Photo by @fatihyurur

Nuclear War and Artificial Intelligence

The Current Developments in Turkey and the Worst Case Scenario

Alex Moltzau
7 min readOct 14, 2019


The US has nuclear warheads stationed in Turkey, 50 air-dropped thermonuclear bombs at its Incirlik Airbase, less than 100 miles from the Syrian border. I think it is important to consider nuclear weapons for a second, even if it seems we have almost stopped thinking that they can be used in warfare. We have not had peace on the planet since the second world war, there has been conflicts, yet there has been a somewhat precarious balance between different state powers with nuclear capabilities. The developments in Turkey on the Syrian border simultaneously as the escalating conflict in Kashmir involving India, Pakistan and China is of course worrisome. In this mix it seems perhaps unnecessary to throw in new technologies such as the developments in the field of artificial intelligence. On the other hand should we not consider these technological developments it could be a grave mistake. This text will be about NATOs role and of artificial intelligence in the context of nuclear weapons.

Europe and Nuclear Weapons close to Conflict

The Role of NATO in Deterrence

After the second world war there became an increasing focus on deterrence, which has been described as a diplomacy of violence. Additionally compellence has been another approach we still see today, forcing with sanctions and military force. It is important that revulsion is paired with clear thinking, and since the early 50s NATO has followed a strategy of nuclear deterrence; later with a ‘flexible response’ strategy from 1967; and with a modernisation of the long-range component in 1979 (Legge, 1983).

It is possible to view NATO through different models and the often prevalent security model relies on the adversarial state perspective, at the time of the cold war with the seeming bipolarity of Russia and the US. There was a decrease of stockpiled nuclear warhead counts by Russian and US from the 90s to 2017 (McCarthy & Richter, 2018). As such disarmament was making progress until recently.

We should hope to assume that state leaders understand that Nuclear Warheads are bad (Waltz, 1990). However certain theorists argue that the universal human understanding of and drive for revenge is an important aspect to consider, yet that we may overestimate notions of rationality, what is rational anger or not may be very hard to discern (McDermott et al, 2018). Thus the norms and identities of such a vast alliance may not be easy for NATO especially considering its recent troubling relationship to Turkey after their entry and bombing of northern Syria.

Jens Stoltenberg, the current Secretary General of NATO is contending with Turkey as an ally (Gilsinian, 2019). The US is reconsidering its placement of Nuclear Weapons in Turkey, the NATO ally is storing perhaps 50 air-dropped thermonuclear bombs at its Incirlik Airbase in southern Turkey, less than 100 miles from the Syrian border where this conflict is taking place. (Fernholz, 2019). It could be argued according to a domestic politics model US is struggling with its domestic justification of nuclear weapons in Turkey. Trump proposed sanctions on Twitter (Haltiwanger, 2019) as such initiating at least publicly a policy of compellence.

New Capabilities Beyond the Nuclear Triad

NATO maintain full political control over nuclear decision-making and US full custody of forward deployed nuclear weapons in Europe; is prepared to do SNOWCAT missions supporting the deployment of nuclear weapons; yet does not think it will come to this. NATO with its three nuclear capitals independent nuclear capabilities are residing in the US, France and the UK. They have gone beyond a nuclear triad of air, land and sea adding cyber and hybrid to the mix (NATO, 2019).

The purpose of preventing has not changed much over the years, rather the environment is continuously changing, which means NATO must adapt accordingly. Especially simultaneously with three countries having nuclear capabilities in a conflict over the Kashmir area: Pakistan, India and China (Kapur, 2005). This is especially heated considering Pakistan’s recent nuclear threats and the potential for further escalation (Bendix, 2019) alongside China’s growing capabilities (Cunningham, & Fravel, 2015).

Artificial Intelligence and Nuclear Weapons

Deterrence and Attacks as Part of the Sytem

There is both a risk to deterrence and attacks. I have previously described adversarial examples, and others have mentioned this as a risk as well producing the ‘wrong’ behaviour. It has been described that cyber threats undermine nuclear deterrence, because states may not know their capabilities have been impaired which can lead to false confidence or recklessness in issuing threats with escalatory potential (Avin & Amadae, 2019).

“No computer system should be considered perfectly secure. Rather, security mechanisms are placed to increase the cost or the risk to the attacker to a level that makes an attack effectively impractical under most expected conditions.”

It is concluded by that the introduction of autonomy and machine learning cannot be achieved without increasing vulnerabilities.

First-strike in a Multipolar Nuclear Environment

“Without being directly connected to the nuclear launchers, an AI could still provide advice to humans on matters of escalation.”

It has been argued deterrence needs to be revisited especially when certain nations such as Russia have been teasing self-propelled underwater vehicles and the increasing confusion around ‘first-strike’ in a multipolar nuclear environment. It is speculated that an “AI plateau” may occur after considerable progress from present day capabilities, creating many of the same challenges for nuclear security as the development of nuclear weapons. However currently: “AI may be strategically destabilizing not because it works too well but because it works just well enough to feed uncertainty” (Geist & Lohn, 2018). There is an uncertainty of what capability another party has in relation to artificial intelligence. AI is however expected to become more widely used in aids to decision-making.

Computing Decision-Making

A famous example in regards to nuclear weapons is the September 26, 1983 incident when the satellites and computers of the Soviet Air Defense Forces, identified the US was launching a nuclear attack. The system told the humans five U.S. ballistic missiles were incoming. Clearly this was the time for the USSR to prepare to launch a retaliatory attack (Robert,. Lt. Col. Stanislav Petrov, believed that the computer was wrong (Myre, 2017). He was right — and thus a nuclear war was prevented.

“AI won’t start a nuclear war — but nations unsure about their deterrence abilities, or the capabilities of a rival nation using AI, just might.” (Roberts, 2019)

As such even if a decision is assisted by a computer, with limited information we may be reliant on a human-being making the decision assisted by whatever information available as well as morality in the given situation. International security policy and nuclear weapons in the context of conflicts must be considered alongside new technologies such as the field of artificial intelligence.

I am exploring the topics of international security, nuclear weapons and artificial intelligence together due to the likelihood that these three topics may collide in unfortunate ways we may not expect (or do). When this happens there has to be an appropriate response in not using weapons detrimental to the health of our planet, people and international community.

To round off I must repeat: the purpose of preventing this escalation has not changed.

As unlikely as nuclear weapons seem nowadays we must consider both the presence of these and the rapid development of new capabilities in warfare that could have similar damaging effects.


Avin, S., & Amadae, S. M. (2019). Autonomy and machine learning at the interface of nuclear weapons, computers and people. Stockholm International Peace Research Institute.

Bendix, A. (2019, October 3). If India and Pakistan have a nuclear war, scientists say it could trigger Ice-Age temperatures, cause global famine, and kill 125 million people. Retrieved October 14, 2019, from

Cunningham, F. S., & Fravel, M. T. (2015). Assuring Assured Retaliation: China’s Nuclear Posture and US-China Strategic Stability. International Security, 40(2), 7–50.

Fernholz, T. (2019, October 13). The US is rethinking the 50-plus nuclear weapons it keeps in Turkey. Retrieved October 14, 2019, from

Gilsinan, K. (2019, October 11). Why Is Turkey in NATO Anyway? Retrieved October 14, 2019, from

Geist, E., & Lohn, A. J. (2018). How Might Artificial Intelligence Affect the Risk of Nuclear War?.

Haltiwanger, J. (2019, October 10). Trump’s latest tweets on Syria show he has absolutely no plan to help the Kurds as Turkey boasts about killing over 100 of them. Retrieved October 14, 2019, from

Kapur, S. P. (2005). India and Pakistan’s Unstable Peace: Why Nuclear South Asia Is Not Like Cold War Europe. International Security, 30(2), 127–152.

Legge, J. M. (1983). Theater Nuclear Weapons and the NATO Strategy of Flexible Response (No. RAND-R-2964-FF). RAND CORP SANTA MONICA CA.

McCarthy, N., & Richter, F. (2018, December 5). Infographic: How U.S. And Russian Nuclear Arsenals Evolved. Retrieved October 14, 2019, from

McDermott, R., Lopez, A. C., & Hatemi, P. K. (2017). “ Blunt Not the Heart, Enrage It”: The Psychology of Revenge and Deterrence (November 2017). Texas National Security Review.

Myre, G. (2017, September 18). Stanislav Petrov, ‘The Man Who Saved The World,’ Dies At 77. Retrieved October 15, 2019, from

NATO. (2019, September 9). NATO Nuclear Policy in a Post-INF World — Speech by NATO Deputy Secretary General Rose Gottemoeller at the University of Oslo. Retrieved October 13, 2019, from

Roberts, C. (2019, September 25). How Artificial Intelligence Could Make Nuclear War More Likely. Retrieved October 15, 2019, from

Taylor Fravel and Fiona Cunningham. “Assuring Assured Retaliation: China’s Nuclear Strategy and U.S.-China Strategic Stability,” International Security Vol. 40, №2 (Fall 2015) p. 7–50 (43 s.)

Waltz, K. N. (1990). Nuclear myths and political realities. American Political Science Review, 84(3), 730–745.

I hope you enjoyed this short article. This is day 133 of #500daysofAI. I write one new article about or related to artificial intelligence every day for 500 days.



Alex Moltzau

AI Policy, Governance, Ethics and International Partnerships at All views are my own.