The most common objection to international AI governance is that it is impossible. A century of evidence says otherwise. Global coordination on existential threats is not a utopian fantasy. It is a documented historical achievement, repeated across different technologies, different adversaries, and different political eras.
Every time humanity has faced a coordinated existential threat, the same objections have been raised: the problem is too complex, the adversaries too numerous, the verification too difficult, the defection too likely. Every time, those who pushed past those objections built frameworks that, imperfect as they are, prevented catastrophic outcomes.
The pattern is consistent enough to be called a law: when the threat is clear, the science is credible, and the advocacy is organised, international coordination becomes possible. Not easy. Not guaranteed. But possible.
The AI threat is clear. The science is credible. The advocacy is building. What is needed now is the political pressure that makes coordination not just possible but necessary. That is what the Nakada Foundation exists to create.
The four cases below are not cherry-picked successes. They are the defining examples of what human civilisation can achieve when faced with threats that no single nation can solve alone. Each began with the same dismissals that AI governance critics offer today. Each succeeded anyway.
Nuclear weapons were used in war in 1945. Within years, multiple nations had developed or were developing them. The prospect of dozens of countries with nuclear arsenals represented civilisation-ending risk. Critics said a non-proliferation treaty was impossible: too many actors, too many national interests, no way to verify compliance.
The Nuclear Non-Proliferation Treaty was opened for signatures in 1968 and entered into force in 1970. It created a framework where most nations agreed not to develop nuclear weapons in exchange for access to civilian nuclear technology. A verification regime through the International Atomic Energy Agency monitored compliance.
Today, nine countries have nuclear weapons when hundreds could have built them. No nuclear weapon has been used in war since 1945. The NPT is imperfect, it has not prevented all proliferation, but it has prevented the catastrophic scenario critics said was inevitable. Eighty years of nuclear peace did not happen by accident.
The International Atomic Energy Agency was established in 1957 with a mandate to promote peaceful nuclear energy while preventing its diversion to weapons. Its inspection and safeguards regime allows international inspectors to verify that nuclear material at declared facilities is not being enriched beyond civilian purposes.
This monitoring regime works because nuclear material, like AI compute, passes through identifiable chokepoints. Uranium must be enriched at specific facilities. Centrifuges are trackable. The IAEA developed the technical mechanisms for monitoring a technology whose dual-use potential posed existential risk.
The parallel to AI is direct. The hardware required to train frontier AI models, specifically the GPU clusters produced by NVIDIA and the advanced semiconductors fabricated by TSMC using ASML lithography equipment, is concentrated in ways that make monitoring technically feasible. A small number of companies control the chokepoints. Export controls already attempt to track these chips.
What the IAEA demonstrates is that international monitoring of dual-use technology is not a theoretical concept. It is an institution that has operated for nearly seventy years, conducting thousands of inspections across dozens of countries. The infrastructure for doing the same with AI compute exists. It requires political will, not technical innovation.
In the mid-1970s, scientists identified that chlorofluorocarbons, CFCs, used in refrigerants, aerosols, and industrial processes, were destroying the ozone layer that shields the planet from ultraviolet radiation. Without the ozone layer, life on earth faces catastrophic consequences. Industry said eliminating CFCs was economically impossible.
The Montreal Protocol was signed in 1987 by 46 nations and entered into force in 1989. It phased out the production and consumption of ozone-depleting substances on a binding international schedule. Developing nations received extended timelines and financial support. The treaty has been ratified by every nation on earth, the only international treaty with universal ratification.
The ozone layer is recovering. Scientific assessments project that it will return to 1980 levels by the middle of this century. The Montreal Protocol is widely described as the most successful environmental treaty in history. The chemical industry, which initially opposed it, developed alternative compounds and continued to operate profitably. The "impossible" turned out not to be.
Chemical weapons had been used in the First World War, in colonial conflicts, and in the Iran-Iraq War of the 1980s. The Chemical Weapons Convention, opened for signatures in 1993 and entering into force in 1997, banned the development, production, stockpiling, and use of chemical weapons and required verified destruction of existing stockpiles.
More than 190 nations have ratified the Convention. Nations with significant chemical weapons stockpiles, including the United States and Russia, destroyed them under international supervision conducted by the Organisation for the Prohibition of Chemical Weapons, which was awarded the Nobel Peace Prize in 2013.
The Chemical Weapons Convention demonstrates that international agreements can achieve verified destruction of weapons that nations have invested decades and billions in developing. The objection that countries will not give up strategic advantages if required by treaty has been proven wrong, repeatedly, across the most adversarial relationships in modern history.
Violations of the Convention occur and are met with international consequences. The framework is not perfect. It has, however, made chemical weapons use a norm-violating act that triggers global condemnation and response, where before the Convention it was a contested military option. That shift in the international norm is itself a form of governance success.
These four cases share a structure. In each, the threat was initially dismissed as too speculative or too technically complex for international governance. In each, the relevant industry argued that regulation was economically impossible. In each, verification was described as a fantasy. In each, the agreement was eventually reached, verification mechanisms were built, and the threat was materially reduced.
AI governance faces every one of these same objections today, in the same form, from the same types of actors. The technology companies say regulation will stifle innovation. The national security establishment says adversaries will defect. Commentators say the problem is too technically complex for legislators to understand. These are not new arguments. They are old arguments that have been wrong before.
The feature that distinguishes AI governance from these precedents is urgency. The NPT was negotiated over years while nuclear arsenals grew but remained human-controlled. The Montreal Protocol was implemented over decades while the ozone layer continued to thin but remained recoverable. AI development is moving faster, and the systems being built may, once sufficiently advanced, resist governance in ways that ozone molecules and nuclear warheads cannot.
This is not a reason to be pessimistic. It is a reason to act faster than the precedents required. We know coordination is possible. We know verification is possible. We know that industry opposition does not prevent international agreements. What we need now is the political will to compress the timeline, and that will is built by people who understand what is at stake and refuse to wait.
Nuclear weapons. Chemical weapons. The ozone layer. Global coordination on existential threats is a documented human achievement, not a hope. The question for AI is not whether it is possible. It is whether we will do it in time. The most common objection, answered →
See Our PlanJoin those who are building the political coalition that will make AI governance possible.