuclearization can bear, but simply that it constitutes a relevant example to illustrate the process under study, particularly when looking at nuclear-armed liberal democracies.

Why nuclear secrecy? Insecurity as a driver of institutional change

HSM posits that the material context in which security-seeking actors evolve plays an important role in determining the forms and shape of the institutions they develop. As Deudney puts it, “[technologies of violence] pose obstacles and opportunities for human institutions as they pursue the long-standing and basic goal of providing physical security. As these material realities change, human institutions are forced to adapt” (Deudney, 1993: 23). HSM, it must be noted, does not argue that states will always find the right solution for security, nor that they would do it immediately. ³ It is, too, agnostic about how change will happen and its outcomes. Though it is a system-level approach, its basic assumption can also be informative for the study of the inner workings of units (see Deudney, 1995: 104–117).

Following this, I argue that the acquisition of nuclear weapons creates a set of constraints in the form of new security challenges for state actors. This leads to institutional change. I call this process nuclearization. The driver of change is the uncertainty intrinsically associated with nuclear weapons. Nuclear policy is a highly dangerous business in a world without restraints from nuclear violence. No state can acquire a certainty regarding its security—nuclear strikes are a constant possibility, and their outcome could be swift annihilation. State actors have incentives to do all they can to prevent that, and to develop institutions designed to assert control over hardware and knowledge related to nuclear policy. Control is not a guarantee against apocalypse: it will not prevent an adversary from striking an actor’s territory. At the same time, control can minimize risks. Physical control over the nuclear arsenal, for example, reduces the risks of an unwanted nuclear explosion, or of an adversary taking control of weapons. This leads to the development of structures of command and control designed to prevent unauthorized nuclear use. Nuclearization is not a deterministic but a relatively open-ended process. Institutional change is driven and constrained by the possibilities contained in nuclear weapons themselves, but its final shape will also depend on the actors’ agency, as well as the other constraints they face. If command and control structures share a common set of constant characteristics, they also show variations dependent on domestic and geo-strategic factors (Shaheen, 2019).

Nuclear secrecy regimes, I argue, are products of nuclearization. Secrecy regimes are defined as a set of institutions, rules, procedures, and practices implemented to manage and control information “to make sure that certain people will or will not have access to certain information at certain times” (Wilsnack, 1980: 468). They are an institutional setting aimed at making information manageable and controllable so that only a defined public can have access to it. They define a state’s secrecy capacity. The use of secrecy as a solution for security derives from the attribution of a security value to nuclear knowledge—knowledge related to nuclear policy writ large (development, production, deployment, strategy). The primary driver of secrecy is the need to maximize the likelihood that any vulnerability or, alternatively, possessed advantage could not be exploitable by an adversary. Failure of information control could allow an adversary to, for example, discover a vulnerability in a state’s nuclear command and control system, and exploit to disable parts of its nuclear forces (Gartzke and Lindsay, 2024: 167–176). From a security standpoint, actors have high incentives to control as much information as possible about their nuclear policy not only to prevent an enemy from acquiring what they see as strategically relevant knowledge but also to strategically reveal certain pieces of information when necessary (Brodie, 1953; Montgomery, 2020). Because the boundaries of nuclear secrecy regimes cannot be objectively fixed, actors have also incentives to extend the boundaries of nuclear secrecy to other domains of policy. For example, the fact that nuclear policy cannot be detached from other domains of state actions, such as military operations or energy production implies that the social space of nuclear secrecy regimes cannot remain limited purely to nuclear arsenals.

The security implications of nuclear knowledge reside therefore in the uncertainty associated to them, and in the known consequences of nuclear war. It does not mean that all nuclear knowledge has implication for security, but at the same time, rewards for transparencies may appear relatively low compared to the potentially incurred costs. Logically, secrecy offers no certainty either: it addresses the problem of insecurity but cannot ultimately solve it since it stems from the existence of nuclear weapons themselves.

Nuclear secrecy and de-democratization

The uncertainty associated with the security implications of nuclear knowledge pushes actors to develop secrecy regimes. What are the consequences of this institutional change for democratic states? To capture these, I rely on the concept of de-democratization.

De-democratization refers to the process through which certain parameters of a state’s political regime vary in such a way that it decreases the quality of public contestation and participation in politics (Bogaards, 2018: 1482). Democracy is not “a secure cave into which people can retreat forever” (Tilly, 2003: 37), but a regime in a dynamic state, which can always see variations in the parameters of its government. Tilly broadly defined these parameters as follows: governmental capacity, breadth and equality of political participation, protection from arbitrary action, and finally public consultation (Tilly, 2003: 38). A state is democratic when each of these parameters are maintained at a high level (Tilly, 2000: 4). Nuclearization incurs variations in the parameters defining democratic regimes. States will not cease to be democracies when they nuclearize and may, in fact, still democratize along certain axes—the United States kept democratizing with regards to minorities’ right to participate in policy long after acquiring nuclear weapons (González and King, 2004). At the same time, nuclearization affects what Tilly calls public consultation, which refers to the level of control exerted by the public over state actions. The development of nuclear secrecy regime creates an institutional barrier to the control of nuclear policy by the public, the legislative and the judiciary power.

It does so through two main mechanisms. The first is a mechanism of exclusion. As Costas and Gray put it, “secrecy serves to create important boundaries,” mapping out relations between actors by erecting “walls” and “defin[ing] the corridors between rooms through which secrets may legitimately pass” (Costas and Grey, 2016: 30, 70). Secrecy can prevent representatives or elected officials from being involved in public consultation. The second is a mechanism of distortion, a situation in which the information made available to an actor does not allow them to acquire exact knowledge of a policy. Distortion can be the result of a deliberate intervention, such as documents tampering, or simply out of what Diane Vaughan called “structural secrecy,” the ways in which information control undermines the attempt to know and interpret situations in all organizations (Vaughan, 1996: 286). Though distortion does not prevent public consultation, it flaws democratic process. Indeed, “only when citizens know the content of state policies they can hold state officials to account for them (Mokrosinska, 2020: 21).” Mechanisms of democratic control cannot function effectively if participants lack information (Lester, 2015: 15; Mulgan, 2003: 121; Sagar, 2007: 427).

Nuclear policy is not carried out entirely out of democratic control, but this control is ultimately limited by the secrecy which surrounds it. Secrecy constitutes nuclear policy as a “restricted domain” of state action. These effects are exacerbated by the possibility of abuse inherent in secrecy. Such abuse is not a necessity. However, as Colaresi (2014: 86) put it in his study on foreign policy secrecy in democratic states, the “potential for abuse does not exist separately from the uses of secrecy that justified having the capacity to keep information from public view in the first place.” Some actors can see an interest in using the newly created secrecy capacity to acquire autonomy from other actors. Security requirements can be invoked by officials to control information given to the public in order to deceive it (Colaresi, 2014: 75–99). There is a demonstrated tendency for over-classification on the parts of bureaucracies, driven by self-interest, notably in concealing blunders, but also by organizational logics (Connelly, 2023; Pasquier and Villeneuve, 2007).

The problem identified here is not that secrecy is so fundamentally incompatible with democracy that any form of secrecy is anti-democratic. Dennis Thompson, in his classic discussion on democratic secrecy, noted that democracy was in a constant dilemma regarding secrecy: if democracy requires publicity, some democratic policies also need secrecy. Thompson’s conclusion was that, for democracy to function, secrecy over policy had to be democratically determined. But this solution, he admitted, was not a silver bullet: the most it could do was to “diminish the damage [caused] to democratic accountability” (Thompson, 1999: 193). In line with this, although nuclear secrecy is not wholly undemocratic, it does affect the level of public consultation over a highly consequential domain of state action. The degree to which state de-democratizes remains an empirical question: not all states will exploit this secrecy capacity to the same extent, and external factors play a role in how actors define the boundaries of secrecy regime. Incentives for transparency over publicity can arise, such as commercial interests or domestic politics (Krige, 2016; Salisbury, 2020; Wellerstein, 2021: 335–383). My claim limits itself to why and through what mechanism nuclear weapons cause de-democratization. In the next section, I turn to an empirical case study to illustrate my argument. Focusing on the case of France, I show how the French state went through a process of nuclearization which led to a de-democratization between 1945 and 1974. The French case is relevant to this study for three reasons. First, France is a nuclear-armed democratic state and therefore is as relevant as any to trace the causal relationship between nuclear weapons, secrecy, and democracy. But it is also a largely understudied case. As noted in a recent effort at opening up French nuclear history, France is the only nuclear-armed democracy which does not have a non-official history available in English, leading to the persistence of unchallenged historical claims about the French nuclear program and its effects (Pelopidas, 2024: 1–2). Moreover, there is a tendency in French nuclear historiography to confuse actors’ stated intentions with actual policy outcomes (Pelopidas and Philippe, 2023: 453), which makes invisible all unintended effects of nuclear policy. This makes it all the more relevant as a case study. Second, the trajectory of France nuclear secrecy is interesting because it varies from other studied cases, such as the United States. Contrary to the United States or Britain, where nuclear weapons were essentially “born secret” (Wellerstein, 2021: 1), in France, scientists and administrators originally made little effort to shroud their research in secrecy. This allows us to study the process through which actors came to see secrecy as a necessary component of nuclear policy. Finally, because the French case spreads over a long timeline, it is possible to distinguish different phases in the process of nuclearization and de-democratization.

Nuclearization and de-democratization: the French case (1945–1974)

In this section, I study the development of nuclear secrecy in France and its consequences on democratic control to show how nuclearization produces de-democratization. I first trace the process through which officials came to develop a nuclear secrecy regime in France and identify security concerns over the potential implications of nuclear research as the primary driver of this development. I show how this process was not immediate. Some, notably scientists, defended that their research was unrelated to nuclear weapons and therefore not related to state security. The nuance was lost on officials who did not want to take any chance, particularly when research acquired a military component. I then show how this secrecy capacity was exploited by proponents of the military option to shield this political choice from any form of domestic control by concealing it entirely until it was well advanced. The boundaries of secrecy evolved over time. Nonetheless, many mechanisms of democratic control remained hampered by nuclear secrecy: elected officials were kept in the dark about key aspects of policy, legislative deliberation and oversight over budget was lacking, and secrecy allowed certain actors to escape accountability for their action. Nuclear policy emerged as a domain of state action without proper democratic control, diminishing the overall level of public consultation.

The origins of nuclear secrecy in France (1945–1954)

How and why did nuclear secrecy emerge in France? The process was conflictual. In October 1945, the French Commission for Atomic Energy (CEA) was created. At its head stood Raoul Dautry, general administrator, and Frédéric Joliot, in charge of the scientific aspects of the enterprise. The CEA did not originally pursue nuclear weapons (Belot, 2015: 141). It did not perceive its role as linked to national security. Information control was minimal, and decided by the CEA leadership itself, with little government control. Joliot perceived secrecy mainly as a way to protect the CEA’s (public) research from private firms. Joliot, and many of the CEA’s scientists, were communists, believed in a statist economy and were wary of private industry. ⁴ In this sense, secrecy was essentially a form of “industrial secrecy.” Joliot himself was opposed to scientific secrecy and had criticized the US choice but wished to limit the diffusion of technical information for fear of leaks to commercial industries. ⁵ Some regulations existed around research sites. ⁶ The direction for Homeland Security (DST—Direction de la Surveillance du Territoire) performed control over parts of the personnel but was mainly concerned with the general criminal and wartime records of applicants. ⁷ Its investigations are internally described as “very shallow.” ⁸ Short of a military program, there existed, for the CEA, no particular desire for secrecy beyond basic caution.

This position was rapidly criticized by outsiders, who saw nuclear research as inherently related to state security and thus worthy of information control. Internationally, US diplomats put pressure on Joliot, warning allies against collaborating with him and pretending to have evidence of his collusion with the Soviets (Forland, 1987: 15; Fraise, 2022a). Domestically, members of the parliament or the security services feared that Joliot, because of his communist engagement, could pass nuclear knowledge onto the USSR and thus threaten French security. In March 1948, a senator publicly criticized the CEA’s leaders “whose thought presents a troubling synchronism with Stalin’s Russia.” He argued that France should engage in a “purge” and create “a college of indisputably independent men” whose role would be to “determine the responsibilities for the preservation of secrecy.” ⁹ The amendment was rejected, but the idea that the absence of information control around nuclear research had security implications was taking roots. In July, a secret Ministry of Defense report provided a similar criticism. Calling for a “true epuration (. . .) starting with the head,” the report accused Joliot of having organized, “following an anticipated and well-established plan,” the “communist intrusion in our scientific research.” ¹⁰

By 1949, only four years after the creation of the CEA, concerns over the institution’s weak information control were high. Several CEA employees—some members of the communist party—were accused of mishandling secret documents. ¹¹ None of them were in trouble, as most of the documents did not relate to the CEA’s research. ¹² The event nevertheless led the Ministry of Justice to investigate the possibility of a law to regulate nuclear secrecy and to “forbid civil servants, and in particular those belonging to the CEA, to bring home secret documents.” ¹³ The project eventually failed as the CEA eventually retracted its support, evidence of the institution autonomy over the matter. ¹⁴ Such autonomy would not last.

In April 1950, Joliot was sacked by the government following his participation in the Stockholm appeal against nuclear weapons (Goedde, 2019: 13). Criticism of the CEA continued anyway, and MPs threatened to engage the Government’s responsibility “if, tomorrow, as a consequence of the previous commissioner for atomic energy, an incident would happen, some documents’ “stroll” toward countries with interest” in France’s atomic research. ¹⁵ In 1951, Henri Queuille, head of the Government, decided to re-establish control over the institution and outlined himself a new policy for the CEA’s personnel, made in consideration of the “importance of the tasks which are incumbent upon the Commission.” The personnel now would have “to show loyalty” to the state, and those who did not offer, from a national perspective, the guarantees that the Government is entitled to expect” should be “eliminated”—meaning, sacked. ¹⁶ For Queuille, nuclear research was an activity related to national security and therefore should be subject to control.

For CEA insiders, security considerations appeared disconnected from the actual research, though they admitted that future research would indeed yield security implications. For Perrin, “the protection of secrecy over the work and research of the Commission did not require particular measures.” ¹⁷ The CEA leadership insisted upon the fact that

the Commission (. . .) has been a teaching establishment, designed to train scientific workers in the field of atomic energy, and that it is unlikely (. . .) that we would have until 1955 to keep any atomic secrets (but we would have industrial secrets). (Baudouï, 1992: 366)

“Nothing,” Dautry insisted, “can plausibly be discovered in France, in the domain of scientific research (. . .) which could reasonably worry anyone.” There “exist[ed] no economic issue more important for the future of France” than the peaceful use of atomic energy, and military applications were, so far, out of the picture. ¹⁸ This exchange reveals how scientists and officials shared common points, in spite of their different policy views. For the CEA, nuclear knowledge would indeed yield security implications at some point in the future, but there was no reason to impose information control over nuclear research now. For the political leadership, the implications of nuclear research were simply too great to take such risk. Who knew what the Soviet competitor could do with the product of French research?

This opposition between two concurring views on secrecy was resolved when, in August 1951, Dautry died unexpectedly. His successors did not share his qualms about secrecy. In late 1951, more than 80 communist employees were fired because of their political allegiance, in line with Queuille’s recommendations. ¹⁹ A reform of personnel control started in late 1952, and a “Secrecy Committee” was created in 1953. ²⁰ In 1954, another change occurred, which made secrecy more urgent: a shift toward military applications for nuclear research.

Concerns over the need for secrecy associated with this shift are visible in the bureaucratic battle waged over whether the CEA or the military should be in charge of this program. For the military, General Bergeron insisted that such a program should be carried by the military and that it “should be carried on very secretly, without it appearing on the military budget and only the few CEA scientists which could be necessary.” ²¹ This, he argued, because secrecy was “a defense imperative and even an inter-allied contractual imperative on the international level” and the CEA was considered unsafe. ²² Bergeron’s first concern was with the security implications of nuclear knowledge and his worries were fueled by an internal note which assessed that “in the current state of things, it is not certain that the CEA is fit to preserve the desirable confidential nature of its work.” ²³ Consequently, his opinion was clear: “the atomic team would be in better conditions in the Armed Forces than in the CEA where there exists a majority of communists which is hard to eliminate.” ²⁴ The new head of the CEA, Pierre Guillaumat, disagreed about this part. He answered Bergeron’s criticism by defending the CEA’s ability to control information, arguing that in “the 8 years it has existed, he had not known of any leak attributed to the CEA,” contrary to the Military, whose issues with secrecy were frequent (cited in Pô, 2000: 49). However, he did not disagree in the slightest about the “imperative” of secrecy. As a former head of the CEA’s security put it in internal notes, in 1954, “with the military orientation,” the CEA administrators felt “required to do something” and strengthened information control. ²⁵

What the study of this process shows is that, between 1945 and 1954, the idea of information control over nuclear research evolved concurrently with the security concerns of officials. In early stages, when research was still stammering, the CEA’s activities were not seen as a matter of national security. After a while, however, anxiety about the potential implications of these research arose. No officials could pinpoint exactly what had to be kept secret. CEA officials argued that secrecy would be justified only when knowledge with clear security implications would be produced. Government officials, on their hand, saw the problem differently: nuclear research, in and of itself, engaged state security because it could benefit an adversary in a catastrophic manner. They were surely aware that the USSR was well in advance in that domain. But uncertainty remained and created incentives for information control. Secrecy slowly came to be considered an “imperative,” and debates ended when a military program was in project. The implications of nuclear technology imposed themselves over the French state.

Democratic consequences of nuclear secrecy (1): the “clandestine” phase (1954–1958)

On December 26, 1954, Prime Minister Pierre Mendès-France officialized the launch a military nuclear program with the CEA in charge. The first step was taken in January 1955, with the creation of the Bureau d’Etudes Generales (Bureau for General Studies, BEG) inside the CEA (Pô, 2000: 87), presented as concerned with “all economic and financial studies linked to the uses of atomic energy” (Mongin, 1991: 378). Its actual goal was the production of nuclear weapons. It would operate secretly.

Research took place in fully secret sites. The BEG established its office in a castle belonging to a private company. Contracts passed for the BEG would use the company’s name (Mongin, 1991: 383). To add another layer of secrecy, the site was internally referred to as Bouchet III, the Bouchet I and II sites being other CEA premises in the same region: “Aside from some insiders, everybody would believe, until 1958, that [Bruyères-le-Chatel] was an extension of that [civilian research] center” (Buchalet, 1985: 48). Similarly, the on-site personnel were forbidden to reveal their true affiliation and instructed to say that they worked in an agricultural business. ²⁶ Officials had even regulated the surroundings as farmland to prevent construction (Buchalet, 1985: 48).

Specific practices and institutions were developed. Regarding personnel control, the practice of DST inquiry was abandoned in 1954 because “it turned out that the investigations, very superficial for that matter, were of a slowness incompatible with the rhythm” of CEA activities. ²⁷ Now, the CEA security service would be responsible for some investigations, and the DST would give opinions—an arrangement not agreed upon in the protocol designed only a year before. ²⁸ The CEA notably required to be informed on any opened case against a personnel member because of the “confidential, and even secret, nature of some research pursued by the Commissariat, and in particular of some results obtained.” ²⁹ In 1956, it was decided that the General Administrator would be solely responsible for security. The personnel were required to sign a declaration according to which they recognized that any violation of secrecy would be punished by forced labor—and not prison (Barrillot, 2005: 61). The military part of the CEA also started to use its classification system, SECRET ATOME, functioning outside of the boundaries of the normal classification. ³⁰

Because of this newly created secrecy capacity, during these years, the French nuclear program took place almost clandestinely. For four years, the public, the parliament, and most members of the Government would be excluded from nuclear policymaking. The historical record is extremely clear as to why secrecy was organized in such a manner. In line with Colaresi’s claim that secrecy regimes come with a potential for abuse, officials from the CEA sought to use secrecy as a resource for autonomy from both the public representatives and the rest of the government. There was a clear consensus among military and CEA officials that the public should not be involved, as its involvement could jeopardize the success of the program. General Bergeron’s plea for total secrecy over the program repeatedly referenced the need to wipe the project from all budget plans, or, at least, to wait for a few years before revealing it, by “integrating those expenses in other spendings.” ³¹ Keeping the program secret allowed to limit an outsider’s ability to get a proper idea of the French effort. But also, it ensured the absence of public debates on nuclear weapons. Bergeron considered public discussion on nuclear weapons “regrettable”: “such discussion, to stay objective (. . .) should even stay in a certain measure in the domain of secrecy.” ³²

Guillaumat was also convinced that information control was necessary to prevent public involvement in nuclear politics. As he declared years later, “public opinion, when consulted, takes whatever position. The only thing that matters is the opinion of the regular government.” ³³ Total secrecy allowed “avoid the illusion (on the right) and the anger (on the left) which might arise from the announcement of research for military application.” ³⁴ The Prime Minister who would endorse the program, Pierre Mendès-France, was also committed to total secrecy, essentially for political reasons. According to Jacques Hymans, secrecy was justified by the conviction that “if public opinion learned of these preparations the government would likely fall” (Hymans, 2006: 103).

During this four-year period, the nuclear secrecy regime played an exclusionary function. The Parliament was not asked to deliberate about this policy, since the choice of a nuclear program was kept secret. Oversight over policy was logically impossible too. Information control made the budgets unavailable to the Parliament. The program was organized and funded based on secret protocols and, with each new government, the head of government would be verbally informed about the program (Buchalet, 1985). To keep this financial contribution hidden, since budgets had to be presented to parliament yearly, vague names and slush funds were used. To account for the military funds transferred to the CEA, a special chapter in the military’s budget was created, simply called “Special studies” (Etudes spéciales). Its official purpose was to fund research into a nuclear-powered submarine. ³⁵ From this chapter, as well as from a fund connected to the Prime Minister simply named “Allowance to the CEA,” funds were transferred secretly, without specific terms. ³⁶ When the “Special studies” section’s creation came to be discussed in the parliament’s Defense Committee, MPs simply noted that it was “natural that the present report [did] not [insist]” more on it, considering that it referred to secret studies. ³⁷ The existence of a program was likely guessed by some. In June 1956, Senator Edgar Pisani defended a project to create a military division inside the CEA, which assumed that such an office existed inside the CEA “in semi-clandestinity.” ³⁸ After being passed by a majority in the Senate, the proposal then completely disappeared, and nothing came out of it. The Parliament and the public were effectively excluded from policymaking in that domain. Nuclear policy remained outside of democratic control as a direct consequence of how secrecy was organized and used strategically by actors seeking more autonomy and less public consultation over specific policy choices.

Democratic consequences of nuclear secrecy (2): the normalization of nuclear policy (1958–1974)

The 1954–1958 period arguably represents an extreme case. Secrecy was total and, consequently, democratic control was totally absent. It would not last forever. In 1958, following regime change and the accession of Charles de Gaulle to power, officials started to talk openly about nuclear ambitions. In October 1958, de Gaulle declared that “everyone knows that that we now dispose of the means to ensure a nuclear armament” and that a French test was approaching (Bendjebbar, 2000: 255).

This change in the public stance toward nuclear acquisition affected the organization of information control. The desire for clandestinity, which had defined how secrecy was organized, was no more. However, security concerns remained, meaning that the secrecy regime would remain in place. Nuclear secrecy slowly extended its reach as nuclear-related activities became more numerous, shielding nuclear politics from normal decision-making practices and perpetuating the de-democratizing effects of nuclearization.

First, personnel control practices evolved, with the official creation of the Department for Security and the Protection of Secrecy (DSPS) as an autonomous branch of the CEA in 1961. ³⁹ In 1963, new protocols were signed with the regular police and military security services, which gave extraordinary power to the CEA, so much so that a civil servant from the Homeland Ministry described it as a project to turn the DSPS into a “parallel police.” ⁴⁰ Agents even had the right to carry weapons, delivered by the CEA. ⁴¹ The inspectors were in charge of personnel control, and served to ensure the “secret nature” of the CEA’s work identified as “an essential objective for foreign intelligence service, necessitating a protection against all forms of penetration.” ⁴² Espionage, and particularly American espionage, were considered to be major risks by the Elysée. ⁴³ By 1966, this service had acquired functional autonomy from police services—to a degree that worried the DST. ⁴⁴ The CEA effectively ran its own police service, in charge of ensuring nuclear secrecy.

Sites related to the production of the nuclear arsenal were still shrouded in secrecy, even though their function in the French nuclear program was known to the national and international public. Security concerns were strong. For example, the official function of the Valduc center, where bombs were being assembled, was known, but it remained a highly secret site about which the CEA did not communicate (SEMIPAR, 2013: 3). As for Marcoule, the site remained so secret that the National Geographic Institute had to request special authorization to map the zone by plane, and its personnel would have to be vetted by the CEA even if they held a security clearance. ⁴⁵ After the creation of the French Strategic Air Forces (Forces Aériennes Stratégiques) in 1964, a new level of classification was created—“Très Secret Rubis”—to control high-level information related to nuclear policy and strategy. This special classification, still in existence today, is in effect the highest level of classification in France (Guisnel and Tertrais, 2016: 16; Warusfel, 2000: 194).

Regarding the once hidden budgets, the Military Program Law for 1960–1964 included for the first time an official military nuclear budget (Cardoni, 2022: 34–49). By late 1962, the rapid growth of the CEA’s size and mission led to calls for more control mechanisms at the budgetary level. ⁴⁶ Gaston Palewski, delegate minister for atomic affairs, refused to cede much ground as he considered autonomy “indispensable” for the CEA’s functioning. He nevertheless proposed the creation of a Finance committee, in charge of examining “the general policy of the Commissariat in financial matters,” which met for the first time in 1963. ⁴⁷ Its task was not to oversee solely military matters, but all nuclear matters. ⁴⁸ Fittingly, the very first point of its first meeting was a “reminder regarding the notion of secrecy.” ⁴⁹ Each budget demand superior to 2 m Francs was supposed to be examined by the commission. It seems, however, that “secret openings”—that is, requests for funds without specification of their use—were proposed in front of the committee, as a way of keeping secret the precise nature of the research done inside the laboratories. ⁵⁰ The CEA’s use of public funds was subject to some judicial control but the institution was granted the right to review judicial reports before publication to remove any information that might violate secrecy rules or leak strategic information. A procedure was also set up in 1960 which allowed the CEA to “apply, when establishing its accounting, all measures it judges necessary to preserve secrecy.” ⁵¹

This new information control regime was built according to a different logic than the “clandestine” one. It was built to fulfill the security needs of nuclear politics while accommodating democratic politics to an extent. Uncertainty remained the key rationale behind secrecy. Information control was designed to keep specific knowledges secret, but rather to maintain the possibility of secrecy over the nuclear enterprise to prevent competitors—that is, both allies and adversaries—from acquiring exact knowledge of the French arsenal. The known vulnerability of the early French arsenal (Pelopidas and Philippe, 2021) likely played a role in the officials’ decision-making process, though it was not articulated explicitly. Secrecy simply was perceived as an imperative in a domain where failure would be nothing short of intolerable. The exceptionality of nuclear secrecy remains today: in 2008, a reform of France’s archival regulation created specific procedures for archives related to nuclear policy. In practice, some documents can legally remain out of the public eye indefinitely (Laurent, 2015). Security requires that some secrets might need to be kept indefinitely.

These changes in the scope and form of information control regime logically imply that its consequences for democratic control were different in this period, compared to the clandestine one. The main change was the fact that nuclear policy could now become an object of public consultation. Actors once excluded from consultation had now more opportunities to participate. However, the persistence of secrecy and the strategic use of information control by actors mean that the de-democratizing effects of nuclear acquisition persisted.

The strictness of information control meant that the Parliament and the public became dependent on state officials to acquire information about policy choices. This allowed the continuation of a form of exclusion. Only a handful of actors were consulted on issues related to the definition of nuclear strategy. Decision-making in nuclear affairs was “totally exceptional,” recalled Minister for Armed Forces Pierre Messmer, because it “privileged certain military officials compared to many ministers,” the latter being kept in the dark. ⁵² The CEA notably attempted to keep the Prime Minister in the dark about its unauthorized work on thermonuclear technology. ⁵³ The president himself, Charles de Gaulle, was likely lied too about policy choices in terms of fissile material production (Pelopidas and Philippe, 2021: 252, fn. 49) and weapon safety. ⁵⁴ Some elements were also purposefully not communicated, such as the fact that “France will not use nuclear weapons on a non-nuclear power and when its territory is not threatened,” something the Minister for Armed Forces Pierre Messmer considered “good to know, but bad to say.” ⁵⁵

When MPs were consulted on policy, secrecy distorted the nature of the choices agreed upon. Though budgetary control was possible, secrecy remained a hurdle. A clear example of that is the case of Pierrelatte’s isotopic separation plant, which represented “the most considerable investment ever realized in France,” the CEA managed to conceal key information about costs, revealing only after a few years that the planned costs were many times larger than expected. ⁵⁶ Similarly, because the CEA’s activities also included the nuclear energy industry, nuclear secrecy also allowed the development of the civil nuclear sector in France “shielded from any form of publicity” (Laurent, 2024: ch. 2). Nuclear secrecy also diminished the level of accountability to which state officials were submitted. The clearest example of that are the radiological consequences of nuclear testing. Secrecy around test sites was strict, designed to prevent external actors from acquiring information about bomb design. At the same time, it allowed state officials to distort information and conceal to the local population that thousands were exposed to radioactive contamination during testing, and thus to escape accountability for the harm caused by nuclear activities (Fraise, 2023; Philippe et al., 2022).

There is supplementary evidence that mechanisms developed to control nuclear knowledge were also used for the control of political opponents in and around the CEA’s plants. The CEA’s security services were explicitly tasked with the surveillance of the “political and social climate” around plants, including unions and social movement. ⁵⁷ As historian Louis Fagon notes, the CEA’s targeting of communist unions was publicly justified by security concerns, but had in fact much to do with the desire to avoid strikes as well as political opposition to the bomb. (Fagon, 2023a: 107–108; 129–132).

The French case demonstrates the de-democratizing effects of nuclear acquisition. Even when nuclear policy was normalized, the level of public consultation remained diminished. Secrecy produced exclusion, limiting the number of actors involved in those consultations. It also produced a distortion of the information accessible to other actors, flawing deliberation processes. Finally, secrecy shielded officials from being called into account by concealing the very existence of forms of harm caused by state activities. France did not cease to be a liberal democracy because of its nuclear arsenal. However, the acquisition of nuclear weapons caused a lasting recoil over the public consultation axis. It created a new and highly consequential domain of policy, while at the same time creating a series of barriers to democratic control over this domain. De-democratization walks hand-in-hand with nuclearization.

Conclusion

Do states change when they acquire nuclear weapons? In this article, I have answered this question positively by showing that nuclear acquisition is not merely a process of material procurement, but also a process of political change. Using the case of the nuclear secrecy regime, I have shown how nuclear acquisition creates security challenges and anxiety among policymakers which leads to institutional change. These changes, in return, deteriorate a state’s ability to act democratically by creating obstacles to public consultation. Nuclearization, in essence, is a cause of de-democratization because it leads to the emergence of a “restricted” domain, limiting democratic control over policy issues with world-changing consequences. The case of France, studied in this article, provides evidence of this process and its mechanisms.

This study has its limitations. It is difficult to confirm a theoretical claim based on a single-case study, and future research should focus on a comparative research design studying the process of nuclearization and its consequences in different states engaged in a nuclear program. Moreover, this study has limited its focus to democratic states, leaving aside the question of nuclearization’s effects on non-democratic states. There is reasonable evidence that even in the USSR, where secrecy tended to be the default policy option (Harrison, 2023), nuclear secrecy was special. Not only was the Soviet nuclear program extremely secret, but it was also innovative compared to other domains of the Soviet state. In the early years of the Soviet program, typists were not trusted, and reports were written by hand, while the entire program was shrouded in extreme secrecy (Holloway, 1994: 202). Later on, Sredmash, the ministry in charge of Soviet nuclear weapons, operated in secret from the rest of the state, including its collaborators working on nuclear power (Schmid, 2015: 49–50). Research and production would take place in secret cities (Brown, 2015). Even in the USSR, nuclear policy seemed to have been an especially restricted domain of policy.

Shifting the focus from what states do with nuclear weapons to what nuclear weapons do to states opens up several paths for inquiry. Beyond nuclear secrecy, what are the other repercussions of nuclear acquisition? A productive path for inquiry could be the effects of command-and-control procedures on the executive branch of government. Have nuclear weapons favored concentration of power in the executive’s hands? Do nuclear weapons favor certain forms of political arrangements? The perspective developed in this article also raises questions about the resistance of certain state structures to change. If nuclear pursuit is bound to produce domestic change, what happens if certain actors are unable or unwilling to accept such change? A hypothesis to be pursued could be that some states may be too reluctant to implement necessary institutional changes and thus prefer to foreswear nuclear weapons altogether. Finally, the findings laid out above have implications for public debates about nuclear weapons and proliferation. Notably, they show that nuclear acquisition implies more than merely acquiring nuclear weapons; it has a real political cost for democracies. Originally built to defend democracy, nuclear weapons ended up hurting it. This should be a cautionary tale for democratic states considering going nuclear.