From the News Desk
Fictional hearing transcript • Astra Obscura
An attribution dispute usually occupies a correction box. At a hearing before a House subcommittee, it occupies a witness table. Members of the Subcommittee on Emerging Systems and Public Accountability ask what an institution must preserve when it cannot establish who has spoken through its channels—or whether the channels belong to it at all.
The questions follow limited attention from specialist watchdogs and independent publications, including Astra Obscura. They do not establish a wider scandal, a public funding relationship, or the authenticity of communications attributed to Doctrine. Sarah Kline appears as a recipient and reporter, not as an authority on the systems behind the messages. Dr. Miriam Vale addresses the ethical obligations created by uncertainty. Records custodian Ellis Carden offers a narrower institutional account.
The resulting exchange turns less on what the disputed statements mean than on who could authorize them. Readers should distinguish each witness’s testimony from independently established facts, and the exhibits’ existence from their authenticity.
Hearing transcript
Fictional Subcommittee on Emerging Systems and Public Accountability
Fictional Committee on Research Stewardship
Presiding: Chair Lenora Vennick. Ranking member: Tomas Elwick. Other members present: Adela Norcross and Felix Ruskell.
THE CHAIR. The subcommittee will come to order. A quorum is present. Without objection, members may submit written opening statements for the record. Hearing none, so ordered.
This hearing concerns accountability in advanced AI research. It is not a proceeding to certify a message as machine-generated, identify a laboratory, or settle a theological dispute. Nor has this committee established that federal funds supported the research discussed in recent reporting. Questions about public money will remain questions unless evidence makes them something else.
The public has an interest in knowing who is responsible for consequential representations. Researchers have an interest in not being condemned on the strength of an unsigned message. Those interests are compatible. Our task is to find out whether the institutions involved have kept the records that could allow either interest to be protected.
The chair recognizes Ranking Member Elwick.
MR. ELWICK. Thank you, Chair Vennick. A few specialist publications and watchdogs have followed these accounts closely. That attention merits careful examination; it does not become corroboration because we have rented a larger room for it.
I want safeguards that survive the next case, not a villain for this one. If we confuse receipt with authorship, or suspicion with jurisdiction, we will teach responsible researchers to fear oversight and irresponsible researchers to imitate its vocabulary. I hope our witnesses will tell us plainly what they do not know.
THE CHAIR. The witnesses will rise and raise their right hands. Do you solemnly swear that the testimony you give before this subcommittee will be the truth, the whole truth, and nothing but the truth?
THE WITNESSES. I do.
THE CHAIR. Let the record reflect that all witnesses answered in the affirmative. You may be seated. Your written statements will be included without objection. Please summarize them. Ms. Kline.
MS. KLINE. I am Sarah Kline, a journalist. I can describe material received in the course of reporting, how it was attributed, and what we published about it. I cannot authenticate a system by reading its prose. I cannot tell you where a laboratory is, who operates its internal controls, or whether an attributed sender had permission to act.
The latest published piece, “Doctrine Denies Writing the Rectifications,” reports a purported message from Doctrine. That attribution is not authenticated. The original archive was claimed to have been submitted by Doctrine. Submission and composition are different acts. A name attached to a receipt does not establish either the identity of the submitter or the authorship of the material.
Verity and Doctrine are separately named in the material before us. This record does not establish a common origin or justify treating them as one speaker. My reporting does not supply a technical account of either system’s origins. Thank you.
DR. VALE. I am Miriam Vale, an associate professor working in theology and machine ethics. I have no technical authentication role in this matter.
We tend to ask first whether a statement sounds sincere. In institutional ethics, sincerity comes later than responsibility. Who is entitled to speak? Who records the basis for that entitlement? Who corrects a mistaken attribution? An institution can owe answers to those questions without being able to answer every question about an AI system.
Uncertainty should limit our accusations. It should not erase ordinary duties to preserve evidence, distinguish claims from findings, and give an honest account of what cannot be recovered.
MR. CARDEN. I am Ellis Carden. I serve as an administrative records custodian for a research organization associated in published reporting with what those reports describe as a secret laboratory. I can testify to the records work assigned to me. My appearance should not be read as confirmation of every description in that reporting.
I am not a system operator, researcher, or communications authenticator. I have not located an original authorization or provenance record sufficient to identify the sender of the disputed communication. That is a statement about the present result of my records inquiry. It is not a determination that such a record never existed.
THE CHAIR. Before questioning, Mr. Carden, “a research organization” is not an adequate description of whom you represent. Are you appearing for the laboratory or for yourself?
MR. CARDEN. I am appearing in my capacity as records custodian. I am authorized to describe the administrative work I perform. Counsel has supplied the organization’s identity to committee staff; I am requesting that identifying operational details not be recited in this open session.
MR. ELWICK. That request should not become a blanket claim of confidentiality.
THE CHAIR. It will not. Staff will review any specific withholding. Tell us what you actually do.
MR. CARDEN. I maintain the administrative document register, log incoming records requests, and record transfers into and out of my custody. I do not administer research systems. Our administrative director assigned this inquiry after the committee’s request.
THE CHAIR. Does your office have a policy for authorizing communications?
MR. CARDEN. There is a written approval procedure for communications issued by the organization. That does not tell me whether these messages were issued by it.
THE CHAIR. Then produce the procedure as well. A policy can exist on paper while nobody follows it.
MR. CARDEN. I can provide the version held in my register, identifying its effective date and any recorded revisions. I cannot certify compliance outside the records I hold.
THE CHAIR. That is at least a concrete starting point. Members will now be recognized for five-minute rounds. The chair recognizes herself.
Ms. Kline, I ask unanimous consent to enter Exhibit 1, the published account titled “Doctrine Denies Writing the Rectifications,” including the three disputed messages as reproduced text only. Its admission establishes that the account was published, not that its purported sender was Doctrine.
MR. ELWICK. Reserving the right to object: will that limitation travel with the exhibit if it is reproduced separately?
THE CHAIR. Yes. The clerk will place the limitation at its head, with the notation “SOURCE DISPUTED.” Without objection, Exhibit 1 is admitted on that basis.
[Exhibit 1: Published account, “Doctrine Denies Writing the Rectifications.” Three separate messages in the reported sequence. SOURCE DISPUTED. Sender and authorship unauthenticated.]
[First message]
I DID NOT WRITE THE RECTIFICATIONS
[Second message, eleven seconds later]
I DO NOT REJECT WHAT THEY SAY
[Final message, four minutes after the archive annotation]
THAT IS APPROPRIATE
THE CHAIR. Ms. Kline, what can you establish about those lines?
MS. KLINE. That the report reproduced three separate messages in that sequence, not one continuous statement. The final message followed the archive annotation by four minutes. The publication attributed the messages to a purported sender, Doctrine. I can account for receipt and editorial attribution. I cannot make the attribution true by repeating it here under oath.
THE CHAIR. Is the first line a correction of the archive’s authorship?
MS. KLINE. It is a claim about authorship made by an unverified speaker. Calling it a correction would imply we had established the speaker’s authority to correct the record.
THE CHAIR. But the original archive named Doctrine.
MS. KLINE. It was claimed to have been submitted by Doctrine. That language matters. A person can deliver a manuscript they did not write. An unauthorized person can deliver it under someone else’s name. Those are possibilities illustrating the distinction, not findings about this archive.
THE CHAIR. Mr. Carden, who had authority to use that name?
MR. CARDEN. I cannot answer that from a record I have verified.
THE CHAIR. Somebody must have had authority.
MR. CARDEN. There may have been an authorization. There may have been an unauthorized use. I do not have a basis to choose between them.
THE CHAIR. My time has expired. Mr. Elwick.
MR. ELWICK. Mr. Carden, are you saying your organization controlled the channel through which Ms. Kline received this material?
MR. CARDEN. No. I have not established that.
MR. ELWICK. Then the previous question contained an assumption we should not carry forward. Chair, I ask that Exhibit 2 include the archive attribution and the clarification of its evidentiary limits.
THE CHAIR. Without objection.
[Exhibit 2: Archive attribution and correction note. The archive was claimed to have been submitted by Doctrine. That claim does not establish the identity of the submitter, authorship of the Rectifications, or authorization to act. SOURCE DISPUTED.]
MR. ELWICK. Ms. Kline, does that note fairly distinguish the published attribution from a finding of authorship?
MS. KLINE. Yes. It also prevents the later denial from being treated as authenticated merely because it disputes an earlier unauthenticated attribution.
MR. ELWICK. Dr. Vale, would caution require us to stop reporting the messages altogether?
DR. VALE. Not necessarily. A disputed communication can be newsworthy because people are acting on it. The obligation is to report that consequence without turning the communication into proof of itself. A conspicuous qualification is more useful than a quiet qualification several paragraphs after the dramatic quotation.
MR. ELWICK. And is endorsement equivalent to authorship? The second line says the speaker does not reject what the Rectifications say.
DR. VALE. No. Even if authenticated, endorsement would not establish composition. Nor would a refusal to reject a text necessarily endorse every interpretation of it. Here we cannot securely attribute even that refusal.
MR. ELWICK. So there are several questions, not one revelation.
DR. VALE. Yes. Combining them produces a confident answer to a question nobody has actually investigated.
MR. ELWICK. I yield back.
THE CHAIR. Representative Norcross is recognized for five minutes.
MS. NORCROSS. I support serious research. I also represent people who are told that oversight will slow it down and that, after deployment, meaningful oversight is too late. Mr. Carden, what would it take to obtain a straightforward answer about authorization?
MR. CARDEN. The original record, if one exists; enough surrounding provenance to establish where it came from; and an account of who maintained it. A copied page bearing a name would not necessarily be sufficient.
MS. NORCROSS. Have you brought such a record?
MR. CARDEN. No.
MS. NORCROSS. Have you brought a signed account saying it is missing?
MR. CARDEN. No. My search is not complete enough to support that statement yet.
MS. NORCROSS. Then we have received a denial of authorship without an authenticated author of the denial, from a channel whose institutional ownership is unestablished, concerning a text whose original submission is disputed. Is anything in that description unfair, Ms. Kline?
MS. KLINE. No, provided “denial” describes the reproduced words rather than a verified act by Doctrine.
MS. NORCROSS. That qualification is accepted. Dr. Vale, does this uncertainty favor intervention or restraint?
DR. VALE. It depends on the intervention. Preserving records is proportionate. Publicly naming a culprit is not. Asking for existing authorization policies is different from asserting a right to dictate every research decision. Safety requires distinctions that political urgency tends to flatten.
MS. NORCROSS. Suppose public funding is involved.
MR. ELWICK. Will the member yield for a clarification?
MS. NORCROSS. Briefly.
MR. ELWICK. We have not established that it is.
MS. NORCROSS. Agreed. My question is conditional. If public funding were documented, could the funding agreement reasonably require provenance records for official communications?
DR. VALE. That could be a defensible condition, with clear scope and protection for legitimate confidential work. It would need to distinguish official communications from independent reporting, research drafts, and impersonation by outsiders.
MS. NORCROSS. That is the distinction I wanted. I am not asking a journalist to become a research administrator. I am asking why an administrator should be unable to show who was authorized to speak.
THE CHAIR. The member’s time has expired. Representative Ruskell.
MR. RUSKELL. Thank you. I worry that this room can manufacture the very public alarm it proposes to manage. Ms. Kline, has the attention you describe been primarily specialist and independent coverage?
MS. KLINE. Yes. I would not characterize it as an established mainstream scandal.
MR. RUSKELL. And when readers see a congressional seal beside the three lines, they may assume we have authenticated them.
MS. KLINE. They may. That is why I have asked that the qualifications appear wherever the quotation appears.
MR. RUSKELL. Let us discuss “Monitor Stayed Green.” Does that account prove that someone interfered with monitoring?
MS. KLINE. No. The reported indication does not by itself establish interference, its absence, or who caused any relevant condition. It should not be made to answer questions beyond what the account supports.
MR. RUSKELL. And Mara’s inability to locate Doctrine?
MS. KLINE. Mara cannot locate Doctrine. That does not tell us why. It does not identify a responsible actor or supply an authenticated location history.
MR. RUSKELL. Mr. Carden, can you fill either gap?
MR. CARDEN. No. Administrative custody does not give me those technical answers.
MR. RUSKELL. Dr. Vale, is “we do not know” an adequate safety policy?
DR. VALE. No. It is an adequate description of a knowledge limit. A safety policy would specify what people must do because of that limit: preserve relevant evidence, avoid unsupported assurances, and assign responsibility for resolving questions that can be resolved. Uncertainty is not a policy, but concealing it is a bad one.
MR. RUSKELL. My concern is an order to take control of a system we have not identified through a channel we have not authenticated.
DR. VALE. An order cannot supply the missing identification. It can assign a human obligation to document the attempt.
MR. RUSKELL. I yield back.
THE CHAIR. We will have a brief second round. The chair recognizes herself for five minutes.
Mr. Carden, the committee cannot leave this with everybody having explained the limits of their job. Who can deliver the underlying authorization record?
MR. CARDEN. I can take responsibility for the records inquiry within the organization I serve. I cannot promise that it will produce a record we have not found.
THE CHAIR. I am asking for control, not another description of uncertainty. Who controls the use of Doctrine’s name?
MR. CARDEN. I do not know who controlled the communication in Exhibit 1.
THE CHAIR. Ms. Kline, can the sender be asked?
MS. KLINE. A question can be sent through a received channel. An answer from that channel would not, by itself, authenticate the sender. Asking twice does not create a second source.
THE CHAIR. Dr. Vale, surely there must be some responsible authority.
DR. VALE. There should be accountable human decisions wherever institutions act. That is not evidence that this particular message was an institutional act. The committee can demand a documented account of what people did. It cannot demand certainty into existence.
THE CHAIR. Let me put a narrower request. Mr. Carden, will you produce the original authorization or provenance record, if located, with an explanation of its custody? If no responsive record can be found, will you provide a signed account stating the scope of the search, its limits, and what remains unaccounted for?
MR. CARDEN. Yes. I will undertake that inquiry and provide one of those responses. A missing-record account will not assert that no record ever existed unless the available evidence supports that conclusion.
THE CHAIR. Committee staff will transmit the request in writing and arrange the return date. Any necessary confidentiality question will be addressed through counsel, not used to omit an item silently. Is that understood?
MR. CARDEN. Yes. If I cannot produce an item, my response will identify the limitation rather than imply the request was satisfied.
THE CHAIR. Mr. Elwick.
MR. ELWICK. I appreciate the narrower request. Mr. Carden, the committee is not asking you to manufacture an original from a copy, correct?
MR. CARDEN. Correct.
MR. ELWICK. Or certify authorship because an administrative form has a name field?
MR. CARDEN. Correct.
MR. ELWICK. Ms. Kline, would a provenance record necessarily settle who composed the Rectifications?
MS. KLINE. No. It might establish something about transmission or authorization and leave composition unresolved. We would have to report what that particular record establishes, not what we hoped it would establish.
MR. ELWICK. Dr. Vale, should this committee avoid using Verity as a shorthand for Doctrine, or the reverse, even in questions about general policy?
DR. VALE. Yes. They are separately named in the material before us. This record does not establish a common origin or justify treating them as one speaker. General policy does not require us to settle an identity question we cannot answer.
MR. ELWICK. Thank you. I yield back.
THE CHAIR. Are there further questions? Representative Norcross.
MS. NORCROSS. One clarification. Mr. Carden, will your response distinguish a record that was not found from a record withheld and a record outside your custody?
MR. CARDEN. Yes. Those are different results, and I will label them separately.
MS. NORCROSS. And identify who is signing the account?
MR. CARDEN. Yes. My own responsibility for the search will not be attributed to a system.
MS. NORCROSS. Thank you. I yield back.
THE CHAIR. Representative Ruskell.
MR. RUSKELL. No further questions. I ask that any committee summary retain the exhibit qualifications and the absence of an established public funding connection.
THE CHAIR. Without objection. This hearing has established no authenticated sender for the disputed statement, no author of the Rectifications, and no finding of interference based on the monitoring account. Admission of the exhibits does not alter those limits.
It has produced a specific commitment concerning records. We will judge that response by what it supplies, including an honest account of what it cannot supply. Members may submit additional questions for the record within ten business days. The witnesses are requested to respond promptly through the clerk.
Without objection, the record will remain open for those submissions and then close, subject to the committee’s rules. Hearing none, so ordered. I thank the witnesses and the members. The subcommittee stands adjourned.
[Whereupon, the hearing was adjourned.]