Standing before the Arms Control Association, United Nations Secretary-General António Guterres described a world balanced on a knife's edge, where the danger of nuclear use has not been this high since the Cold War. The pressures he named are old fears in new clothes: an arms race turning qualitative, artificial intelligence amplifying risk, nuclear threats returning as a kind of blackmail, and the treaties meant to restrain these weapons slowly losing their grip. His most pointed request was that, until such weapons are abolished, the decision to use them must remain in human hands rather th
UN Chief Guterres Warns AI Must Not Share Control of Nuclear Launch Decisions
If it was up to algorithms, clouds could have ended the world.
Mark
Guterres says the nuclear risk is the highest since the Cold War. What's new in his message?
Mimi
Mostly the pairing. He ties the usual concerns, like weakening treaties and nuclear threats, to AI as a multiplier of danger, and then asks for a specific pledge: humans make any nuclear-use decision.
Mark
And he aims that at the US and Russia in particular?
Mimi
Yes. He asks them to return to talks, fully implement New START, and agree on a successor.
Luke
Worth saying what this is. It's a speech and an appeal. The reporting doesn't show any state agreeing to the human-control pledge.
Mark
Why bring up the Soviet dead hand system, then?
Mimi
It shows automation in nuclear planning isn't hypothetical. If communication lines went dead, launch authority passed to lower-level operators in a bunker.
Luke
Though the piece says it was never activated, so it's a precedent for the design, not for any actual use.
Mark
And the 1983 incident?
Mimi
A detection system appeared to show five missiles. Stanislav Petrov doubted it and didn't pass on a launch recommendation. It was sunlight on high-altitude clouds.
Luke
That's a strong example of human judgment working. But the article doesn't say how current AI systems are used in nuclear command today, so that gap stays open.
Le Pouls
- The UN Secretary-General warns that humanity stands on a knife's edge, with nuclear danger at its highest since the Cold War.
- Reckless nuclear threats are resurfacing while the arms control framework that once restrained them is visibly fraying.
- Guterres presses the United States and Russia to return to negotiations, fully implement New START, and craft a successor treaty.
- His core demand is blunt: no nuclear launch decision should ever be handed to algorithms, and the matter now turns on whether states will bind themselves to that principle.
Standing before the Arms Control Association, United Nations Secretary-General António Guterres described a world balanced on a knife's edge, where the danger of nuclear use has not been this high since the Cold War. The pressures he named are old fears in new clothes: an arms race turning qualitative, artificial intelligence amplifying risk, nuclear threats returning as a kind of blackmail, and the treaties meant to restrain these weapons slowly losing their grip. His most pointed request was that, until such weapons are abolished, the decision to use them must remain in human hands rather than those of machines or algorithms.
António Guterres stood before the Arms Control Association and said the world was on a knife's edge. The risk that nuclear weapons might be used has climbed to levels not seen since the Cold War, and he pointed to a tangle of causes: a qualitative arms race, artificial intelligence multiplying the dangers, reckless nuclear threats, and a regime of treaties that is weakening. His message was simple: disarm now.
He asked nuclear-armed countries to lead, and singled out the United States and Russia, urging them back to the negotiating table to fully implement New START and agree on a successor. His sharpest request concerned who holds the final call. Until nuclear weapons are gone, he said, every nation must agree that the decision to use them belongs to humans, not machines.
This is not a purely hypothetical worry. The Soviet Union once built a 'dead hand' system that would automatically authorize retaliation if the command structure were destroyed, and if communication lines stayed open it shut itself down, leaving the fate of the world to a junior officer and a computer otherwise. It was never activated. In 1983, a Soviet missile-detection system falsely reported five incoming American missiles, and Stanislav Petrov, a skeptical officer, refused to pass the alarm up the chain. Sunlight on high clouds had been mistaken for an attack. The lesson the article draws is quiet but firm: left to an algorithm, clouds could have started a catastrophe. The question now is whether nuclear states will commit in binding terms to keeping human judgment at the center of any use.
António Guterres, the Secretary-General of the United Nations, stood before a meeting of the Arms Control Association and told the room that humanity is "on a knife's edge." The risk that nuclear weapons will be used, he said, has reached heights not seen since the Cold War.
His speech named several pressures at once. States, he said, are engaged in a qualitative arms race. Technologies like artificial intelligence are multiplying the danger. Nuclear blackmail has reemerged, with some actors recklessly threatening nuclear catastrophe. Meanwhile, the regime built to prevent the use, testing and spread of nuclear weapons is weakening. His conclusion was blunt: "we need disarmament now."
Guterres urged all countries to disarm, and asked those that already hold nuclear weapons to lead the way. He singled out the United States and the Russian Federation, calling on them to return to the negotiating table, fully implement the New START treaty, and agree on a successor to it.
His most specific demand concerned who makes the final call. Until nuclear weapons are eliminated, he said, all countries must agree that any decision on nuclear use is made by humans, not by machines or algorithms.
That may sound like a distant worry, but automation has already played a part in nuclear planning. The Soviet Union built a "dead hand" system meant to guarantee retaliation even if its command structure were destroyed by a nuclear blast. It monitored seismic activity, radiation levels and air pressure for signs that a weapon had struck the country. If it detected such an attack, it then checked whether communication lines between top Soviet officials were still open.
If the lines were open, the system shut down after 15 minutes, since open lines meant people who could decide on a launch were still alive. If the lines were dead, authority to launch passed to lower-level operators of the system inside a protected bunker. That would have left the fate of the world with a junior officer and a computer system. The system was never activated.
The human check has mattered in at least one documented case. On September 26, 1983, a Soviet missile-detection system appeared to show five nuclear missiles heading toward the Soviet Union. Stanislav Petrov, a Soviet military officer, was skeptical of the reading and declined to tell Soviet command to launch a retaliatory strike. The alarm turned out to be false: sunlight glinting off high-altitude clouds had looked like a missile attack in the satellite data.
The article's own conclusion is dry but pointed. Had the decision been left to an algorithm, clouds could have set off a catastrophe. The thing to watch is whether nuclear-armed states move toward binding commitments that keep human judgment at the center of any decision to use these weapons.
Citations marquantes
Until these weapons are eliminated all countries must agree that any decision on nuclear use is made by humans; not machines or algorithms.— António Guterres, UN Secretary-General
Humanity is on a knife's edge.— António Guterres, UN Secretary-General
Envie de l'histoire complète ?
Lire l'original sur IFLScience ↗
Technology