AI and the future of mutually assured destruction Will nuclear deterrence survive this revolution?

The swift progress of artificial intelligence in recent years has led many observers to predict it will disrupt global politics and the military balance. Some have even argued that advanced AI systems might “establish one state’s complete dominance and control, leaving the fate of rivals subject to its will.” Yet some experts argue that as long as nuclear deterrence systems function, the advantages conferred by AI will not enable any state to fully impose its political will on others.

AI is undeniably transformative, with the potential to enhance the economic, political, and military pillars of national power. Yet winning the AI race will not automatically grant a state unchallenged dominance over major rivals.

As an article published by the Foreign Affairs magazine points out, the enduring power of nuclear weapons—the most consequential invention of the last century—remains a powerful brake on the sweeping change AI might otherwise bring. 

The article provides the case of the US as an example, noting that it boasts an economy nearly 15 times the size of Russia’s and roughly 1,000 times that of North Korea’s, yet still struggles to compel either nation’s compliance—largely because both possess nuclear arsenals.

Some experts contend that AI could alter this equation. To erode nuclear deterrence, AI would have to dismantle its central premise: a state’s ability to launch a devastating nuclear counterstrike after absorbing a nuclear attack, known as second-strike capability.

In theory, AI might make a “splendid first strike” more feasible by pinpointing the locations of nuclear submarines and mobile launchers. It could also cripple a rival’s retaliatory capacity by disrupting command-and-control systems or bolstering missile defenses enough to nullify threats of reprisal.

Such breakthroughs could allow a state to escape mutual assured destruction, making it uniquely able to coerce adversaries—a vision increasingly common in discussions of AI-enabled dominance.

Still, tipping the nuclear balance is far from straightforward. Even the most advanced AI-powered detection and targeting systems may fail to spot a mobile launcher concealed under a bridge, distinguish a nuclear-armed submarine’s signal from ocean noise, or flawlessly coordinate simultaneous strikes on hundreds of land, sea, and air targets.

Since the dawn of the atomic age, adversaries have consistently countered each new technological challenge with defensive innovations of their own—and would do so again.

Even if AI fails to overturn nuclear deterrence, the article argues that it could still foster suspicion and risky behavior among nuclear-armed states. Measures aimed at hardening second-strike capabilities could alarm opponents, fueling costly and dangerous arms races.

There is also the possibility that AI could suddenly leap in capability, giving its possessor a sharper and harder-to-counter advantage.

Foreign Affairs urges policymakers to prepare for such a scenario by fostering regular dialogue between AI and nuclear specialists.

Leaders should also act to reduce risks of accidents or unintended escalation by evaluating nuclear systems for AI-related vulnerabilities and keeping communication channels between nuclear powers open. These efforts would help preserve nuclear stability—and not just deterrence—amid the rise of AI.

By Nazrin Sadigova