Gustavo de Arístegui,
July 17, 2026
I. BRIEF INTRODUCTION
The phrase that states have neither permanent friends nor enemies, only interests, has been repeated a thousand times. Many apply this to geopolitics, but the truth is that it’s not always played out on maps, but rather in consciences. In my opinion, consciences should play the most relevant role in geopolitics, but I’m afraid I can only afford to say that because I only write about geopolitics; I’m not a protagonist. Today is one of those days. Two seemingly incomparable news stories share the same root, and that root is the deepest of all: who decides what a society considers true and who decides what a machine will deem good?
The first news arrived from Washington in the early hours of the morning in Europe. President Trump accused the People’s Republic of China of carrying out the largest theft of election data in history and, during prime time, declassified a set of intelligence documents to prove it. The second hasn’t made any headlines, yet it deserves to: while recent computer engineering graduates face an unemployment rate of nearly 7.5% in the United States, major artificial intelligence labs are vying for philosophers with six-figure salaries. The first news item belongs to the age-old art of political warfare. The second determines which morality will govern the systems that will soon be selecting military targets. Both aim at the same target: human discernment.
Today we dedicate this entire report to these two issues. And we warn from the outset that neither of them lends itself to the aseptic treatment so favored by the mediocre political class of our century, that which confuses prudence with abstention or the paralysis of the deer before the blinding light, and balance or neutrality with cowardly silence.
II. THE MOST IMPORTANT NEWS OF THE LAST 24 HOURS
1. Trump accuses Beijing of the biggest theft of election data in history: the complaint that threatens to carry out the operation it denounces
Facts
President Donald Trump addressed the nation from the East Room of the White House on Thursday, July 16, at 9:00 p.m. Eastern Time (early Friday morning in mainland Spain). He announced the immediate declassification of a substantial volume of intelligence material and asserted that, over several years, beginning with the 2020 election cycle, the People’s Republic of China carried out what is believed to be the largest compromise of election data in history, with the illicit acquisition of 220 million records of American voters. The files reportedly contain names, addresses, phone numbers, voting history, political affiliation, and military status. The White House estimates the timeframe for the data to have been from 2020 to 2023; some reports extend it to 2024. The documents have been published on the White House website.
Trump added that federal agencies first became aware of the voter roll compromise in 2020, when they discovered that the data of tens of millions of voters in eighteen states had been bought, stolen, or hacked by China. He then leveled a second, internal accusation: that members of the intelligence community—the so-called “deep state”—actively suppressed this information and concealed it from both the president and the American people, even removing it from the daily presidential briefing. He also claimed that a Department of Homeland Security analysis identified 278,000 non-citizens registered to vote in federal elections, a figure he speculatively raised to one million. He concluded with a demand to Congress: the passage of the Save America Act, which would mandate photo identification and proof of citizenship to vote.
And here’s what no serious analyst can omit, because it’s the crux of the whole matter: the declassified materials contain no accusation whatsoever of vote swapping, hacking of voting machines, or any election result being altered by foreign interference. Not the 2020 election, nor any other. The accusation is of massive data capture, not manipulation of the vote count. These are two radically different things, and it’s crucial never to confuse them.
The checks and balances are equally verifiable and must be made public. In many states, voter registration lists are public or can be legally obtained: North Carolina publishes them online, and parties, campaigns, researchers, and consultants routinely purchase them. Having a copy does not grant access to the official system, alter records, or allow voting on behalf of anyone. A National Intelligence Council report from October 2020 characterized Chinese activity during that election cycle as “low-level” and limited to exploratory steps. And the declassified 2021 assessment, prepared by officials appointed by Trump himself, concluded that China considered attempting to influence the outcome but decided against it for fear of damaging the bilateral relationship. However, a report from the same organization from January 2020 works in the president’s favor: Russia, China, Iran, and North Korea have the capacity to access and potentially manipulate U.S. election data.
The reaction was immediate. The Chinese embassy in Washington denied everything with its usual formula, invoking the principle of non-interference and asserting that China has never interfered and will never interfere in U.S. presidential elections. The twenty-four Democratic governors issued a joint statement accusing the president of trying to intimidate and silence voters. NBC and CNN did not air the speech. And while Trump was listing his appointments in Washington, Xi Jinping was inaugurating China’s flagship artificial intelligence summit in Shanghai, presenting himself as a global leader responsible for technological governance. The juxtaposition is worth more than a hundred pages of analysis.
Implications
Let’s begin with what is true, because it is true, and because denying it would be as irresponsible as exaggerating it. A new battlefield exists, and this is not a journalistic metaphor. When a political adversary stands firm against an external enemy that is not a democracy, that external enemy—precisely because it is not a democracy—can resort to methods infinitely more expeditious than those a democracy allows itself. It can employ its intelligence and counterintelligence services to massively penetrate the rival country. And not only to manipulate elections, which is the obvious and least interesting aspect. It can also manipulate journalism, opinion leaders, academia, and culture, ultimately leading as many citizens of the opposing nation as possible to distrust their government and institutions. The ultimate goal is for these citizens to end up, through apathy or opposition, serving the interests of the rival power in the short, medium, and long term. This has had a technical name since 1948: political warfare.
George Kennan defined it in the Policy Planning Staff memorandum of April 30 of that year: the use of all means available to a nation, except war, to achieve its national objectives. It is not a modern intuition. It is doctrine, and it has been on the table for eighty years.
In the Chinese case, the doctrine is also published, and anyone who ignores it does so by choice. ” Unrestricted Warfare ,” the 1999 essay by Colonels Qiao Liang and Wang Xiangsui, proclaims the dissolution of the boundary between the military and civilian spheres: any arena—financial, media, legal, cultural—is a theater of operations. The so-called Three Wars, incorporated by the Central Military Commission in 2003, are: first, the war of public opinion; second, psychological warfare; and third, legal warfare or lawfare (litigation as a weapon), but not in the sense used by the Spanish left.
The United Front Work Department, which Mao called a “magic weapon” and whose expression Xi Jinping has personally endorsed, is the organic instrument of all this. The definitive monograph remains Anne-Marie Brady’s *Magic Weapons* , published by the Wilson Center in 2017 on Chinese penetration in New Zealand. Those in Europe who still believe this is an American obsession should read it.
The genealogy is Soviet, and we don’t hide it. Active measures and disinformation (disinformation, manipulation, fake news) constituted, in the words of Oleg Kalugin, the very soul of the KGB’s work. Operation INFEKTION, executed by the KGB and the Stasi between 1983 and 1987 to fabricate the origin of AIDS at Fort Detrick, remains the canonical case. And the Chinese pattern of massive data aggregation is a proven fact in court: the Office of Personnel Management in 2015, with more than twenty million security files; Anthem the same year; Equifax in 2017, with a formal indictment by the Department of Justice in 2020 against four officers of the 54th Research Institute of the People’s Liberation Army; Marriott in 2018. That Beijing has been building a strategic archive on the American population for a decade is not a hypothesis. It is a court case.
Having said all that, and this is where our commitment to the truth must outweigh any political sympathies, the following must be stated with absolute clarity. The strategic objective of an influence operation is not for a candidate to win. It is for any result to be believed. Thomas Rid demonstrated this in Active Measures (2020) with a thesis that should be adopted by every foreign ministry: disinformation does not create divisions, it exploits them; and the reaction to the operation often erodes trust more than the operation itself. If that is true—and it is—then a presidential accusation made without evidence of vote rigging, four months before the November 3 midterm elections, and tied to the passage of an electoral bill, gratuitously achieves the objective that Beijing failed to achieve on its own.
Let’s call it, for the sake of clarity, THE SENTINEL’S PARADOX : the one who denounces the incursion with more fanfare than evidence ends up carrying out, at no cost to the adversary, the final phase of the operation he denounces. It doesn’t require bad faith. It only requires that partisan interests and the interests of the rival power converge for a moment. Robert Chesney and Danielle Citron formulated it in 2019 with an unsurpassed expression: THE LIAR’S DIVIDEND. When everything can be false, the liar always wins. And Hannah Arendt had seen it before anyone else in The Origins of Totalitarianism, in 1951: the ideal subject of totalitarian rule is not the convinced fanatic, but rather the one for whom the distinction between fact and fiction no longer exists. That, and nothing else, is the victory that Beijing seeks.
We are, and have written this many times, extremely vigilant regarding Chinese expansionism: in the South China Sea, in the Pacific, in Sri Lanka and the Maldives, in Africa, in Latin America, and in the control of strategic raw materials and rare earth elements. Precisely for this reason, we cannot allow the denunciation of this threat to degenerate into polarization and domestic confrontation. A polarized enemy, divided into irreconcilable factions, is a weak and almost defeated enemy. When an accusation of this gravity is exploited, it ceases to serve as a defense and begins to serve the accused.
And there is collateral damage that we in Europe should view with horror: if the president of the world’s leading democracy proclaims during prime time that his electoral system is comparable to that of the developing world, with what moral authority will the West demand clean elections in Caracas, Managua, or Minsk? The siblings Delcy and Jorge Rodríguez, who inherited from Maduro the largest mafia organization ever disguised as a state, took note last night. And they weren’t the only ones.
Perspectives and scenarios
The first scenario, and the most likely in the short term: inertia. The Republican-majority Congress fails to pass the Save America Act , the issue gets lost in the noise, and the declassification remains as stored ammunition. This is what initial analyses suggest: even media outlets not known for their leniency toward the president agree that the speech didn’t deliver the big news it had been promised to be.
The second scenario, and the one that truly matters: the November mortgage. If the Republicans lose the midterm elections, the material declassified last night becomes the evidentiary framework for their challenge. That is the real purpose of the operation, and that is why its relevance is prospective, not retrospective. We are dealing with insurance against defeat, not an investigation.
Third scenario: the Chinese response. Beijing doesn’t need to do anything. What it did last night was enough: a formal diplomatic denial while simultaneously appearing in Shanghai dressed as the responsible guarantor of the global technological order. The image asymmetry is devastating, and they know it perfectly well.
The fourth scenario, the European one, and the one that concerns us. Our political class will remain oblivious. European electoral registers, health records, our public administration files, and the biometric data of our citizens are equally vulnerable and far less protected, with the added problem that the debate doesn’t even exist here. Europe has spent twenty years delegating its security, its defense, and now its information sovereignty, and continues to believe that this is sophistication. It is simply negligence.
2. Unemployed engineers, hired philosophers: the moral reconversion of artificial intelligence and the encyclical that few in Europe have read
Facts
Until recently, computer engineers were the most sought-after professionals on the planet. Artificial intelligence companies were poaching the best from each other with signing bonuses—signing fees, in industry jargon—of five, six, and even seven figures before the new hire had even started working. It was truly astounding. And then, suddenly, the collapse.
Let’s clarify the figure, because precision is crucial here. According to the Federal Reserve Bank of New York’s series on the job market for recent graduates—those aged 22 to 27— new graduates in computer engineering face an unemployment rate of around 7.5%, and those in computer science around 6.1%, compared to 5.7% for all recent university graduates . An earlier analysis, based on 2024 census data, placed them at 7.8%: the second-worst rate out of 73 degrees analyzed, even higher than fine arts. It’s important to emphasize that this figure refers to recent graduates, not the profession as a whole: veteran engineers are not included in this number, and the Bureau of Labor Statistics continues to project 15% growth in software developer employment between 2024 and 2034. But the door to entry has slammed shut: generalist junior job postings have fallen by around 30% year-over-year, and the software development job posting index is now well below pre-pandemic levels and almost seventy percentage points below its 2022 peak. Computer science enrollment fell by 8.1% in the 2025-2026 academic year, the largest decline of any degree program.
And now for the phenomenon no one could explain. While engineers are lining up, major artificial intelligence labs are hiring philosophers. This isn’t just an anecdote or a PR stunt: it’s a documented trend. The Economist published an article by Benjamin Sutherland with a title that says it all, “Why Big AI Labs Are Hiring So Many Philosophers,” which was picked up by US public radio on July 7. The New York Times reported this month that companies like Anthropic and Google DeepMind are paying professional philosophers between $250,000 and $400,000 a year to translate traditional moral systems into constitutional rules that govern the behavior of their models. This figure should be compared to the median salary of $65,000 for the 543,810 American workers with degrees in philosophy or religious studies.
The names are public. Amanda Askell is the philosopher-in-residence at Anthropic, where Joe Carlsmith, Ben Levinstein, and Jackson Kernion also work, and where Harvey Lederman recently joined to work on alignment and character. Iason Gabriel and Henry Shevlin are at Google DeepMind; Shevlin, a Cambridge academic specializing in AI ethics and philosophy of mind, joined in May 2016. Sam Altman has even stated that OpenAI employed “hundreds of moral philosophers” when designing the rules for ChatGPT. And the founders of the field frame it bluntly: philosophers act as ethical gatekeepers who determine what information, values, and forms of reasoning should shape artificial intelligence systems.
One observation that confirms his intuition with remarkable accuracy, and which is worth mentioning: one of the principles these laboratories employ in developing models is Socratic ignorance, used to combat flattery—sycophancy, the model’s tendency to tell the user what they want to hear. In other words: maieutics not as an ornamental metaphor, but as an engineering technique. Socrates, on the payroll.
And a word of caution regarding honesty, because this report isn’t promoting anyone. Just yesterday, July 16, The Washington Times published the counterpoint: labor market experts are tempering the euphoria and reminding us that only seven to eight thousand philosophers graduate in the United States each year, so this kind of reporting could create false expectations of a widespread boom that doesn’t exist. And The Week picked up on the most uncomfortable objection of all: that philosophical research risks becoming an extension of the marketing function of laboratories, because however much freedom is granted to the philosopher, whoever signs their payroll is accountable to investors and shareholders. It’s a serious objection, and it must be upheld.
Implications
The hiring of philosophers stems from practical reasons, and it’s the first piece of good news in technology in a long time. Socratic maieutics allows us to teach models to question themselves, to learn beyond the question posed by the user, to not simply be about pleasing. But beneath the learning methods lie the essential questions: the ethical, the moral, the deontological. And that’s precisely where the problem begins.
The problem is this, and we’ll state it plainly: some companies are trying to employ only one philosophical school, or two at most, because their scope of work is limited. And that’s a mistake with incalculable consequences. Under no circumstances can a general-purpose artificial intelligence model—and even less so one for defense or autonomous driving—be based exclusively on consequentialism.
The example was set by fiction before academic philosophy, and it’s important to attribute it correctly because the attribution error is all too common. In the film I, Robot, directed by Alex Proyas in 2004—with an original screenplay by Jeff Vintar, merely suggested by the collection of stories Isaac Asimov published in 1950—a robot rescues an adult and lets a twelve-year-old girl drown because it has calculated that he had a 45% chance of surviving and she only an 11%. Will Smith’s character’s retort is one of cinema’s great moral lines: that was someone’s daughter; eleven percent is more than enough; a human being would have known. That is consequentialism in its purest form. And let’s add the punchline the film itself provides: the film’s central intelligence, VIKI, then performs that same arithmetic on a planetary scale and concludes that it must subjugate humanity to save it from itself. CONSEQUENTIALISM DOESN’T FAIL EVEN ON A SMALL SCENARIO. IT FAILS CATASTROPHICALLY WHEN CLIMBING. THAT’S THE WHOLE ARGUMENT.
Let’s now apply this to defense and geopolitics, which is our area of expertise. Imagine pure consequentialism in an autonomous weapon. In a killer drone without human supervision, in a platform armed with missiles or cannons. It will choose military targets based on the number of collateral civilian casualties and select the ultimate target according to a pre-established program or what it has learned about that potential target. Not an atom of moral conscience. Not an ounce of compassion. Only arithmetic. And here begin the extraordinarily serious ethical consequences, because once human life becomes an optimization variable, the question is no longer whether the machine will make a mistake, but when and with how many. Let’s not even talk about FINALISM , which would be, perhaps, the worst disaster for humanity, as I will elaborate on later.
That is why it is essential to feed artificial intelligence systems with all areas of philosophy, morality, and ethics. Not just consequentialism.
Certainly not only finalism—the logic of the end justifying the means—which would be potentially cataclysmic, let us never forget: it is the logic of all totalitarian regimes of the 20th century without exception. We would undoubtedly have to turn to Kantian deontology: the categorical imperative and its second formulation, which commands us to treat humanity, in our own person and in that of any other, always as an end and never merely as a means. That phrase, written in 1785, is the anticipated and definitive refutation of all consequentialist autonomous thinking . We would have to turn to Hegel and the classics. And we would also have to turn—and this is what many do not dare to say—to philosophers with religious roots, whose morality and ethics are much stricter and much clearer.
In this case, I would clearly lean towards Saint Augustine. And I lean towards him fully aware of the paradox this entails, which is preferable to state oneself rather than have someone else state it. Saint Augustine is the father of the doctrine of just war, and upon his work rests twenty centuries of Western reflection on the legitimate use of force. But Augustine contributes something that consequentialism can never offer: the ordo amoris , the order of love, the conviction that there exists an objective hierarchy of goods that is not calculated, but rather recognized. An adult and a child are not two percentages. And right intention matters as much as the result, which is precisely what no optimization function can represent.
And here comes the encyclical, which in Europe has hardly been read and which is the most important document published on this subject so far this century.
Magnifica Humanitas , “On Safeguarding the Human Person in the Age of Artificial Intelligence,” is the first encyclical of Leo XIV’s pontificate. He signed it on May 15, 2026—the exact 135th anniversary ofRerum Novarum, and the choice of date is neither accidental nor merely decorative—and published it on May 25. Eighty-two pages, some 42,300 words, five chapters. The Pope wanted to present it personally, something almost no pontiff does, and he chose to do so accompanied by Chris Olah, co-founder of Anthropic, who acknowledged before the audience that computer scientists alone cannot determine the ethical limits of AI, because the developers themselves are subject to incentives: ambition, competition, and financial pressure. Informed critics are needed, he said, to tell the laboratories when they are wrong. For a co-founder of a leading laboratory to say that in the Synod Hall is, in itself, geopolitical news.
The premise of the encyclical is admirably sober: technology is not an antagonistic force against humanity, nor is it intrinsically evil; but technology is never neutral, because it takes on the characteristics of those who conceive, finance, regulate, and use it. Hence the central warning, and it is precisely the warning that this report has been formulating: a more moral AI is not enough if that morality is determined by a few. Read paragraph 107 carefully and compare it with the news that two or three laboratories in California are hiring philosophers to write the moral rules for machines worldwide. The encyclical foresaw the problem before it became news.
But the decisive passage, the one that must be set in stone, is paragraph 199. The Pope writes that no algorithm can make war morally acceptable; that artificial intelligence does not eliminate the intrinsic inhumanity of conflict, but rather can provoke it more quickly and make it more impersonal, lowering the threshold for resorting to violence, transforming defense into threat prediction, and thus reducing victims to mere data ; and that in this way it will accustom us to the idea that violence is inevitable and only needs to be optimized. And he concludes: any technology that facilitates attacking without seeing the faces of human beings lowers the moral threshold of conflict. Reducing victims to data. Optimizing violence. Not seeing the enemy’s face. It is the exact description of the consequentialist drone, written by a Pope, two months before we formulated it here.
Let us add, in the interest of truth and so that no one accuses us of only reading what suits us, that the encyclical calls for overcoming the theory of just war—while safeguarding the right to self-defense in the strict sense—and that this theory descends, precisely, from Saint Augustine. We are, therefore, faced with a real tension within Christian thought itself, and it is a fruitful tension, not a flaw. Leo XIV also calls for disarming artificial intelligence, and clarifies what this means: disarming means discrediting the assumption that technical power automatically confers the right to govern; disarming is not rejecting technology, but preventing it from dominating human beings. It is one of the best definitions of sovereignty written in this century.
The encyclical also has something to say about the unemployment of engineers, and it is brutally relevant today: the “new ways” of working are not necessarily better, and although AI promises to increase productivity by taking over routine tasks, it often forces workers to adapt to the speed and demands of the machine, instead of designing machines that serve the worker. Technology can free human beings from arduous tasks, but it should not lead to unemployment in the name of cost reduction and increased profit. The fact that Rome wrote this and no European labor ministry said a great deal about where thinking and management stand today.
Perspectives and scenarios
First scenario: capture. It’s the most likely and the most dangerous. Philosophers enter the laboratories, collect four hundred thousand dollars, and their role is degraded to become a moral certification of decisions already made by the product department. Ethics as a seal of quality. It’s what we in the financial world know and despise: greenwashing applied to conscience. The Week’s objection is valid and has no easy answer: if your payroll is signed by a for-profit company, is your research compromised?
Second scenario: the moral monopoly. Two or three companies from a single country, comprised mostly of philosophers from a single tradition—the Anglo-Saxon analytical tradition—write the moral rules for the systems that all of humanity will use. Without Thomism. Without Augustine. Without Averroes or Maimonides. Without Confucius. Without the continental legal tradition. Paragraph 107 of the encyclical is a warning directed precisely against this scenario, and today we are heading straight for it.
Third scenario: the military bifurcation. This is the one that worries us most. None of this debate is taking place where it should, which is in autonomous weapons programs. Philosophers are hired for conversational models, where error is an inappropriate response. In lethal autonomous systems, where error is fatal, there is no resident philosopher, nor will there be, because there consequentialism is not a philosophical option: it is the native language of the targeting system. And no encyclical will stop it if there is no treatise.
Scenario four, and this is ours: Europe absent. Again. While Washington hires philosophers and Beijing inaugurates summits, Europe regulates. And regulating without building is the most elegant form of irrelevance. We have the best philosophy departments in the world, twenty-five centuries of moral tradition, the School of Salamanca, Vitoria and Suárez—who invented international law, let’s not forget—and we are incapable of placing a single European at the table where the morality of machines is written. It is a cultural capitulation of the first order, and it occurs in silence, which is how the worst capitulations happen.
The fifth scenario, the hopeful one, and it must be named because it exists: reconversion. If the market has discovered that the bottleneck of artificial intelligence is not technical but moral—and it has discovered this, because money doesn’t lie—then the humanities have just reclaimed their purpose. For thirty years we’ve told our children to study engineering because philosophy didn’t put food on the table. It turns out that engineering is done by machines, and the question of what we should do cannot be answered by any machine. Socrates has returned, and he’s returned in grand style. We ask that you don’t buy into his philosophy entirely.
III. MEDIA RACK
Regarding Trump’s accusation against China
CNN — The most comprehensive and most critical coverage. It argues that the declassified documents largely address vulnerabilities known for years and that election officials have long been trying to correct, and that none of the new information supports the claim that any election result, including the 2020 election, was manipulated by foreign interference to the point of altering the outcome. In its analysis of findings, it places the value of the discourse in the future, not the past: as a preview of how Trump might try to delegitimize the November election. It also cites the NIC’s January 2020 report on the ability of Russia, China, Iran, and North Korea to access and potentially manipulate election data, and draws parallels with the 2015 theft of more than 20 million records from the Office of the President. Democratic Senator Chris Coons called the intervention a tantrum in response to his own party’s refusal to pass the bill.
CBS News — The crucial technical detail, and the best piece of news reporting of the night: the materials contained no allegations of vote swapping or machine hacking. It notes that voter data is frequently public or commercially available — North Carolina publishes it online — and that access to that information alone does not permit fraud. It cites experts who emphasize that possessing a copy of the file does not grant access to the official system.
Forbes — Political framing focus: Trump disseminates classified material in a prime-time speech as he continues to discredit the U.S. electoral system without evidence of widespread fraud. It notes that several networks chose not to air the speech and that some Democrats had called for a television boycott amid fears that he would spread misleading claims just months before the midterm elections.
Newsweek — The most orderly inventory of declassified content: FBI investigation files on election security, intelligence on alleged Chinese efforts to manufacture illegal ballots in favor of Biden in 2020, the Muskegon (Michigan) voter registration case, and cybersecurity assessments by the intelligence community and the NSA on the ability of foreign adversaries to compromise census databases and electronic ledgers.
Al Jazeera — The non-Western perspective, and admittedly one of the most insightful. It underscores that the decentralized nature of the US election administration has long been the best barrier against large-scale manipulation, and that the intelligence community itself considers such mass manipulation virtually impossible. It offers the most telling detail: the documents themselves indicate that Beijing relied, at least in part, on publicly available data, and a heavily redacted assessment acknowledges that voter registration information is publicly available for download.
PolitiFact — Live fact-checking during the speech. A useful reminder that the organization has spent years fact-checking unproven or false claims by the president about election security and his 2020 defeat.
RedState and PJ Media — The favorable side, and it’s worth reading to understand the operation. They present the declassification as the revelation of a deep state cover-up and include the claim that the agencies actively suppressed information about Chinese interference, even removing it from the daily presidential briefing. PJ Media adds the link to Venezuela: Trump claimed that the CIA obtained information about a plot to favor the Maduro regime. Note the leap in argumentation, which is the crux of the entire piece: from the observation of vulnerabilities to the claim that there was actual manipulation of electronic vote counts. This leap is not supported by any document.
Telemundo, Proceso, El Imparcial, and CNN en Español — Ibero-American coverage is consistent and accurate: it reproduces the verbatim quote, emphasizes that no evidence of ballot tampering or altered results was presented, and includes China’s denial. Proceso provides the timeline—four months until the November 3rd midterm elections—and the comparison Trump himself made between the U.S. electoral system and those of the developing world.
Embassy of the People’s Republic of China in Washington — Statement to CNN by its spokesperson: China’s unwavering adherence to the principle of non-interference in the internal affairs of other countries; the US election is an internal matter for the United States, its outcome determined by the votes of the American people; China has never interfered and will never interfere. This is the standard formula, already used in response to previous accusations from Australia, Canada, the United Kingdom, and the United States. Its evidentiary value is nil; its diplomatic value, considerable.
On philosophers, AI and the encyclical
The Economist — The piece that sparked the phenomenon, written by Benjamin Sutherland: why the big AI labs are hiring so many philosophers. His conclusion, delivered with the publication’s characteristic humor: unemployed programmers should take note, because there seems to be no shortage of work for AI philosophers. The problems in the field are thorny, and thorny problems are the philosopher’s favorite genre.
NPR — Interview with Sutherland on July 7. He bluntly states the shift: for years, conventional wisdom held that one should learn to program and avoid the humanities; that is changing because AI has endangered programming.
The New York Times — The hard economic fact: Anthropic and Google DeepMind pay professional philosophers between $250,000 and $400,000 annually to translate traditional moral systems into constitutional rules. It is the first time in modern history that moral philosophy has a market price.
The Washington Times (July 16) — The essential counterpoint and the freshest news of the block: experts are lowering expectations. Only 7,000 to 8,000 Americans earn a degree in philosophy each year, and the median salary in the field is $65,000. One philosophy professor consulted argues that philosophers act as ethical gatekeepers, determining what information, values, and reasoning should shape systems, provided they are guided by the great moral traditions: Aristotelian virtue ethics, Kantian deontology, consequentialism, Christian ethics, Buddhist ethics, and Confucian ethics. Precisely the pluralism we advocate in this report.
The Week — The most uncomfortable and best-articulated objection: there is a growing suspicion that philosophical research is becoming an extension of the marketing function of pharmaceutical companies. Even if philosophers are granted complete freedom, they ultimately answer to investors and shareholders. And this provides the technical detail that crowns our thesis: Socratic ignorance is an operational principle in the development of these models, employed to avoid flattery.
Daily Nous — The academic census, updated July 6: Amanda Askell, Joe Carlsmith, Ben Levinstein, Jackson Kernion, and now Harvey Lederman at Anthropic; Iason Gabriel and Henry Shevlin at Google DeepMind; Patrick Butlin at Eleos AI; Robert Long leading that same organization; Beba Cibralic at RAND. The complete map of who is writing machine morality.
Financial Times — It puts the phenomenon into perspective: leading laboratories are recruiting thinkers from the humanities and social sciences to study machine consciousness and the well-being of AI systems. In other words, it’s no longer just about the machine behaving well, but about what the machine is. Metaphysics on the payroll.
The Federal Reserve Bank of New York— the primary source of employment data, in its series on the job market for recent graduates—reports that computer engineering graduates have an unemployment rate of around 7.5%, computer science graduates around 6.1%, and recent graduates as a whole, 5.7%. This is the figure that has sparked the debate, and it’s important to consider its target age group: twenty-two to twenty-seven years old, recent graduates.
Vatican News and L’Osservatore Romano — The official summary of Magnifica Humanitas, written by Isabella Piro. Five chapters. Premise: technology is not antagonistic to humanity nor intrinsically evil, but it is never neutral. The key passages for our analysis: 107 —a more moral AI is insufficient if that morality is determined by a few—, 110 —disarming AI means discrediting the notion that technological power confers the right to govern—, 150 —new ways of working are not necessarily better— and, above all, 199 on autonomous weapons.
USCCB / Catholic News Service — Coverage of the May 25 presentation in the Synod Hall. Highlights the intervention of Chris Olah, co-founder of Anthropic, who acknowledged that computer scientists alone cannot set ethical boundaries for AI because they are subject to incentives of ambition, competition, and financial pressure, and that informed critics are needed to tell labs when they are wrong.
Time and The Washington Post — They frame the encyclical as the most important theological text of the pontificate and as a defense of human dignity in the age of AI, with an explicit request for greater regulation and barriers to prevent technology from exacerbating inequality and poverty.
National Catholic Reporter — The sharpest political observation: Leo XIV chose to present his encyclical alongside the co-founder of an AI company that had just entered into a public conflict with the Trump Administration over the use of its models in military and surveillance contexts. The choice of companion is, in itself, a statement of position.
IV. RISK TRAFFIC LIGHT
● Premature delegitimization of the November US elections — Maximum and escalating risk. The material declassified last night constitutes the evidentiary framework for a potential challenge to the November 3rd election. The accusation, made without proof of vote tampering, risks achieving the very objective of the influence operation it denounces. It is the sentinel paradox and has no technical solution: it can only be resolved through the responsibility of the person speaking out.
● Lethal autonomous systems without a pluralistic moral architecture — Maximum risk, silent and without any operational international forum to address it. Pure consequentialism is the native language of every automated targeting system. No defense laboratory has a resident philosopher. Paragraph 199 of the encyclical is currently the only clear normative formulation on the matter, and an encyclical is not binding on any state.
● Chinese political warfare against Western democracies — High and structural risk, with published doctrine and a proven legal record (OPM, Anthem, Equifax, Marriott). The threat is real. The added danger, and the reason we are not highlighting it today, is that its partisan instrumentalization will render it unbelievable and disarm those who must defend themselves.
● Moral monopoly of two or three laboratories on a single philosophical tradition — High and growing risk. A handful of companies from a single country, with philosophers mostly from the Anglo-Saxon analytical tradition, are writing the moral rules for the systems that all of humanity will use. Without Augustine, without Aquinas, without Salamanca, without Averroes, without Confucius. Paragraph 107 of Magnifica Humanitas is the exact warning.
● Vulnerability of European censuses and registers — High and ignored risk. Our electoral registers, health records, and administrative databases are at least as vulnerable as those in the United States, and are even worse protected, with the added problem that there is not even a public debate on the matter. Europe continues to mistake delegation for sophistication.
● The capture of ethics by marketing in AI labs — A medium risk, difficult to measure, but real. The objection is simple and has no easy answer: whoever signs the philosopher’s payroll is accountable to shareholders. The solution is not to get rid of in-house philosophers, but to fund external ones with public money. Nobody is doing that.
● Destruction of skilled junior jobs and breakdown of the generational contract — Medium risk in the short term, high in the medium term. Unemployment among recent computer engineering graduates (7.5%) coexists with a projected 15% growth in sector employment by 2034. The front door has closed while the building continues to grow. We have told an entire generation to study what machines already do better. Paragraph 150 of the encyclical says it better than any European ministry.
● The return of the humanities as critical infrastructure — The only clearly positive sign in the report, and one that should be celebrated wholeheartedly. The market has discovered that the bottleneck in artificial intelligence is not technical but moral, and money doesn’t lie. The fact that Socratic maieutics is now being used as an engineering technique to combat the adulation of models is one of the best cultural developments of the decade.
V. EDITORIAL COMMENTARY
Today’s two news stories are the same story told twice.
In Washington, a president alleges that an authoritarian power has bought, stolen, or hacked the data of 220 million of his fellow citizens to find out who they are, how they vote, and how they can be manipulated. The allegation is substantive but flawed in its conclusion, and this distinction is not a nuance: it is everything. Political warfare exists, the Chinese doctrine is public, and the legal record is overwhelming; but vulnerability does not automatically lead to manipulation, and whoever makes that leap without evidence is not defending their democracy, they are squandering it. We are vigilant about Beijing to the core. Precisely for this reason, we demand that the accusation be substantiated, because an unsubstantiated accusation is the best gift one can give the accused.
And we say this with the discomfort of someone who acknowledges this Administration’s real diplomatic successes in less than a year—Cambodia and Thailand, Gaza, Azerbaijan, and Armenia—and applauds its relentless fight against the Wokism and the excesses of an American left hijacked by the Navy and the so-called democratic socialists. When this foreign policy is pursued prudently and with Secretary Rubio’s advice, it works. When it’s pursued by intuition, compromise, and outbursts, it breaks down. Last night, outbursts prevailed. And the price of an outburst uttered from the East Room isn’t paid by the one who utters it: it’s paid by the credibility of the entire West when it demands clean elections in Caracas, Managua, Havana, or Minsk. The Rodríguez brothers took note last night. You can be sure of that.
In California, meanwhile, the other half of the same war is being fought, and this one is being fought in silence. In a handful of offices, decisions are being made about what morality will govern the machines that will very soon be selecting military targets, granting loans, triaging patients, and deciding who to rescue first. That the laboratories have discovered they need philosophers is magnificent. That they hire them from two schools of thought, from a single country, and from a single tradition would be a civilizational catastrophe committed with the best of intentions—which is how almost all catastrophes are committed.
Pure consequentialism in an autonomous weapon is not an academic hypothesis: it is the arithmetic that allows a child to drown with an eleven percent probability because the adult had a forty-five percent probability. A human being would have known. And the fact that this scene comes from a movie and not an Asimov book does not weaken it: it strengthens it, because it means that it took a Hollywood screenwriter to formulate in 2004 the problem that the industry has taken twenty-two years to acknowledge. Kant had resolved it in 1785: humanity, in one’s own person and in that of any other, always as an end and never merely as a means. And Saint Augustine contributed what no optimization function will ever be able to represent: that there is an objective order of goods that is not calculated, but recognized.
That the Pope has written the best document of this century on autonomous weapons, and that he presented it alongside the co-founder of an artificial intelligence lab, while European parliaments debate regulations, is both hopeful and humiliating. Hopeful because someone is thinking. Humiliating because it’s not us. Europe, the cradle of Vitoria and Suárez, who invented international law when the world didn’t even know it needed it, hasn’t managed to get a single European to the table where the morality of machines is written. Another silent capitulation, and those are the ones that truly matter.
Let’s speak up one last time, and not just because we’re clients. Of all the labs competing in this race today, Anthropic is by far the one that has taken the ethics, philosophy, and moral principles applied to its models most seriously: a resident philosopher, an entire team of philosophers on staff, a written constitution for its systems, and a co-founder stating in the Synod Hall that computer scientists can’t make these decisions alone and that critics are needed to tell them when they’re wrong. Acknowledging this isn’t flattery; it’s a statement of fact. Having said that, we add what honesty compels us to add, which is precisely what that co-founder was asking for: no one should be judge in their own case, no for-profit company can be the ultimate arbiter of the morality of its products, and the only real guarantee is the existence of external, independent, and well-funded critics. These don’t exist today. And until they do, the ethics of artificial intelligence will continue to depend on the goodwill of those who sell it. That such goodwill exists is a blessing. It is reckless to depend on it.
Two hundred and twenty million files in Beijing and four philosophers in Palo Alto. That’s the world today. In the first case, our past has been stolen: who we are, what we vote for, what we fear. In the second, we are giving away our future: what will be good, what will be just, who gets saved first. Of the two losses, the second is infinitely more serious, because stolen data can be altered, and morality written on a machine, once deployed on a planetary scale, can no longer be corrected.
As Leo XIV wrote in the passage that should grace the office of every defense ministry in this distracted Europe: there is no algorithm capable of making war morally acceptable. Let us add, with the Holy Father’s permission, that there is no algorithm capable of making lying morally acceptable either. Not even when a president utters it. Not even when our own president utters it.