Though the number of nuclear weapons has gone down significantly since the 1980s, the current global stock of nuclear weapons is still high and rising. As of March 2024, the nuclear-armed states possessed a combined total of about 12,100 nuclear weapons. Roughly 3,804 are operational nuclear weapons. Of those, 2,000 are on “high-alert status,” which means they stand ready to be launched at any time.

Of these “Nuclear Nine,” nearly all actively maintain a policy of nuclear retaliation in self-defense. Further, only two states have vowed never to use nuclear weapons first, known as a “no first use” policy. There are also nuclear-capable states: those that do not possess physical nuclear weapons but have the resources and expertise to construct a successful nuclear program. Nuclear-capable states include: Iran, Japan, Germany, and South Korea.

Nuclear winter is a climate event that emerges from the large-scale use of nuclear weapons. If weapons are used on targets with large fuel loads, such as cities, the resulting firestorms would loft soot high into the atmosphere. The dust, smoke, and soot that then fill the atmosphere would then block sunlight from warming the earth’s surface, resulting in drastic temperature changes which would dramatically reduce food production for 10 years. A full scale nuclear exchange between the United States and Russia would cause below freezing summer temperatures and lead to the deaths of more than 5 billion in the first 2 years alone. However, military doctrine and planning traditionally fails to account for nuclear winter. 

In a crisis, leaders may only have 15 minutes to decide on a response to an incoming attack. The nine nuclear powers possess varied command and control structures for approving a nuclear launch. Despite differences in governance structures, the decision to launch a nuclear strike indeed often remains in the hands of the president or head of state to enable them to respond in time. In the US, France, and North Korea, the head of state has sole authority to launch nuclear weapons. Other nuclear launch authority structures often take the form of a small, 10 to 15 person decision-making body, but it is unclear how many of these states would function in a nuclear crisis. Once weapons are launched, there is no way to recall or remotely disable them.

For instance, in 1983, Stanislav Petrov, a Soviet officer, received early warning signals suggesting five incoming U.S. ICBMs. Despite protocol requiring him to report the alerts, which could have led to a retaliatory nuclear launch, Petrov chose to disregard them based on his intuition that the signals were a false alarm, preventing an accidental nuclear exchange. The readings were later found to be a malfunction from sunlight glinting off of clouds. Beyond 1983, there have been several more occasions when the world came dangerously close to a nuclear exchange—the Cuban Missile Crisis and the Able Archer 83 incident are just two more examples.

During the Cold War, the world only had to truly maintain deterrence between the US and the USSR. Today, with the US, China, and Russia possessing comparably large nuclear arsenals, the strategic dynamic has evolved. Since arsenal sizes are predicated in part on the ability to destroy an adversary’s arsenal, between three states attempting to deter one another, the calculus for how much offensive power is needed to achieve such a “counterforce” strike has changed. Negotiations also become harder between three parties. As other states such as India rise to world power status, this strategic dynamic is likely to become even more uncertain.

Many of these international legal frameworks have expired or been withdrawn from in the last several years. Many of the remaining treaties will soon expire. These treaties include: the Strategic Arms Limitations Treaty (SALT), the Comprehensive Test Ban Treaty (CTBT), New START, JCPOA, and the Treaty on Conventional Armed Forces in Europe (CFE). The expirations, as well as the current global security climate, have weakened the effectiveness of treaties that still exist, like the NPT. Furthermore, amidst rising tensions, it seems unlikely that new treaties will be negotiated.

With increased overlap of nuclear and conventional systems, adversaries will struggle to distinguish which systems are intended for conventional or nuclear strikes. This could lead to inadvertent escalation where neither state wants to use nuclear weapons but feels they face a “use it or lose it” situation. For instance, a conventional strike in a conventional conflict on an element of an adversary’s command, control, and communications platforms (C3), such as a radar station or satellite, may unintentionally make an adversary feel that their arsenal is threatened if it introduces communication gaps and information shortages in their nuclear chains of command. Ambiguity over whether an incoming missile is nuclear-armed may force a state to make hasty and escalatory decisions. Even in the absence of an active conflict, an adversary’s build up of dual-use weapons systems may make a state uncertain about the other’s intentions.

Globally, only roughly $40-50 million per year is invested by philanthropy in mitigating nuclear risk. That number is expected to drop as the MacArthur Foundation, previously the source of around 30% of funding, had left the field as of 2024. In comparison, defense companies spent $117 million lobbying governments in 2020 alone — making a return of $236 in nuclear weapon-related contracts per dollar spent. Governments worldwide are increasing their spending on nuclear weapons with $91.3 billion spent in 2023. While this spending doesn’t necessarily mean a more dangerous world, it shows that states still view nuclear weapons, and the threat of their use, as key to their security.

While states are not seriously considering giving AI models direct control of nuclear weapons, militaries are actively looking to incorporate this emerging technology into a wide array of applications, including assisting military decision making and early warning sensors. AI models are susceptible to a wide range of failures and attacks. In the conventional realm, the technology might cause unintended escalation through automated procedures and shortened decision times, especially when military forces are deployed during a crisis. Even while working as intended, these tools can enable states to better target previously hidden weapons such as submarines or mobile launchers leading states to perceive their nuclear forces are vulnerable. In a domain where even small mistakes could be catastrophic, the past failures of these new technologies is concerning.