Foreign Affairs has published an analytical article by Erin Dumbacher warning that the rapid development of artificial intelligence and deepfake technology could create unprecedented risks for global security, particularly in the sphere of nuclear deterrence. The author argues that the combination of automated early-warning systems and AI’s ability to generate convincing fake videos, images, and audio recordings could lead to fatal mistakes at critical moments of decision-making.

The author recalls that since the Cold War, humanity has repeatedly come close to catastrophe because of false signals from early-warning systems. One of the best-known examples is the 1983 incident in which a Soviet system mistakenly detected what appeared to be the launch of American nuclear missiles. The world avoided disaster only because the duty officer chose not to trust the automated alert. Today, the author warns, similar scenarios could unfold in even more dangerous forms due to the use of artificial intelligence.

Particular concern surrounds deepfakes, which can convincingly imitate statements by political leaders or create the illusion of military actions that never actually occurred. The article references cases in which such fabricated videos have already been used for disinformation during Russia’s war against Ukraine and warns that, in a nuclear confrontation, even a few minutes of misinformation could have irreversible consequences.

The author also notes that some countries, including the United States, are actively integrating AI into defense and security structures. At the same time, she stresses that nuclear command, control, and early-warning systems must remain under strict human oversight. In her view, automation in this sphere carries more risks than benefits, as algorithms can make mistakes, “hallucinate”, or become targets of deliberate manipulation.

The article further highlights the need to reassess nuclear doctrines developed during the Cold War era. Today’s information environment, saturated with fake content and AI-generated material, requires new approaches to data verification, crisis communication, and decision-making. Experts call on nuclear-armed states to agree on a fundamental principle: no decision to use nuclear weapons should ever be made without direct and conscious human involvement.