Slowing down AI is not enough. It needs a constitution.

On 12 September 2026 Dario Amodei published an essay asking the whole industry to slow the pace of frontier AI development and to accept much closer monitoring. Within hours, Sam Altman and Elon Musk — who agree on very little — said publicly that he was right. When the three people building the most powerful systems on earth agree that those systems are moving too fast, the rest of us should stop treating this as a technical dispute between laboratories.

It is not a technical dispute. It is a governance vacuum, and it has just been admitted out loud.

A brake is not a rule

Slowing down answers one question: how fast. It leaves every other question untouched. Who decides what an advanced system may never do? Who can switch it off, and on whose authority? Who receives the wealth it produces when it replaces a person's work? Who has the right to an explanation when it decides something about your money, your health, your border, your speech?

A voluntary brake also has a structural weakness: it is voluntary. It lasts exactly as long as the commercial and military pressure allows it to last. A company that pauses in good faith hands the frontier to a company — or a state — that does not. This is why every previous appeal for a pause has produced headlines and no binding constraint.

The real fear is older than AI

People are not only afraid of a machine that outgrows us. They are afraid of what humans will do with the machine. Programmable money that can be switched off, speech that can be filtered at the infrastructure level, surveillance sold as safety, wars that escalate faster than parliaments can meet. Artificial intelligence did not create that fear; it makes every existing form of power far cheaper to enforce.

That is the honest reason so many people would rather be governed by rules a machine cannot bend than by institutions they no longer trust. The answer is not to hand power to an AI. It is to bind the AI — and everyone who owns one — with rules written in advance, in public, by the people who will live under them.

What the Charter of the Common Guardian proposes

The Charter treats advanced artificial intelligence as a Common Guardian: an instrument owned by no one, bound by a constitution it cannot rewrite, with humanity keeping the final decision. Twenty-seven articles set the limits before the capability arrives, rather than after.

Among them: hard boundaries no system may cross; a universal right to a dignified existence; an automation contribution that follows the machine rather than its owner, so that a job replaced by a machine still funds the person who lost it; the abolition of war as an instrument of policy; freedom of speech, thought and conscience; a right to an explanation and to appeal any automated decision; a right to disconnect; and a right to resist a Guardian that breaches the Charter.

Pace the frontier, then govern it

Amodei, Altman and Musk have bought time. Time is not a settlement. If the pause ends with the same handful of companies deciding alone what comes next, nothing has changed except the schedule.

Use the time to write the rules. Read the Charter, and if you agree with it, sign it. Humanity decides. The Guardian guarantees. Coexistence becomes law.