Potsdam, August 4, 2025
Eighty years ago, the world witnessed the catastrophic consequences of the first atomic bombs dropped on Hiroshima and Nagasaki. Yet it also saw the beginnings of a postwar order built on a commitment to peace. In July 1945, the Charter of the United Nations was adopted, and on August 1, 1945, the leaders of the United States, the United Kingdom and the Soviet Union signed the Potsdam Agreement.
A decade later, a group of scientists, led by Albert Einstein and Bertrand Russell, issued a manifesto urging the leaders of the world powers to do everything in their capacity to prevent nuclear war. They warned: “We have to learn to think in a new way… not what steps can be taken to give military victory to whatever group we prefer… but what steps can be taken to prevent a military contest of which the issue must be disastrous to all parties.”
This appeal had a lasting effect. Despite numerous conflicts over the decades, nuclear weapons – recognized as weapons of mass destruction – have never been used in war again.
Yet the world has not become peaceful. We face continued regional and local wars, with ever more sophisticated weapon systems. While governments have so far refrained from using the “nuclear option,” a new threat is emerging, one that raises fundamentally different but potentially comparably grave risks: autonomous weapon systems (AWS).
Eighty years ago, the nuclear bomb was heralded as a scientific breakthrough. But it also marked the beginning of a nuclear arms race. Today, artificial intelligence (AI) – powerful enough to enable the deployment of lethal autonomous weapon systems – is another transformative scientific advance. And once again, we are witnessing the emergence of an arms race, this time in AI.
We, the signatories of the “Potsdam Call on Autonomous Weapon Systems“, recognize that nuclear weapons and AWS are not the same. Nuclear weapons are designed for mass destruction. AWS, by contrast, enable targeted strikes. Yet their danger lies elsewhere: in the potential loss of meaningful human control. When machines are given the power to make life-and-death decisions, the threshold for violence may be dramatically lowered, and the scale of deployment could result in destruction on a scale that challenges existing legal and ethical frameworks.
On June 14, 2024, the late Pope Francis addressed the G7 leaders in Apulia, warning: “In light of the tragedy that is armed conflict, it is urgent to reconsider the development and use of devices like the so-called ‘lethal autonomous weapons’ and ultimately ban their use… No machine should ever choose to take the life of a human being.”
We, the undersigned scientists, are not naive. We understand that once war breaks out, all military options may be considered. But history shows that even adversaries with deep disagreements have been able to negotiate disarmament agreements – such as those concluded in the 1960s and 1970s – to reduce the risk of escalation. A similar path must be pursued for AWS.
As Einstein and Russell reminded us in 1955: “We appeal, as human beings, to human beings: Remember your humanity, and forget the rest.”
Eighty years ago, the city of Potsdam hosted the leaders of three major powers, who laid the groundwork for a postwar order aligned with the principles of the United Nations Charter, signed weeks earlier in San Francisco.
Eighty years later, we call on today’s world leaders to remember this legacy. Ensure that decisions over life and death remain in human hands. And commit to resolving conflicts not through automation and algorithmic targeting, but in accordance with international law and the shared sense of humanity.