Observers writing for Foreign Affairs warn that the rapid advance of artificial intelligence is becoming a direct threat to nuclear security. In their view, deepfakes generated by AI could push leaders of nuclear-armed states toward launching a strike, while an even more dangerous scenario would be one in which an AI system is given the authority to decide on the use of nuclear weapons. The authors argue that the rapid spread of artificial intelligence has already worsened risks to nuclear stability.

The article recalls that in December this year, Philip Schellekens, chief economist at the Asia-Pacific bureau of the UN Development Programme (UNDP), said the use of AI in the military sphere could take on an existential dimension and lead to mass casualties. He stressed that states have an obligation to regulate such technologies and expressed concern over the possibility of delegating nuclear launch decisions to machines.

Foreign Affairs also notes that AI-based tools make it much easier to produce deepfakes — highly realistic fake videos, images and audio recordings used to spread false information about people and events. This instrument, the authors argue, is extremely dangerous: such fabrications can be used to create a pretext for war, to boost public support for a conflict, or to sow confusion and unrest.