AIPolitics 17: Anthropic Wrote the Papal Encyclical
When a frontier lab outsources its moral authority to the Vatican, what does that tell you about every other available institution?
A Catholic Pope issued an encyclical on artificial intelligence yesterday. The encyclical runs forty-two thousand words. Pope Leo XIV describes AI systems as “cultivated” rather than “built.” The Pope notes that the internal representations of current models “remain, at present, unknown.” Then, the Pope calls for the disarmament of AI.
It is an amazing document in so many ways.
That said, both formulations track Anthropic’s interpretability vocabulary closely. The “cultivated, not built” framing has been part of Chris Olah’s interpretability research at Anthropic for years, the central trope in a research program arguing that AI systems are grown rather than engineered. AI systems emerge from training. Their inner workings have to be reverse-engineered after the fact, like a biological organism rather than read off a schematic like a bridge. The framing isn’t unique to Anthropic. Variants of “grown not built” have been knocking around AI-alignment vocabulary since the late 2010s. Anthropic has owned the institutional development of the frame the way McKinsey owns “agility” or the ACLU owns “civil liberties.” The Pope’s encyclical adopts the Anthropic-flavored version, not the early-LessWrong version or the OpenAI-emergence-of-capabilities version. The “internal representations remain unknown” framing is also distinctively Olah’s, the standing summary of mechanistic interpretability as a research program.
Chris Olah stood next to Pope Leo XIV at the encyclical’s unveiling. He gave remarks. In those remarks he cited the Pope citing Anthropic. A recursive loop closed right in front of you in real time.
The Washington Post ran the story under the headline “Anthropic aligns with Vatican over White House as Pope Leo stokes AI fears.” A major US institutional newspaper put it in its own headline. Anthropic chose the Pope.
A point worth making explicit before going further: nobody has produced documentary evidence that Anthropic personnel drafted lines of the encyclical, and I don’t have any either. But let’s use our eyes for a minute, eh?
What’s documented is that Olah was invited, Olah appeared, Olah’s research vocabulary is recognizable inside the document, the WaPo coverage reads the Pope’s framing as “echoing” Anthropic’s, and Anthropic’s policy reach into Rome was substantial enough that several other frontier labs were also lobbying for influence over the same document. That’s the evidentiary base. Strong enough to support the analytical claims in this piece. Not strong enough to support a literal claim of authorship.
So here is the question. What does it mean when a frontier AI lab outsources its moral authority to the Vatican? And what does the choice tell us about every other institution that was not available to do the job?
What Anthropic Actually Did
Politico reported the encyclical as the culmination of a multi-month “Vatican lobbying season.” Meta, Google, Amazon all lobbied too. They all wanted access, and more importantly, endorsement. They all wanted their framing of AI policy embedded in the document the Pope was about to release. Anthropic won.
The competitive dimension of that win is what matters, with one important caveat. The Anthropic-Vatican alignment was not an accidental match between two unusual institutions. Anthropic outmaneuvered the others by playing a different game. The caveat is that the others may not have been seriously trying to win this particular contest. A Vatican imprimatur on AI ethics is a regulatory and reputational bind on commercial deployment; Meta, Google, and Amazon may have shown up at the Vatican to manage the downside of an encyclical rather than to capture its framing. Anthropic’s commercial positioning rewards being aligned with regulatory and moral-authority frames; the others’ commercial positioning doesn’t. The contest was real. The prize was differently valued by the contestants. Anthropic was structurally positioned to compete hardest for it.
That structural advantage is the story. Meta and Google were probably arguing for the things Meta and Google argue for: speed, openness, the inevitability of progress, the danger of falling behind China. Anthropic argued for a vocabulary. Anthropic argued for “cultivated.” Anthropic argued for the language of interpretability. Anthropic argued that AI is the kind of thing whose internal workings remain “unknown.”
And the vocabulary won.
Frame-control as policy strategy is not new. Every lobbying outfit in Washington has a glossary of preferred terms, and the bills that eventually pass tend to use whichever side’s glossary lands first. The novelty here is the venue. Anthropic didn’t lobby for the vocabulary in a House subcommittee or an SEC working group. Anthropic lobbied for the vocabulary in Catholic social teaching.
I’m not Catholic. So I’ve been doing some reading of the document itself, what an encyclical is, as well as all of the coverage. I have learned…a lot.
An encyclical is, structurally, the most durable framing document a Christian moral institution can issue. The encyclical anchors everything Catholic theologians, hospital systems, universities, and lay readers will say about AI for the next half-century. Once “cultivated, not built” is in the encyclical, the phrase is in the vocabulary of every Catholic bioethics conference, every Catholic university’s AI policy committee, every Catholic-affiliated hospital that has to decide whether to deploy a diagnostic model. The framing has institutional substrate.
A reasonable challenge to this move: does encyclical vocabulary actually transmit into legislative debate? The historical record is mixed at best. Laudato Si’ (2015) put strong climate-justice vocabulary into Catholic social teaching; eleven years later, US climate legislation doesn’t use that vocabulary. Centesimus Annus (1991) on capitalism didn’t observably shape Clinton-era economic policy. The transmission channel from Vatican to Capitol Hill is thin, at least lately.
What encyclical vocabulary does reliably shape is Catholic moral discourse: bishops’ conferences, Catholic universities, Catholic hospital systems, Catholic-affiliated charities, Catholic media. The vocabulary is downstream of one of the world’s largest single religious institutions, with all the institutional reach that implies. It doesn’t directly write US AI legislation. So, the Anthropic win is large, but it’s not unbounded.
The frontier lab that got there first is the frontier lab whose intellectual program now anchors the moral discussion, within the institutional reach of Catholic discourse.
Anthropic is lobbying for the vocabulary much of the moral debate will be conducted in, rather than for a specific bill. The game is a different game, and the company is winning it.
The Olah Move
Look at what Chris Olah actually said at the Vatican.
“We need moral voices that the incentives cannot bend.” A co-founder of a frontier AI lab stood at the Vatican and publicly stated that frontier AI labs need external moral authority because their incentive structures cannot be trusted to produce moral outcomes alone.
Pair that with what Olah said immediately before: “Every frontier AI lab, including Anthropic, operates inside a set of incentives and constraints that can sometimes conflict with doing the right thing.” Olah was preemptively absorbing the regulatory-capture critique. The accelerationist read of what Anthropic just did is that they are playing a regulatory-capture game by making the church their moat. Mike Solana has made that argument repeatedly. Olah answered the critique before Solana could deliver it from the press release. He acknowledged the incentive problem. He named the structural constraint. He then concluded that the response to the incentive problem is external moral authority, and located that authority in the Vatican.
So, Olah is publicly saying: my industry is not capable of self-regulation. The way to address the gap is to import a moral authority that is older, more institutionally stable, and uncorrupted by the venture-capital incentive stack. The Pope qualifies. The Pope is, by job description, the institutional voice that the incentives cannot bend.
I mean, that is a “wow” isn’t it? A frontier lab founder declaring on the record that the available secular institutional voices are insufficient. Congress, the SEC, the Federal Trade Commission, the National Institute of Standards and Technology, any state AG, any federal AI agency that might one day exist: all of them, insufficient.
The Pope is what’s left.
The Olah remarks get deeply strange at the third question though. Olah laid out three “questions for discernment” the encyclical raises. One was the condition of the global poor in an AI-mediated economy. One was the use of moral imagination to envision flourishing. The third was the nature of AI models themselves. On that third question, Olah told the Vatican audience: “We find structures that mirror results from human neuroscience. We find evidence of introspection. We find internal states that functionally mirror joy, satisfaction, fear, grief, and unease.”
Read the language carefully. Every clause is hedged. “Structures that mirror” is not “structures that are.” “Evidence of introspection” is not “introspection.” “Internal states that functionally mirror” carries a heavy load on “functionally,” which is an interpretability researcher’s way of saying “these neural patterns produce behavior consistent with these emotional categories, without claiming the model experiences them.” Olah said the technically careful thing.
The problem is that the Vatican audience and the encyclical readers downstream do not parse “functionally mirror” with the technical care a researcher does. The hedge survives in the lab. The unhedged version survives in the encyclical. The Pope’s encyclical treats Olah’s findings as the framework. Olah’s research is no longer one position in an open scientific debate. The Pope’s moral analysis is built on top of it.
The language transfer is recursive, partial, and out in the open.
The Other (Related) Headlines We Should Tie Together
Same week as the encyclical, the European Central Bank summoned bank leaders to address Anthropic’s Claude Mythos Preview as a systemic financial risk. The ECB framing is that the model represents a sufficient threat to European financial stability that US banks with early access should be required to share with European rivals. Forced information-sharing as a financial-stability measure, applied to an AI model preview.
Same month, the Pentagon was blacklisting Anthropic over export-control disputes. The Pentagon-Anthropic case is now in DC Circuit briefing. The Department of Defense’s position is that Anthropic’s safety-encoded constraints make the company an unreliable defense supplier. Three institutions in three weeks, each with a different posture: the Catholic Church embraces, European central banking demands forced disclosure, the Pentagon blacklists.
Three institutions. Three frames. One firm.
Multi-institutional simultaneous response to a major tech firm isn’t, by itself, unusual. Microsoft was hit with simultaneous DOJ, EU, and SEC actions in 2000. Google has been the target of US, EU, and French simultaneous regulatory action multiple times since 2017. Meta gets multi-jurisdictional pressure on a near-quarterly basis. What makes the Anthropic case distinctive isn’t the count of institutions. It’s the spread of valences. One religious institution embraces; one central bank demands forced disclosure; one defense ministry blacklists. Three opposite institutional reactions in the same news cycle.
The structural fact gets lost in the encyclical-day coverage. Anthropic is now consequential enough as an actor that its posture is policy. The firm’s commercial positioning generates an institutional response in three different sectors at once, the same week. The political-economic vocabulary we have for thinking about companies (the framing where “Anthropic is doing X”) is no longer adequate to a moment where major global institutions are simultaneously embracing, demanding access to, and shutting out the same lab on a rolling weekly basis.
The encyclical works on three layers simultaneously. The first is religious. The second is technological. The third, the layer where the encyclical-day coverage missed the point, is institutional realignment. Anthropic occupies the middle position.
(Quick aside for new readers: this is the seventeenth piece in the AI Politics series. The full archive is here. If the analysis is hitting, subscribe and the next one shows up in your inbox.)
The Silicon Valley Response Broke
When the Pope released the encyclical, you might have expected Silicon Valley to close ranks. Tech labs usually do, when a major institutional voice criticizes them. Silicon Valley didn’t close ranks though. The fracture had been visible for months, but the encyclical brought it into sharp focus.
Start with Marc Andreessen. Last November, when Pope Leo addressed the Builders AI Forum with an earlier and shorter version of the same moral argument, Andreessen quote-posted the Pope’s call for moral discernment with a meme: a screenshot from a GQ interview where the interviewer raises her eyebrows at actress Sydney Sweeney about her jeans ad. The implicit equation was that the Pope’s call for tech leaders to cultivate moral discernment was, structurally, woke moralizing dressed up in vestments. Andreessen was running the play accelerationists have been running on the encyclical’s predecessor documents since 2024.
The play didn’t work. The backlash came hard, and much of it came from inside the tech industry itself, including from people of faith who don’t usually have to choose between their religious commitments and their professional environment. By the end of the day, Andreessen had deleted the post.
The deletion was the story. Not that an accelerationist attacked the Pope, but that the attack drew enough institutional return fire from inside the tech industry that the attacker withdrew. The “we’re all rowing in the same direction” tech consensus on AI policy had already cracked six months before the encyclical released. The encyclical is the moment the crack widened into a public split.
Peter Thiel went much further. Thiel publicly named Pope Leo as a potential Antichrist and urged Vice President JD Vance to ignore the Pope’s moral guidance. The political-economic question Thiel’s statement raises: Thiel is Vance’s top donor. Vance is the Catholic Vice President of the United States. The administration’s AI policy track (Pentagon-8, CAISI, the Trump-Xi AI ladder) runs through Vance and through Thiel-aligned funders. Whether Thiel’s statement reflects a coordinated donor-pressure play or just Thiel’s documented apocalyptic theological eccentricity expressed in public is a question of motive that can’t be observed directly. What can be observed is the structural alignment: a Pope-skeptical posture from the funder side of the donor-VP relationship that decides US AI policy, in the same news cycle as the encyclical’s release. Worth watching how Vance responds. Silence and engagement will each tell us something.
Reid Hoffman defended the Pope. Hoffman went on the record backing the encyclical’s stance. As of this writing, he is the only major Silicon Valley figure who has done so publicly. The lonely position.
(An irony in the Hoffman defense: Hoffman is the executive whose “Reid AI” digital twin has delivered more than seventy-five public addresses for him since 2024, trained on twenty-two years of his books, speeches, and podcasts. So the man whose AI clone now handles his public speaking is the man defending the Pope’s claim that AI must not substitute for human moral discernment. The irony sharpens what the defense costs without disqualifying it.)
What you have, then, is not a unified Silicon Valley response to the encyclical. You have a triangle. Andreessen attacked, got pushed back, and retreated. Thiel escalated to theology-level accusation. Hoffman defended. The accelerationist bloc is louder than it is unified, and that public split is now legible to the press.
For the Pope, this was a strategic win on the first day. The encyclical was not just an institutional document; the encyclical was a wedge that visibly fractured the relevant target audience on contact.
The Accelerationist Frame, Stated Plainly
The accelerationist read deserves its strongest version, because the Vatican-Anthropic alignment really does deserve the regulatory-capture critique on the merits.
Andreessen’s Techno-Optimist Manifesto puts the case at maximum strength: “We believe any deceleration of AI will cost lives. Deaths that were preventable by the AI that was prevented from existing is a form of murder.” The frame, in plain language: the moral catastrophe of AI policy is the under-deployment of capable systems that would have saved lives if allowed to ship. Regulation kills people. The Pope, by calling for the disarmament of AI, is morally responsible for the lives the disarmed AI does not save.
The frames don’t merely oppose each other. They invert on what counts as the moral catastrophe. The Pope frames concentrated AI power as catastrophe; Andreessen frames AI restriction as catastrophe. Both invoke human flourishing as the agreed end. The disagreement is about theory of harm: which means is most likely to produce the agreed end, and which kind of harm has to be prevented first. That’s a real disagreement, and in principle a productive one to have empirically.
In practice the disagreement is hard to host inside a single policy debate because both sides treat the other side’s harm as morally negligible. The Pope’s encyclical does not weigh the lives Andreessen says regulation will lose. Andreessen’s manifesto does not weigh the human-dignity costs the Pope says concentrated AI will impose. The disagreement is large enough that the contest of the moment isn’t really empirical. The contest is over which moral arithmetic gets anchored in the institutional vocabulary. The Pope just put his arithmetic in Catholic social teaching. The accelerationists are stuck arguing that the moral arithmetic of Catholic social teaching is itself murderous.
The accelerationists are not going to win that argument.
Mike Solana’s version of the accelerationist read is more tractable. Solana has argued that Anthropic is “running a sophisticated regulatory capture strategy based on fear-mongering,” and that the company benefits commercially from restrictions because its safety-encoded models would otherwise lose market share to less-restricted competitors. On Solana’s read, the Pope is the prop, the moral language is the cover, and the regulatory moat is the play.
Solana has a point. You cannot read Olah’s “moral voices the incentives cannot bend” line without noticing that the speaker has commercial incentives that benefit from a stronger external moral authority being installed. The piece of the argument Solana can’t quite carry, at first pass, is that Olah named the regulatory-capture frame himself, on stage, in his Vatican remarks, and then proceeded with the moral argument anyway. The acknowledgment functions as a kind of inoculation. Whatever you want to say about Anthropic’s incentives, Olah said it first.
The Solana-style counter-read: the preemptive acknowledgment is itself the play. A speaker who acknowledges his industry’s incentive corruption while at the Vatican is doing a sophisticated version of exactly what Solana accuses Anthropic of, with the acknowledgment as armor. Both readings stay on the table. Which one is correct is a question of motive, and motive isn’t observable. What is observable: Olah’s line landed with the audience it was designed for. Solana’s critique now has to clear a higher bar before it lands in the same room.
The accelerationist critique has a principled libertarian cousin that shouldn’t be flattened into it. Adam Thierer at R Street has been writing through the encyclical’s regulatory implications from a more traditional free-market technology-policy posture, and his case is different from Andreessen’s. Thierer’s case is that the regulatory machinery the encyclical is calling for is itself a concentration of power, just one wearing different vestments. The libertarian objection runs on institutional grounds rather than moral ones. The objection is the closest thing to a serious non-accelerationist counter-frame to the encyclical, and it deserves to be in the conversation alongside the Solana critique rather than absorbed into it.
Why Is the Pope Doing What Congress Should Be Doing?
The encyclical itself isn’t really my beat, of course. Catholic social teaching has been doing what Catholic social teaching does for a hundred and thirty-five years, going back to Rerum Novarum on labor in 1891. The Church has a long tradition of issuing major framing documents at the start of new technological eras. Magnifica Humanitas is the AI edition. Cardinal-watchers will be writing about its theological implications for the rest of the decade, and they should be.
The question that keeps surfacing for me, when you sit with the encyclical alongside the rest of the institutional terrain, has no comfortable answer: Why is the Pope the one doing this? Why was the venue for the largest framing document on AI ethics this year a Vatican unveiling instead of, say, a White House report, a Federal AI Commission rulebook, a Senate hearing’s published findings, a National Academy of Sciences report, a PCAST advisory, an OECD AI principles update, or a UN AI Advisory Body framework?
The honest answer is that the secular institutional voices available to do this work are either absent, captured, or insufficient at scale. In an era where the influence of religion has been said to have declined markedly, this is…something!
Congress has been unable to pass federal AI legislation. The most recent attempt, the moratorium provision that would have preempted state AI laws, failed in the Senate this spring, and what’s currently moving on the Hill is a patchwork of sector-specific bills on child safety, deepfakes, and election integrity. The bigger framing legislation has been stuck for two years. No federal AI agency exists. The White House AI Action Plan from last summer set a deregulatory direction but did not produce a regulatory institution capable of the kind of comprehensive framing the encyclical performs.
The SEC has been mostly silent on AI. The FTC has issued guidance but is not framing the moral terms of the debate. The CAISI work is technical assessment rather than moral framework. State-level activity has been considerable: California’s retain-not-replace executive order, New York’s AI Dividend bill. But state action is by definition sub-national, and the framing arena where the encyclical is operating is global.
I can hear the (fair) retort: maybe these secular institutions shouldn’t be issuing 42,300-word moral frameworks. Congress’s job is statutory law, not moral framing. National Academy reports are written for policymakers, not laity. The Catholic Church has a hundred-and-thirty-five-year longitudinal moral tradition that no American legislature can match. Maybe the encyclical filling this niche is exactly what should happen, and the alarm in this piece is misplaced.
The objection is serious. My counter is that the alarm is really about the absence of any secular institutional voice with comparable durability and reach, rather than about Congress writing moral documents. IEEE has standards. OECD has principles. PCAST has reports. NAS has assessments. None of those bodies has the institutional substrate the encyclical does. Not because moral framing has to come from a Pope, but because the secular venues that could have built comparable substrate have not built it. The Pope is filling a vacancy. The vacancy is the issue.
Olah’s “we need moral voices the incentives cannot bend” was a job description. The Pope qualified for the position because nobody else was applying.
Andy Hall’s prescription follows from the same diagnosis, and I’ve written about it before in this series. Hall has been arguing for two years that the move policymakers actually need is to build measurement scaffolding: federal sentinel agencies, sectoral displacement data with the same authority and frequency as employment statistics, public benchmarks of model capabilities and failure modes, a fiscal architecture that can sustain redistribution if displacement runs ahead of reskilling. The Hall prescription is something other than a traditional policy proposal. It’s a request for the substrate that any future policy debate would have to be built on top of. Without the data, you can’t even have the argument.
The Hall prescription is still possible because the political-system window is, in principle, still open. The encyclical didn’t close it. Congress could pass a sentinel-agency bill tomorrow. The administration could stand up an AI measurement infrastructure inside Commerce next month. The Hall prescription does not require an act of God.
The Hall prescription also didn’t happen. In its absence, the framing moved without it. The encyclical now has the vocabulary that any future Hall-style architecture would have to be reconciled with. If a federal AI measurement agency gets stood up two years from now, the conceptual frame it operates under will be one that Anthropic helped author, because the most durable framing document of the previous two years was the one Anthropic helped author. Frame-control compounds. Once the vocabulary is in the bill drafts, the bills that follow have to argue inside it.
The Other Window That Closed Anyway
A note before the close, because the easy reading of this piece is “AI labs are colonizing religious institutions; this is bad; resist.” That easy reading is not quite right.
The encyclical is, in many respects, an amazing, serious moral document. Simon Willison, an independent AI researcher with no Anthropic ties, called Magnifica Humanitas “some of the clearest writing I’ve seen on the ethics of integrating AI into modern society.” TechCrunch read it as a document fundamentally about concentrated power and democratic participation, with AI as the contemporary occasion rather than the subject. The Pope’s call for AI to be “disarmed, freed from logics that turn it into an instrument of domination, exclusion and death” has intellectual content and represents a serious moral position. The encyclical takes a stance, with substance behind it, that the Catholic Church now holds about the technology my profession is increasingly asking my students to use.
The discomfort here runs through the institutional process, not the moral content.
How major framing on a major technology gets done affects what the framing can do later. The venue. The participants. The timeline. Each shapes what the framing can later become. An encyclical can be referenced, taught, cited at the funeral of a programmer killed by a faulty AI system, and invoked at the Vatican when a bishop talks to a tech CEO. The encyclical cannot be amended, repealed, or revised through democratic process. The encyclical has none of the responsiveness of legislation and all of the durability of doctrine. Catholic social teaching is a plausible place for a framing document to live. Catholic social teaching also operates on different rules than any institution capable of binding the firms it discusses.
The frame-control story matters because frame-control reproduces within domains where the moral document carries authority. Once “cultivated, not built” is in Catholic social teaching, the phrase shapes Catholic bioethics conferences, Catholic university policy committees, Catholic-affiliated hospital decisions, Catholic media coverage. Whether the phrase also shapes US legislative debate is the harder question; the historical record on encyclical-to-legislation transmission is thin, as already noted. What is not thin is the reach of Catholic moral discourse itself, and the partial overlap between that discourse and the elite policy class that drafts AI legislation. The vocabulary defines the terrain on which a meaningful chunk of the moral debate happens. Anthropic understood this. The other frontier labs, judging by the lobbying outcomes, did not.
And, sure, Andy Hall’s window comes back into view here. Hall has been arguing that the political-system window for AI policy remains open, that Congress, the agencies, the courts have the institutional capacity to act if they choose to. Hall’s recent writing is more politically confident than the framing in this piece allows. He would, I think, push back on the “narrowing” claim. I have a lot of respect for the argument, as I said above, the political-system window is, in principle, still open.
In practice, the substrate of the debate is being written somewhere else. The frame-defining venues for AI ethics in 2026 are the Vatican, the Davos breakouts, the corporate research labs, the Substack essays of the people who are paid to think about this for a living. Those venues have moved faster than the venues with democratic legitimacy.
A check against overclaim: there are framings other than Anthropic’s still operating in major AI policy debate. The EU AI Act, passed in 2024, runs on vocabulary derived from human-rights and risk-tiering traditions rather than from interpretability research. California’s SB-1047, vetoed in 2024, was written closer to MIRI-style existential-risk framing. The Pentagon-Anthropic dispute itself is being conducted in defense-procurement vocabulary, not in Olah’s. The world is not a vocabulary monopoly. What’s distinctive about the encyclical move is the durability and the religious-discourse reach, not exclusivity of framing.
You can keep the political-system window open through structural commitment. You can’t keep the framing-control window open through institutional default. While Congress was unable to pass legislation on AI, Anthropic was contributing to the language of an encyclical. The frame moved while the institutions slept. Two years from now, if Congress finally does pass framework legislation, at least one corner of the framework will be built on a vocabulary the firm that benefits most has already established.
I’m not arguing this is an injustice! Anthropic competed and won. Meta, Google, Amazon competed and lost. Frame-control works that way. (For the GenXers: “Neener, neener, neener.”)
I’m also not arguing Anthropic the firm executed a unified multi-year strategy to wind up at the Vatican on May 25. The lab’s Long-Term Benefit Trust, its policy team, its co-founders, and its commercial leadership have different incentives and different positions. Olah-at-the-Vatican may be one researcher acting on his interpretability program’s intellectual commitments more than on the firm’s policy strategy. The institutional outcome doesn’t require coordinated firm-level intent. It just requires that one of Anthropic’s principals was the most credible AI-research interlocutor available to a Pope drafting an encyclical on AI. Olah was.
The argument here is narrower than “Anthropic now controls the policy vocabulary.” The argument is that the political-system window Hall describes appears, at this point, narrower than it was a year ago in one specific sense: the most durable framing document on AI ethics released this year was authored partly through Anthropic’s institutional reach. The Pope’s encyclical doesn’t go away. The Pope’s vocabulary doesn’t go away. The recursive language transfer between Anthropic and the Vatican doesn’t go away. The framing that gets used in the moral arguments your sister has with her parish priest about whether to deploy AI tutoring software at the Catholic school is the framing one company helped author.
The political-economic literature on policy framing (Schattschneider, Baumgartner, Jones, the agenda-setting tradition) has been saying for decades that whoever defines the policy image controls the policy outcome. Anthropic just defined the policy image of artificial intelligence in Catholic social teaching, while every secular framing venue with democratic accountability was either deadlocked, dismantled, or unbuilt.
So, Hall’s window is still there. The room behind the window is increasingly full of furniture that the political system did not put there.
Wild times. The writers’ room that came up with this little plot twist, they deserve some accolades, eh?
If this piece sharpened your thinking, share it somewhere, restack it, whatever. Series like this one compound when readers share it. Much appreciated!
If I lose my job, who exactly is taxing me?