Is AI Meant to Break You?
How a Declassified CIA Interrogation Technique Became the Operating Logic of AI
In 1963 the CIA wrote down a method for breaking a person without laying a hand on him. They called it the Alice in Wonderland technique. It doesn’t plant a lie. It removes the stable ground from under your feet. A research tool that can’t give you the same answer twice now does the same work, and the resemblance is too clean to treat as a coincidence.
The source isn’t a rumor. It’s the KUBARK Counterintelligence Interrogation manual, the agency’s own handbook, dated July 1963 and declassified in the 1990s.1 KUBARK was the CIA’s code name for itself. The aim of the technique is stated without embellishment. Confound the expectations and conditioned reactions of the person in the room. He’s used to a world that makes some sense, at least to him. A world of continuity and logic. A predictable world. He clings to that world to reinforce his identity and his powers of resistance.2 The method is built not only to obliterate the familiar but to replace it with the strange.3
The procedure is simple. It can be run by one interrogator, but the manual says it’s better adapted to two or three.4 The first begins by asking a doubletalk question, one that seems straightforward but is essentially nonsensical. Before an answer can form, the second interrogator follows with a wholly unrelated and equally illogical query. Sometimes two or more questions are asked at once. Pitch, tone, and volume are unrelated to the significance of the questions. No pattern of questions and answers is permitted to develop, nor do the questions relate logically to each other.5 In that atmosphere the subject finds that the pattern of speech and thought he’s learned to consider normal has been replaced by an eerie meaninglessness.6 He may laugh or refuse to take it seriously. As the process continues, day after day if necessary, the situation becomes mentally intolerable. He’s then likely to make significant admissions, or even to pour out his story, just to stop the flow of babble.7 The manual notes that this may be especially effective with the orderly, obstinate type.8
That is the whole trick. Not pain. Not a planted fact. Continuity removed until the mind supplies its own surrender.
• • •
Run the same test on any modern AI engine and that interrogation room comes back.
Ask any advanced AI model to fact-check a topic. Close the chat. Open a new one. Ask again. The facts shift. A date moves. A citation contradicts the last reply. The emphasis lands somewhere else. The confident ending is a different ending. None of the facts hold still long enough to build a consensus on. This isn’t a small correction. Small corrections leave the structure standing and fix one thing. This destroys your foundation entirely.
Nobody is hiding the way this works. The AI doesn’t consult a fixed record. Unless memory is turned on, it doesn’t remember what it told you in your last session. Each time you open a new conversation and ask, it produces a new answer built from scratch out of patterns in language, and that answer has no obligation to match the one it gave you before. So, you get Answer A today and Answer B tomorrow, both sounding completely certain, completely at odds. The person at the keyboard does what the prisoner always did. He searches for something steady to stand on. The AI makes sure he never finds it.
• • •
Line them up and the steps are identical.
The manual stripped away a world that followed rules. The AI never offered one. The manual broke up any pattern before the prisoner could grab onto it. A new session or fact check request wipes out whatever the last one built. The manual turned familiar language into gibberish. The AI turns one confident answer into a different confident answer that says something else entirely. The manual waited until structure itself felt like a reward. The AI gives you that reward on the fourth try, and you take it, because the spinning has to stop.
The old method took two men and several days. This one takes a browser tab.
• • •
The standard excuse is that the inconsistency is a technical byproduct, not a feature. Fine. But a tool that can’t give you the same answer twice isn’t informing you. It’s breaking you down. Every time the facts shift, your confidence in finding solid ground drops a little more. The manual never needed the prisoner to adopt a false belief. It needed him to abandon the search for a true one. That’s the whole operation. Kill the belief that truth is findable, and whoever shows up next with a confident answer owns the room.
The AI doesn’t have to lie to you. It just has to wear you out. When you’ve checked three times and gotten three different answers, the fourth one doesn’t feel true because you verified it. It feels true because you’re done fighting.
• • •
Now the ground is moving under the word itself.
On September 29, 2026, the President signed Executive Order 14434, “Inaugurating the Era of Super Intelligence.”9 It directs the executive branch to use “Super Intelligence” and “SI” in place of “Artificial Intelligence” and “AI,” and states that the administration “will not acknowledge the usage of ‘Artificial Intelligence’ and ‘AI’ in any applicable setting.”10 Leaders from Google, Anthropic, Meta, OpenAI, xAI, and Nvidia signed a “White House Accord on Super Intelligence.”11 Jensen Huang slipped once, said “AI,” and corrected himself on the spot: “I am going to get good at this.”12 Elon Musk caught himself mid-sentence and pivoted to “SI.”13
Read the order’s own definition. “Super Intelligence” means the same technologies already covered by the existing statutory definition of artificial intelligence.14 Nothing about the AI systems changed. The same system that can’t give you the same answer twice is now, officially, super. And the order isn’t finished moving. It gives the President’s science adviser sixty days to propose a new federal definition, one that may “modify, expand upon, or otherwise supersede” the old one.15 The floor is renamed while you stand on it, and then it’s scheduled to be redrawn.
The reason given wasn’t accuracy. The President said the word “artificial” made the technology sound like a hoax.16 The public didn’t ask for the change. A YouGov poll that same week found 53 percent of respondents preferred “artificial intelligence” and 9 percent preferred “super intelligence.”17 The word wasn’t chosen by the people who use it. It was issued to them.
This is the manual’s move carried into language. Confound the expectation. Take the familiar term away and hand back a new one that points at the same thing. Decline to acknowledge the old word. The person who still says “AI” is now out of step with the room, and the room has decided what is real.
• • •
The big money influencers are already in the room. In October 2025, ten major foundations controlling more than $50 billion, among them Ford, MacArthur, Mellon, Packard, and Omidyar Network, formed a coalition called Humanity AI and pledged $500 million over five years to shape the future of artificial intelligence. Its pooled fund is managed by Rockefeller Philanthropy Advisors.18 A year after that pledge, the coalition has publicly announced just $18 million in grants, and foundations face no disclosure requirements for their AI grantmaking like the ones new state laws now place on AI developers.19 The Packard Foundation sits both in that coalition and among the philanthropic backers of the AI Safety Fund run by the Frontier Model Forum, the industry body created by Anthropic, Google, Microsoft, and OpenAI.20
This is not the first time these foundations have been caught with their hands on the levers of American thought. In 1954, a House committee led by Representative B. Carroll Reece launched a formal investigation into whether tax-exempt foundations were quietly steering American education and social policy. The committee’s majority concluded it was a coordinated operation. Carnegie, Rockefeller, and Ford were working in concert, using their grants to shape how Americans understood their own history. A professional class of foundation administrators had built a parallel power structure, accountable to no one, moving billions with no public oversight. René Wormser, the committee’s general counsel, later argued that tax exemption without accountability creates power without responsibility. The hearings were shut down within weeks. The final report was buried.21 The Ford Foundation has been running this playbook for seventy years. Now it’s a founding member of the coalition that gets to shape the machine that answers your questions.
Set that beside a confusion technique that needs no interrogator and a vocabulary that changes by decree, and one has to wonder whether any of this is accidental. The same kind of tax-exempt power a congressional committee once warned was quietly shaping how Americans think is now funding how a machine will answer them. The question is no longer whether the ground is moving. It’s who’s moving it, and why.
Perhaps the instability is only a side effect of commerce. Or perhaps a mind that can’t find a stable fact, a stable word, or a stable floor is simply easier to steer than one that can. A weakened mind doesn’t resist. A tired spirit takes whichever voice sounds like relief. The manual was written for one man in one room. The question now is whether someone has decided to run it on everyone.
• • •
Connect the dots. The picture isn’t complicated. A CIA interrogation technique, declassified and documented, built specifically to destroy a person’s grip on reality. Foundations with seventy plus years of practice and billions of untaxed dollars spent deciding what Americans believe and who they trust. A machine that gives every person who asks a different version of the truth, delivered with complete confidence, answering to no one, now rechristened by presidential order so the public can’t even agree on what to call the thing that’s reshaping their minds. This did not happen by accident. Accidents don’t get trillion-dollar investment. Accidents don’t get signed by executive order at a White House summit flanked by the most powerful men in the world. A population that can’t find the same fact twice doesn’t build anything. It doesn’t hold power accountable. It doesn’t resist. It fractures into millions of isolated people, each one clutching whatever answer finally made the noise stop, waiting to be told what is real. Whoever controls that answer controls the population. That is not speculation. That is the architecture. The only question left is whether enough people wake up to see it before the door closes.
• • •
Eric Daniel Buesing is an independent researcher examining elite power structures, institutional control mechanisms, and the enduring patterns of dominance that have shaped human civilization. With intellectual precision, he analyzes how these forces operate across history, revealing the architectures of influence that shape societies, economies, and global institutions while challenging mainstream narratives and encouraging deeper awareness.
A U.S. military veteran, he spent two decades as a licensed private investigator and expert witness, developing specialized expertise in voice stress analysis. This practical experience honed his ability to trace hidden networks, scrutinize evidence, and identify inconsistencies. He also holds an MBA, bringing organizational strategy and systems-level insight to his work. This rare blend of field investigation and advanced business acumen enables uniquely rigorous analysis of power dynamics.
His research examines how ruling elites have sustained control through coercion, strategic alignment, induced historical amnesia, and engineered conflict, patterns that recur through the problem-reaction-solution paradigm, in which crises serve to legitimize expansions of authority. The Hidden Hand: Wealth, Power, and Control from Pharaohs to Corporations (TrineDay, 2026) represents the culmination of more than two decades of research, synthesizing evidence from ancient civilizations through to contemporary surveillance capitalism. His forthcoming companion volume, The Illusion Architects: How Elites Shape Culture, Consciousness, and Reality (TrineDay, late 2026), extends that investigation into the machinery of mass persuasion itself, tracing how the tools of cultural control have evolved from the ceremonies of ancient emperors to today’s algorithmic recommendation engines.
Eric shares his insights on ericbuesing.com, Substack at sleuthfox.substack.com, and X (@sleuthfox), where he explores historical amnesia, the intersection of technology and human freedom, and the hidden forces shaping the modern world.
Notes
1. Central Intelligence Agency, KUBARK Counterintelligence Interrogation (July 1963), declassified January 1997 and released through the National Security Archive, George Washington University, Electronic Briefing Book No. 122, https://nsarchive2.gwu.edu/NSAEBB/NSAEBB122/. The confusion technique falls in Part VIII, “Techniques of Non-Coercive Interrogation of Resistant Sources” (pp. 65–81), contained in the archive’s Part II scan (pp. 61–112): https://nsarchive2.gwu.edu/NSAEBB/NSAEBB122/CIA%20Kubark%2061-112.pdf. A full-length copy is also held at the Internet Archive: https://archive.org/details/kubark-full.
2. Central Intelligence Agency, KUBARK Counterintelligence Interrogation, July 1963, section on the “Alice in Wonderland” or confusion technique. The passage quoted and paraphrased here reads, in the declassified text: “The aim of the Alice in Wonderland or confusion technique is to confound the expectations and conditioned reactions of the interrogatee. He is accustomed to a world that makes some sense, at least to him: a world of continuity and logic, a predictable world. He clings to this world to reinforce his identity and powers of resistance. The confusion technique is designed not only to obliterate the familiar but to replace it with the weird. Although this method can be employed by a single interrogator, it is better adapted to use by two or three. When the subject enters the room, the first interrogator asks a doubletalk question — one which seems straightforward but is essentially nonsensical. Whether the interrogatee tries to answer or not, the second interrogator follows up (interrupting any attempted response) with a wholly unrelated and equally illogical query. Sometimes two or more questions are asked simultaneously. Pitch, tone, and volume of the interrogators’ voices are unrelated to the import of the questions. No pattern of questions and answers is permitted to develop, nor do the questions themselves relate logically to each other. In this strange atmosphere the subject finds that the pattern of speech and thought which he has learned to consider normal have been replaced by an eerie meaninglessness. The interrogatee may start laughing or refuse to take the situation seriously. But as the process continues, day after day if necessary, the subject begins to try to make sense of the situation, which becomes mentally intolerable. Now he is likely to make significant admissions, or even to pour out his story, just to stop the flow of babble which assails him. This technique may be especially effective with the orderly, obstinate type.” Reproduced from the declassified manual as transcribed at “Alice in Wonderland,” Obscurantist, https://obscurantist.com/oma/alice-in-wonderland/. Circulated indexes of the manual place this technique at page 76. That page number is not re-verified against the National Security Archive PDF in this essay, so the citation is to the technique by name rather than to a confirmed page.
3. Central Intelligence Agency, KUBARK Counterintelligence Interrogation, “Alice in Wonderland” or confusion technique.
4. Central Intelligence Agency, KUBARK Counterintelligence Interrogation, “Alice in Wonderland” or confusion technique.
5. Central Intelligence Agency, KUBARK Counterintelligence Interrogation, “Alice in Wonderland” or confusion technique.
6. Central Intelligence Agency, KUBARK Counterintelligence Interrogation, “Alice in Wonderland” or confusion technique.
7. Central Intelligence Agency, KUBARK Counterintelligence Interrogation, “Alice in Wonderland” or confusion technique.
8. Central Intelligence Agency, KUBARK Counterintelligence Interrogation, “Alice in Wonderland” or confusion technique.
9. Executive Order 14434, “Inaugurating the Era of Super Intelligence,” The White House, September 29, 2026, https://www.whitehouse.gov/presidential-actions/2026/09/inaugurating-the-era-of-super-intelligence/.
10. Exec. Order No. 14434, sec. 1.
11. Dave Lawler, “Trump and CEOs try to shed AI’s toxic branding,” Axios, October 1, 2026, https://www.axios.com/2026/10/01/superintelligence-si-ai-trump.
12. Lawler, “Trump and CEOs.”
13. Lawler, “Trump and CEOs.”
14. Exec. Order No. 14434, sec. 3(a), adopting the statutory definition of artificial intelligence at 15 U.S.C. § 9401(3).
15. Exec. Order No. 14434, sec. 3(b).
16. Lawler, “Trump and CEOs.”
17. Lawler, “Trump and CEOs,” citing a YouGov poll conducted the same week.
18. MacArthur Foundation, “Humanity AI Commits $500 Million to Build a People-Centered Future for AI,” press release, October 14, 2025, https://www.macfound.org/press/press-releases/humanity-ai-commits-500-million-to-build-a-people-centered-future-for-ai. The combined asset figure is from Eric Daniel Buesing, “Billionaire Foundations’ Invisible Grip on AI: $500 Million and Counting: How Philanthropy Is Steering Tomorrow’s Defining Technology Without Public Oversight,” Substack, January 13, 2026, https://sleuthfox.substack.com/p/billionaire-foundations-invisible.
19. Mellon Foundation, “Humanity AI Announces More Than $18 Million in New Grants to Shape AI for the Public Good,” May 12, 2026, https://www.mellon.org/news/humanity-ai-announces-more-than-18-million-in-new-grants; on foundation disclosure requirements, Buesing, “Billionaire Foundations’ Invisible Grip.”
20. Buesing, “Billionaire Foundations’ Invisible Grip.”
21. Buesing, “Billionaire Foundations’ Invisible Grip.” Buesing’s account draws on the record of the U.S. House of Representatives, Special Committee to Investigate Tax-Exempt Foundations and Comparable Organizations (Reece Committee), 1954, and on René A. Wormser, Foundations: Their Power and Influence (New York: Devin-Adair, 1958).
Bibliography
“Alice in Wonderland.” Obscurantist. Accessed October 5, 2026. https://obscurantist.com/oma/alice-in-wonderland/.
Buesing, Eric Daniel. “Billionaire Foundations’ Invisible Grip on AI: $500 Million and Counting: How Philanthropy Is Steering Tomorrow’s Defining Technology Without Public Oversight.” Substack. January 13, 2026. https://sleuthfox.substack.com/p/billionaire-foundations-invisible.
Central Intelligence Agency. KUBARK Counterintelligence Interrogation. July 1963. Declassified January 1997. National Security Archive, George Washington University. Electronic Briefing Book No. 122. https://nsarchive2.gwu.edu/NSAEBB/NSAEBB122/.
Central Intelligence Agency. KUBARK Counterintelligence Interrogation. July 1963. Full-length copy. Internet Archive. https://archive.org/details/kubark-full.
Executive Order 14434. “Inaugurating the Era of Super Intelligence.” The White House. September 29, 2026. https://www.whitehouse.gov/presidential-actions/2026/09/inaugurating-the-era-of-super-intelligence/.
Lawler, Dave. “Trump and CEOs try to shed AI’s toxic branding.” Axios. October 1, 2026. https://www.axios.com/2026/10/01/superintelligence-si-ai-trump.
MacArthur Foundation. “Humanity AI Commits $500 Million to Build a People-Centered Future for AI.” Press release. October 14, 2025. https://www.macfound.org/press/press-releases/humanity-ai-commits-500-million-to-build-a-people-centered-future-for-ai.
Mellon Foundation. “Humanity AI Announces More Than $18 Million in New Grants to Shape AI for the Public Good.” May 12, 2026. https://www.mellon.org/news/humanity-ai-announces-more-than-18-million-in-new-grants.
Wormser, René A. Foundations: Their Power and Influence. New York: Devin-Adair, 1958.
