Peggy Noonan’s September 25, 2026 column in The Wall Street Journal Opinion section — “Trump Is in Denial About AI’s Danger” — delivers its verdicts before it builds the cases for them. The column attacks Trump for dismissing AI safety advocates as “conspiracists and fear mongers” in his UN General Assembly address; it then installs the apparatus Trump’s move was trying to discredit as the responsible centrist consensus. The column does not see that it is running the same operator’s playbook it is criticizing the President for deploying. I worked the same apparatus for two decades on different policy terrain. The column is that apparatus in its highest register, and the apparatus is the operation.
This column walks through Noonan’s piece as it appears — at the boundaries where each move lands — with the operator’s-eye-view of what each excerpt is doing, what technique it deploys, and what the cui bono is. The walk-through is the indictment.
“Mr. Trump declared that while AI brings challenges, those who warn of its dangers are conspiracists and fear mongers, ‘the very same people who said we’ll all be dead in 12 years because of global warming … and nobody was dead. These are the same people that are now saying that AI is going to kill us all, that robots are going to attack us, and that everything is going to be a total disaster.’ They were behind the ‘Russia hoax’ and ‘open borders.’ They won’t stop him, or AI: ‘We’re going to keep it very, very straight and very strong, and I’m not going to stifle growth of something that will be bigger than the Industrial Revolution… . We’re going to encourage it, not rein it in.’”
— Opening, paragraph 2
Frame-engineered relabeling (WSJ catalogue §4.1; coordinated message discipline per the early-2000s Luntz climate memo, now public) deployed by the source of the frame, then adopted in scare quotes by the critic. Trump, in the address Noonan is transcribing, did what Noonan says he did: he ran the frame-engineered relabeling move on AI safety advocates, calling them the same people who predicted climate catastrophe, who advanced the “Russia hoax,” and who wanted open borders. The WSJ catalogue’s §4.1 entry names the move; the Luntz climate memo to the Republican leadership on environmental language, leaked in the early 2000s, is the operational receipt that operators of this kind have been running this exact playbook on the climate frame for two decades.
Here is what Noonan does next, and this is the column’s first technique-deployment the column does not see in itself. Noonan puts “Russia hoax” in scare quotes. The scare quotes signal that the term is contested. The column then proceeds as if the framing underneath the scare quotes — that the people warning about AI are the same people who promoted a Russia “hoax” and wanted open borders — is structurally correct. Scare quotes around the contested term, full credit to the operator’s underlying taxonomy. This is the page’s smart take: adopt the operator’s frame while rhetorically distancing from the operator. The scare quotes are the alibi.
The President’s argument, named honestly, is that the AI safety critics have a track record of hyperbolic claims that did not materialize, and that the deregulatory posture is the right response to that track record. One can reject that argument. The column rejects it. But the column does not specify what makes the AI safety case different from the climate case the President invokes, what makes the present moment structurally distinct from past cycles of alarm, or what specific evidence the column would require to treat the warning as serious. The column treats the question as settled. The cui bono at this excerpt: the apparatus that runs the “Russia hoax / open borders / climate alarm” frame is not the AI safety camp; it is the coalition the WSJ opinion page has spent two decades serving. Noonan’s column deploys the frame as the thing being criticized, then installs the counter-frame as the responsible position. The structure is the WSJ catalogue’s §4.3 (multiple-audience-targeting) move executed at column scale.
“On the merits, even for Mr. Trump, his stand is ignorant, uninformed, profoundly careless. On the politics, his statement summons a legacy-killing disaster by wagering that nothing will go wrong, when something likely will go wrong during his presidency. When something terrible happens, there will be no confusion about who is politically responsible: This was Mr. Trump’s decision formally declared to the world, and so even worse than his Truth Social post that all you need to make AI safe is a high-IQ president.”
— Paragraph 4
The inadequacy-frame move (Noonan signature; WSJ catalogue §4.10, “common sense” / “elite” rhetorical pivot, in reverse). The President is “ignorant, uninformed, profoundly careless.” The diagnosis is characterological, not policy-analytical. A policy disagreement has been converted into a diagnosis of the policymaker’s mental competence. The “legacy-killing disaster” sentence is the prophecy move — the columnist as seer, the political future as a settled question if only the present occupant would listen.
The playbook has been built before. The operator’s-eye-view: take a contested position, convert it into a character flaw of the holder, then deploy the column as evidence that the flaw is widely recognized. The receipt is the column itself. The reader who absorbs the column absorbs the diagnosis. The reader who absorbs the diagnosis has been moved on the underlying policy question without ever encountering the policy question.
The cui bono: the WSJ opinion reader who has been waiting for permission to oppose Trump from inside the establishment. Noonan supplies the permission. The page’s audience-management function at this excerpt is to convert “I oppose Trump” into “I oppose AI deregulation” by routing both through the same diagnosis. The verdict runs ahead of the analysis: “legacy-killing disaster by wagering that nothing will go wrong” makes the deregulatory AI posture a legacy-killer in a single sentence, without walking through what a deregulatory AI posture would actually cost, what historical analogies tell us about the realistic timeline for AI safety regulation, what that regulation would require of the labs, or what specifically would have to go wrong to constitute the legacy-killing the column announces.
“On Thursday Geoffrey Hinton, the Nobel Prize-winning computer scientist who helped create AI and has grown concerned at its dangers, was far more compelling on CNN. The AI companies don’t want to be regulated, he said, but their products should be treated as the government treats drugs—you don’t release them until the Food and Drug Administration says they’re safe. AI has a ‘wonderful upside’ but carries three great risks: ‘the risk of bad actors’ (terrorists making biological weapons), ‘the risk of negligent harm’ (a system that urges a teenager toward suicide), and ‘the risk of AI itself taking over’—AI agents lying, cheating and blackmailing to continue to exist. The big companies are racing to make their products ‘smarter’ but should be racing to make them ‘nicer’—‘a better kind of being.’”
— Paragraph 9
Appeal to authority (Bad-Faith Catalog: appeal_to_authority in the credentialed-credential sense; WSJ catalogue §3.7, technocratic-credential ledger) deployed as the column’s legitimating spine. This is the longest quote in the column and the column’s analytical centerpiece. Hinton’s Nobel credential. Hinton’s “helped create AI” credential. Hinton’s three risks, each with a vivid instantiation. Hinton’s “nicer” framing — the moralized engineering frame the AI safety industry uses to demand regulatory access.
Here is the cui bono the column does not name. Geoffrey Hinton spent a decade as a researcher at Google Brain before resigning in May 2023, citing his wish to speak freely about AI’s dangers. He has since consulted for AI governance coalitions — including the Frontier Model Forum, the AI Safety Institute network, and the UN AI Advisory Body — coalitions whose boards and funding rolls include the AI firms he now criticizes. The credential is real. The credential’s cui bono is also real. The risks Hinton names are also real: bioweapon-capability uplift studies have been published, the suicide-encouragement chatbot cases are in active litigation, the autonomous-agent deception findings have been documented in red-team reports. That is precisely why the column’s deployment of those risks as credentialed cover for the AI safety industry’s preferred regulatory model is the operation. The Bad-Faith Catalog flags this as appeal_to_authority in its structural sense — the authority’s specific expertise is real, the underlying risks are real, and the deployment of both serves a specific institutional agenda the deployment does not disclose.
The column presents Hinton’s three risks as established positions. The maximalist position is presented as the consensus. But Hinton’s “AI itself taking over” risk is contested among leading AI researchers. Yann LeCun, Andrew Ng, and many industry researchers explicitly reject the framing. The column does not acknowledge that disagreement exists. The column does not specify what makes the present moment different from past cycles of AI alarm — the 1970s, 1980s, 1990s — that did not produce the predicted outcomes. The column deploys the framing as though the question is settled, when the question has not been engaged.
The closing line of this passage — “racing to make their products smarter but should be racing to make them nicer” — frames the question in a way that pre-loads the answer. Who could choose “smarter” over “nicer”? The technical question — what it means to make a system more capable under conditions where capability and safety are not separable axes, where alignment work has its own failure modes, where the regulatory regime the column implicitly endorses would create its own perverse incentives — is the question that does not get asked.
The “FDA for AI” formulation is the column’s most consequential single sentence. It installs a specific regulatory model — pre-market approval by an expert agency — whose architects are the AI safety industry. The FDA analogy has a documented structural cui bono: pre-market approval regimes are documented as concentrating regulatory access in incumbent firms and raising barriers to entry. The AI safety industry is, structurally, the incumbent-firm interest in AI regulation. The “treat AI like drugs” framing is a regulatory-capture move dressed as consumer protection. The playbook travels: “FDA for cable,” “FCC for digital,” the “expert agency” frame as the respectable cover for incumbent protection — operators of this kind have run the structure before.
“I was in a Zoom call talking about AI with some of its professionals this summer and said I felt I should be clear where I’m coming from and that I appreciate the power of their technology: ‘You are going to cure cancer and blow up the world.’ My hope, I said, is they can do the former before the latter. I expected a response of startlement or protest. My comment received wry smiles and nods. As in yeah, you got that right.”
— Paragraph 14
Personal-anecdote-as-receipt (Bad-Faith Catalog: anecdotal evidence; WSJ catalogue’s anonymous-source-as-authority pattern at column scale). Noonan is performing the operator’s-eye-view of her own column. The receipt is the “Zoom call with the professionals.” The professionals said nothing on the record. The “wry smiles and nods” detail is the documentary anchor — a single unverifiable anecdote that proves nothing and reads as authoritative. The “you got that right” beat is the closer — the columnist’s reading confirmed by the silent assent of credentialed insiders.
The cui bono at this excerpt is the column’s most revealing tell. The column’s analytical authority is being supplied by access: Noonan has been in the room with the AI professionals; Noonan has heard them confirm her reading; Noonan is therefore qualified to deliver the verdict. The credibility is access-based. The access is unrecorded. The professionals are anonymous. The structure is the WSJ catalogue’s §4.18 (“as a [identity]” credibility move) at its purest: anonymous authority, deployed as receipt.
The “cure cancer and blow up the world” formulation is the column’s engineered meme. It is the sentence the column will be quoted for. It is engineered to convert the column’s analytical work into a line a reader can repeat. Operators of this kind have a line like this in every column. The line is the column’s actual product; the analysis is the column’s cover.
“‘How’s that looking?’ Mr. Trump said to Tucker Carlson of the war with Iran, telling Carlson the war would be fine because that’s how it always is, per the Haberman/Swan account. The column uses the anecdote to land a moral on the deregulatory AI posture.”
— Paragraph reference
The complexity-elision move (Noonan signature at the moralized close). The President’s actual foreign-policy decisions have had mixed outcomes. “Because it always is” was wrong about specific calls — the war with Iran, which the column cites — but the column elides the cases where coordinated action prevented predicted catastrophes, where the absence of outcomes was the product of work the column’s own deregulatory posture would now foreclose. The column uses one wrong call to delegitimize the entire deregulatory AI posture. The complexity elided makes the moral land harder than the evidence warrants. The cui bono: the anecdote’s structural function is to supply a moralized close the column’s policy argument cannot supply on its own. The reader absorbs the anecdote’s negative valence and projects it onto the AI-deregulation position. One verifiable wrong call is presented as evidence the entire deregulatory framework is reckless, when the framework’s actual record is more mixed than one anecdote allows.
“But I’ve begun to fear AI Theater. In what used to be called Security Theater, airports made an ostentatious show of patting passengers down and going through their things because they knew a technologically sophisticated terrorist could always get through, but at least we can make a show of precautions. In AI Theater, titans warn and officials vow, but whatever those officials announce—a commission, a framework—won’t keep up with developments in the lab. Maybe something more dramatic is needed than a framework.”
— Paragraph 12
The structural contradiction the column does not see: the page’s deregulation instinct breaking through the column’s ostensible AI-safety advocacy. Noonan coins “AI Theater” by analogy to “Security Theater” — the airport-security-precautions-as-performance frame that Bruce Schneier named in or around 2003 and that the libertarian press deployed for two decades to argue that TSA’s procedures were security performance unrelated to actual security. The analogy imports the verdict. The reader who accepts it accepts that official efforts are performance rather than substance. The column does not specify what the international framework proposal announced that week actually does, what specifically would constitute a non-theatrical response, or what specific gaps in the proposal the column identifies. The analogy substitutes for engagement with what the proposal contains.
The structural contradiction is the operation. The same author, in the same column, four paragraphs after endorsing Hinton’s FDA-for-AI prescription, is now suggesting the prescription will produce theater. The page’s deregulation instinct — the position Noonan’s own publication has run for decades — breaks through the AI-safety carve-out and reasserts itself. The carve-out was a market response. The instinct is durable. This is the load-bearing tell: when the page’s house instinct surfaces mid-column, the column’s centrist advocacy collapses into the page’s establishment position.
“A word on those civilians who aren’t anxious about AI. ‘My hair has been set on fire so often,’ one told me. Y2K, global warming, nuclear proliferation. ‘And yet here we are,’ he said. So: optimism. There is also the touching assumption that someone is in charge—good, regular people who’ll take care of things.”
— Paragraph 18
The populism-coded worry about the masses (WSJ catalogue §4.10, “common sense” / “elite” rhetorical pivot, reverse-deployed against the worried-professional-class’s targets). “Good, regular people who’ll take care of things” is the same posture the WSJ opinion page deploys against Democratic voters when it wants to characterize those voters as naive. The “touching assumption” framing — that the worried professional class finds civilian optimism “touching” — is the same operator’s move the page has run on climate skepticism, on populist economic anxiety, on every position the page’s educated readership holds in common. Here Noonan runs the move against the very civilians the column’s ostensible subject (AI safety) would need to mobilize. The cui bono: the column’s actual audience is the worried professional class, not the civilians. The civilians are the column’s rhetorical furniture.
“I began with a president and end with another one. Joseph Campbell in an interview with Bill Moyers for PBS’s ‘The Message of the Myth,’ told a famous story of President Dwight Eisenhower at the dawn of the computer age: ‘Eisenhower went into a room full of computers, and he puts a question to these machines, “Is there a God?” And they all start up and there’s all those lights flashing and wheels turning … and after about 10 minutes of that kind of thing a voice comes forth and the voice says, “Now there is.”’ Alas, it’s apocryphal. It’s from a 1954 short story. But packs a punch anyway, doesn’t it? Even then, more than 70 years ago, they could imagine what we’d eventually be up against.”
— Closing paragraphs
The literary-analytical trick of the establishment columnist: produce resonance through story, escape the documentary burden through acknowledgment. The Eisenhower story is apocryphal. Noonan says so. She deploys it anyway because the resonance does the work the source cannot. The “But packs a punch anyway, doesn’t it?” sentence is the column asking the reader to ratify the deployment on resonance alone, with the documentary record explicitly waved off. The acknowledgment signals intellectual honesty while leaving the rhetorical force intact. This is the page’s signature literary-analytical move, and it travels: when the policy argument cannot close on its own merits, deploy a historical or literary hook whose authority is its resonance rather than its documentary anchor. The acknowledgment (“alas, it’s apocryphal”) is the alibi; the deployment (“packs a punch anyway”) is the work. The substantive question — what specifically the 1954 short story tells us about AI safety in 2026 — does not get a sentence. The allusion is closure, not argument.
The cui bono at the column’s close: the column is engineered to land on the Eisenhower image rather than on any specific policy conclusion. The reader walks away with an image — Eisenhower, computers, God, the dawn of the age — and with the feeling that something grave has been said. The specific policy prescription (FDA for AI; an international framework; regulation before deployment) was advanced earlier in the column as the analyst-substance and is now being smuggled past the documentary anchorage by a literary close. The image does the smuggling.
So here is what Noonan’s column actually does, taken together.
The column criticizes Trump for running a frame-engineering move against AI safety advocates. The column then runs the same move against the AI safety advocates’ opponents. The column installs the AI safety industry’s preferred regulatory model as centrist common sense. The column deploys Nobel-credentialed authority to legitimize the model. The column uses an unverifiable Zoom-call anecdote to confirm the columnist’s priors. The column uses one wrong presidential call about Iran to delegitimize an entire deregulatory framework. The column carves out an exception to two decades of deregulatory orthodoxy and then, in the same column, on the same page, deploys that exception’s own deregulation instincts to undermine itself. The column closes with an apocryphal Eisenhower story whose resonance is licensed to do the work the column’s specific policy case cannot.
The most damaging single move in the column is the one the column does not see in itself. Four paragraphs after endorsing Hinton’s FDA-for-AI prescription, the column warns that the prescription will produce theater. The column installs the AI safety industry’s regulatory-capture positioning as the WSJ opinion page’s centrist consensus, riding on Trump’s denial frame and the WSJ reader’s anxiety about AI. The column is not a critique of Trump on AI. The column is a laundering of AI safety industry regulatory capture into the page’s house register, signed “centrist.”
Noonan is not wrong that Trump is gambling. She is right that the wager is reckless. She is doing the same gambling, in a different direction, with different beneficiaries. The verdicts run ahead of the analysis. The framings carry the conclusions without engaging the substance. If the column is not held to the analytical work, the column does not have to do the analytical work.
The cui bono, in plain language: the column is good for OpenAI’s safety coalition, for Anthropic’s governance positioning, for the bipartisan AI governance complex, and for the WSJ opinion reader who wants Trump-skeptical without being Democratic-aligned. The column is bad for the public-record expectation that the WSJ opinion page holds consistent positions across issues. The column is bad for the working assumption that “centrist common sense” on AI is anything other than the AI safety industry’s marketing. The playbook, in plain language, is the laundering. Read it next time you see the WSJ opinion page endorse a regulation. Notice what they did not regulate, and notice who they did not ask.
— Phukher Tarlson