Issued by the FW de Klerk Foundation on 26/03/2024


Introduction: Opportunities, Challenges and Strategies

The global political atmosphere has become exceptionally dynamic, riotous and uncertain. Artificial Intelligence (“AI”) is disruptive, and disruption requires consistent and effective planning. AI offers immense power and benefits of predicting nuclear threats, solving complex problems, as well as transforming nuclear launches – wherein human autonomy is faced out. It is also, however, potentially destructive, as it could be used for nuclear proliferation. Hence, better and more nuanced conversations about what AI may mean for nuclear proliferation is needed.

This article is the second in a series on nuclear weapons and the increased threat these once again pose to civilisation. In this article, we will assess the opportunities and challenges that emerge from the ‘intersection’ between nuclear production programmes, artificial intelligence and cybersecurity. This analysis closes by assessing mitigation strategies in relation to AI integration in nuclear power. 


Origins of the Nuclear Age

World War II marked the onset of what historians would refer to as the “nuclear age”, characterised by the rapid development and proliferation of nuclear weapons among nations, notably the United States, Nazi Germany and the Soviet Union, as they vied for supremacy. This era reshaped global power dynamics and underscored the significance of technological advancements in warfare. 


AI Transforming Atomic Bombs? Studying the Impact

Fast forward to the present day, the advent of generative intelligence and machine learning heralds a new epoch – the technological age. As these technologies continue to evolve, they are poised to revolutionise various aspects of society, raising critical questions about their implications for international security and stability, including the potential risk of nuclear proliferation. 

With the rapid advancement of AI, the potential for AI-enabled cyber-attacks to compromise nuclear command and control systems has become a significant concern. Threats like jamming and hacking could illicitly breach nuclear weapons systems, posing grave risks to global security. Moreover, the proliferation of such capabilities raises the spectre of non-state actors, including terrorist groups, exploiting these systems to launch attacks or obtain sensitive information about nuclear weapons locations. 

It is important to note that there is also a widening opportunity for AI to aid in the illicit and convert production of nuclear weapons programmes (Figure 1). Nuclear material production phases include mining, milling, conversion, enrichment, fuel fabrication, energy generation and reprocessing – with the enrichment of uranium as one of the most crucial steps in attaining nuclear capability. Thus, improved access to technologies that are able to generate information in relation to these processes pose excessive risks.  

AI presents both vertical and horizontal proliferation risks. Vertical proliferation risks describes the stockpiling of nuclear weapons by nuclear powers to improve existing nuclear weaponry. Horizontal proliferation risks, by contrast, includes the spread of programmes of nuclear weapons to new countries. 

Figure 1: AI Revolutionizing Nuclear Materials, Towards Opportunities? 

Source: He and Degtyrev (2023).


A World Gone “M.A.D”? Understanding Warfare in a Digital Age

With the development of nuclear weapons, the theory of ‘Mutually Assured Destruction’ (“MAD”) gained traction. MAD outlines that if one nuclear armed state employs the use of nuclear weapons against another nuclear armed state, the state being attacked will retaliate with its own nuclear strike ultimately causing both parties mutually assured destruction. MAD, at least to a certain extent, has brought about stability, as states with nuclear weapons choose to avoid nuclear confrontation. 

The MAD logic may, however, be undermined with the emergence of AI: As military powers increasingly invest in harnessing the possibilities and capabilities of AI, which could ultimately result in AI improving militaries ability to locate, track, target and destroy a rival state’s nuclear weapons – effectively resulting in a zero-sum game, thus threatening nuclear escalation and peace. In April 2021, Israel employed this precise technology with the objective of causing physical damage to Iranian uranium enrichment facilities. 

AI may result in the modern-day usage of nuclear weapons, which will have negative impacts upon society and the environment. As of February 2024, nuclear powers China and Russia agreed to increase AI cooperation and capabilities aimed at enhancing military warfare and preparedness, while the U.S Defence Department begun to design AI programmes to aid in warfare. Hence, conventional warfare may be eventually replaced by hybrid warfare, whereby states engage in warfare with each other using conventional military means, AI and nuclear capability. 


Mitigating the Integration Between AI and Nuclear Programme

Ultimately, to maintain nuclear security within a thriving digital age, AI control measures must be adopted to effectively regulate and control its use. This strategy mirrors the approach pursued during the atomic age when nuclear experts sought to control the spread of nuclear weapons through international agreements like the Treaty on the Non-Proliferation of Nuclear Weapons, commonly known as the Non-Proliferation Treaty or (“NPT”). Hence, experts and sceptics of AI have called for the implementation of such non-proliferation regimes to be placed upon AI systems and development as well (Figure 2). 

Figure 2: Non-Proliferation Treaty during the AI Age, Towards Strategies?

Source: Scharre and Lamberth (2022).


AI, Atomic Bombs and Cybersecurity: Understanding Current Trends 

In 2023, the Annual Meeting of the Arms Control Association, in Washington D.C, regarding nuclear weapons stated that there would be a strong push for nuclear armed states to ban ever allowing AI to possess command and control functions over nuclear weapons systems. This highlights that at present, policymakers and experts recognise that there will be an influence, or there is, in fact, a flirtation amongst nuclear armed states and non-state actors with AI aimed at strengthening the advantages it may bring by infusing such technologies with the intention of driving greater nuclear competitive advantage. Hence, just as conflict experts view the physical conflict landscape to be constantly changing and evolving, it is crucial that security experts view the AI realm as constantly undergoing rapid change and evolution – which will bring with it limitless opportunity, but also countless challenges to the security landscape. 

Such advancements in AI also coincides with a time of heightened geo-political tensions between Russia and western nations, with Russian President Vladimir Putin uttering that Russia is ready for nuclear war should NATO nations send troops into Ukraine – after which French President Emmauel Macron uttered that he would not rule out sending western troops into Ukraine to fight alongside Ukrainian forces. 



We are at a juncture where the intersection of nuclear production programmes, AI development and cybersecurity present profound opportunities and challenges for global security. While AI indicates a new wave of human civilisation, it also poses a destructive threat. Thus, as AI advances, the risk of AI-enabled cyber-attacks undermining nuclear command systems grows. Arms control measures, akin to those implemented during the atomic age, are imperative to mitigate these risks and maintain nuclear security. It is crucial for policymakers to heed the calls for non-proliferation regimes to encompass AI systems and development, ensuring a safer and more stable world amidst evolving geopolitical tensions and technological advancements.